int main(int argc,
char **argv) {
char buf[32];
char *filein, *fileout, *fontdir, *textstr;
l_int32 n, i, maxdepth, ntext, border, lossless, display, showtext;
l_float32 scalefact;
L_BMF *bmf;
PIX *pix1, *pix2, *pix3, *pix4, *pixd;
PIXA *pixa, *pixad;
static char mainName[] = "displaypixa";
if (argc != 3 && argc != 4 && argc != 7 && argc != 8) {
fprintf(stderr, "Syntax error in displaypixa:\n"
" displaypixa filein fileout [showtext]\n"
" displaypixa filein scalefact border"
" lossless disp fileout [showtext]\n");
return 1;
}
filein = argv[1];
if ((pixa = pixaRead(filein)) == NULL)
return ERROR_INT("pixa not made", mainName, 1);
pixaCountText(pixa, &ntext);
if (argc == 3 || argc == 4)
fileout = argv[2];
if (argc == 4)
showtext = atoi(argv[3]);
/* Simple specification; no output text */
if (argc == 3 ||
(argc == 4 && (ntext == 0 || showtext == 0))) { /* no text output */
pixaVerifyDepth(pixa, &maxdepth);
pixd = pixaDisplayTiledInRows(pixa, maxdepth, 1400, 1.0, 0, 10, 0);
pixDisplay(pixd, 100, 100);
if (pixGetDepth(pixd) == 1)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
return 0;
}
/* Simple specification with output text */
if (argc == 4) { /* showtext == 1 && ntext > 0 */
n = pixaGetCount(pixa);
bmf = bmfCreate(NULL, 6);
pixad = pixaCreate(n);
for (i = 0; i < n; i++) {
pix1 = pixaGetPix(pixa, i, L_CLONE);
pix2 = pixConvertTo32(pix1);
pix3 = pixAddBorderGeneral(pix2, 10, 10, 5, 5, 0xffffff00);
textstr = pixGetText(pix1);
if (textstr && strlen(textstr) > 0) {
snprintf(buf, sizeof(buf), "%s", textstr);
pix4 = pixAddSingleTextblock(pix3, bmf, buf, 0xff000000,
L_ADD_BELOW, NULL);
} else {
pix4 = pixClone(pix3);
}
pixaAddPix(pixad, pix4, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
}
bmfDestroy(&bmf);
pixaVerifyDepth(pixad, &maxdepth);
pixd = pixaDisplayTiledInRows(pixad, maxdepth, 1400, 1.0, 0, 10, 0);
pixDisplay(pixd, 100, 100);
if (pixGetDepth(pixd) == 1)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
pixaDestroy(&pixad);
return 0;
}
/* Full specification */
scalefact = atof(argv[2]);
border = atoi(argv[3]);
lossless = atoi(argv[4]);
display = atoi(argv[5]);
fileout = argv[6];
showtext = (argc == 8) ? atoi(argv[7]) : 0;
if (showtext && ntext == 0)
L_INFO("No text found in any of the pix\n", mainName);
bmf = (showtext && ntext > 0) ? bmfCreate(NULL, 6) : NULL;
n = pixaGetCount(pixa);
pixad = pixaCreate(n);
for (i = 0; i < n; i++) {
pix1 = pixaGetPix(pixa, i, L_CLONE);
pix2 = pixConvertTo32(pix1);
pix3 = pixAddBorderGeneral(pix2, 10, 10, 5, 5, 0xffffff00);
textstr = pixGetText(pix1);
if (bmf && textstr && strlen(textstr) > 0) {
snprintf(buf, sizeof(buf), "%s", textstr);
pix4 = pixAddSingleTextblock(pix3, bmf, buf, 0xff000000,
L_ADD_BELOW, NULL);
} else {
pix4 = pixClone(pix3);
}
pixaAddPix(pixad, pix4, L_INSERT);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
}
bmfDestroy(&bmf);
pixaVerifyDepth(pixad, &maxdepth);
pixd = pixaDisplayTiledInRows(pixad, maxdepth, 1400, scalefact,
0, 10, border);
if (display) pixDisplay(pixd, 20, 20);
if (pixGetDepth(pixd) == 1 || lossless)
pixWrite(fileout, pixd, IFF_PNG);
else
pixWrite(fileout, pixd, IFF_JFIF_JPEG);
pixDestroy(&pixd);
pixaDestroy(&pixa);
pixaDestroy(&pixad);
return 0;
}
int main(int argc,
char **argv)
{
char *filein, *fileout;
l_int32 ret;
l_float32 deg2rad;
l_float32 angle, conf, score;
PIX *pix, *pixs, *pixd;
static char mainName[] = "skewtest";
if (argc != 3)
return ERROR_INT(" Syntax: skewtest filein fileout", mainName, 1);
filein = argv[1];
fileout = argv[2];
setLeptDebugOK(1);
pixd = NULL;
deg2rad = 3.1415926535 / 180.;
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
/* Find the skew angle various ways */
pix = pixConvertTo1(pixs, 130);
pixWrite("/tmp/binarized.tif", pix, IFF_TIFF_G4);
pixFindSkew(pix, &angle, &conf);
fprintf(stderr, "pixFindSkew():\n"
" conf = %5.3f, angle = %7.3f degrees\n", conf, angle);
pixFindSkewSweepAndSearchScorePivot(pix, &angle, &conf, &score,
SWEEP_REDUCTION2, SEARCH_REDUCTION,
0.0, SWEEP_RANGE2, SWEEP_DELTA2,
SEARCH_MIN_DELTA,
L_SHEAR_ABOUT_CORNER);
fprintf(stderr, "pixFind...Pivot(about corner):\n"
" conf = %5.3f, angle = %7.3f degrees, score = %f\n",
conf, angle, score);
pixFindSkewSweepAndSearchScorePivot(pix, &angle, &conf, &score,
SWEEP_REDUCTION2, SEARCH_REDUCTION,
0.0, SWEEP_RANGE2, SWEEP_DELTA2,
SEARCH_MIN_DELTA,
L_SHEAR_ABOUT_CENTER);
fprintf(stderr, "pixFind...Pivot(about center):\n"
" conf = %5.3f, angle = %7.3f degrees, score = %f\n",
conf, angle, score);
/* Use top-level */
pixd = pixDeskew(pixs, 0);
pixWriteImpliedFormat(fileout, pixd, 0, 0);
#if 0
/* Do it piecemeal; fails if outside the range */
if (pixGetDepth(pixs) == 1) {
pixd = pixDeskew(pix, DESKEW_REDUCTION);
pixWrite(fileout, pixd, IFF_PNG);
}
else {
ret = pixFindSkewSweepAndSearch(pix, &angle, &conf, SWEEP_REDUCTION2,
SEARCH_REDUCTION, SWEEP_RANGE2,
SWEEP_DELTA2, SEARCH_MIN_DELTA);
if (ret)
L_WARNING("skew angle not valid\n", mainName);
else {
fprintf(stderr, "conf = %5.3f, angle = %7.3f degrees\n",
conf, angle);
if (conf > 2.5)
pixd = pixRotate(pixs, angle * deg2rad, L_ROTATE_AREA_MAP,
L_BRING_IN_WHITE, 0, 0);
else
pixd = pixClone(pixs);
pixWrite(fileout, pixd, IFF_PNG);
pixDestroy(&pixd);
}
}
#endif
pixDestroy(&pixs);
pixDestroy(&pix);
pixDestroy(&pixd);
return 0;
}
/*!
* pixRunlengthTransform()
*
* Input: pixs (1 bpp)
* color (0 for white runs, 1 for black runs)
* direction (L_HORIZONTAL_RUNS, L_VERTICAL_RUNS)
* depth (8 or 16 bpp)
* Return: pixd (8 or 16 bpp), or null on error
*
* Notes:
* (1) The dest Pix is 8 or 16 bpp, with the pixel values
* equal to the runlength in which it is a member.
* The length is clipped to the max pixel value if necessary.
* (2) The color determines if we're labelling white or black runs.
* (3) A pixel that is not a member of the chosen color gets
* value 0; it belongs to a run of length 0 of the
* chosen color.
* (4) To convert for maximum dynamic range, either linear or
* log, use pixMaxDynamicRange().
*/
PIX *
pixRunlengthTransform(PIX *pixs,
l_int32 color,
l_int32 direction,
l_int32 depth) {
l_int32 i, j, w, h, wpld, bufsize, maxsize, n;
l_int32 *start, *end, *buffer;
l_uint32 *datad, *lined;
PIX *pixt, *pixd;
PROCNAME("pixRunlengthTransform");
if (!pixs)
return (PIX *) ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (PIX *) ERROR_PTR("pixs not 1 bpp", procName, NULL);
if (depth != 8 && depth != 16)
return (PIX *) ERROR_PTR("depth must be 8 or 16 bpp", procName, NULL);
pixGetDimensions(pixs, &w, &h, NULL);
if (direction == L_HORIZONTAL_RUNS)
maxsize = 1 + w / 2;
else if (direction == L_VERTICAL_RUNS)
maxsize = 1 + h / 2;
else
return (PIX *) ERROR_PTR("invalid direction", procName, NULL);
bufsize = L_MAX(w, h);
if ((pixd = pixCreate(w, h, depth)) == NULL)
return (PIX *) ERROR_PTR("pixd not made", procName, NULL);
datad = pixGetData(pixd);
wpld = pixGetWpl(pixd);
if ((start = (l_int32 *) CALLOC(maxsize, sizeof(l_int32))) == NULL)
return (PIX *) ERROR_PTR("start not made", procName, NULL);
if ((end = (l_int32 *) CALLOC(maxsize, sizeof(l_int32))) == NULL)
return (PIX *) ERROR_PTR("end not made", procName, NULL);
if ((buffer = (l_int32 *) CALLOC(bufsize, sizeof(l_int32))) == NULL)
return (PIX *) ERROR_PTR("buffer not made", procName, NULL);
/* Use fg runs for evaluation */
if (color == 0)
pixt = pixInvert(NULL, pixs);
else
pixt = pixClone(pixs);
if (direction == L_HORIZONTAL_RUNS) {
for (i = 0; i < h; i++) {
pixFindHorizontalRuns(pixt, i, start, end, &n);
runlengthMembershipOnLine(buffer, w, depth, start, end, n);
lined = datad + i * wpld;
if (depth == 8) {
for (j = 0; j < w; j++)
SET_DATA_BYTE(lined, j, buffer[j]);
} else { /* depth == 16 */
for (j = 0; j < w; j++)
SET_DATA_TWO_BYTES(lined, j, buffer[j]);
}
}
} else { /* L_VERTICAL_RUNS */
for (j = 0; j < w; j++) {
pixFindVerticalRuns(pixt, j, start, end, &n);
runlengthMembershipOnLine(buffer, h, depth, start, end, n);
if (depth == 8) {
for (i = 0; i < h; i++) {
lined = datad + i * wpld;
SET_DATA_BYTE(lined, j, buffer[i]);
}
} else { /* depth == 16 */
for (i = 0; i < h; i++) {
lined = datad + i * wpld;
SET_DATA_TWO_BYTES(lined, j, buffer[i]);
}
}
}
}
pixDestroy(&pixt);
FREE(start);
FREE(end);
FREE(buffer);
return pixd;
}
/*!
* pixGenerateSelRandom()
*
* Input: pix (1 bpp, typically small, to be used as a pattern)
* hitfract (fraction of allowable fg pixels that are hits)
* missfract (fraction of allowable bg pixels that are misses)
* distance (min distance from boundary pixel; use 0 for default)
* toppix (number of extra pixels of bg added above)
* botpix (number of extra pixels of bg added below)
* leftpix (number of extra pixels of bg added to left)
* rightpix (number of extra pixels of bg added to right)
* &pixe (<optional return> input pix expanded by extra pixels)
* Return: sel (hit-miss for input pattern), or null on error
*
* Notes:
* (1) Either of hitfract and missfract can be zero. If both are zero,
* the sel would be empty, and NULL is returned.
* (2) No elements are selected that are less than 'distance' pixels away
* from a boundary pixel of the same color. This makes the
* match much more robust to edge noise. Valid inputs of
* 'distance' are 0, 1, 2, 3 and 4. If distance is either 0 or
* greater than 4, we reset it to the default value.
* (3) The 4 numbers for adding rectangles of pixels outside the fg
* can be use if the pattern is expected to be surrounded by bg
* (white) pixels. On the other hand, if the pattern may be near
* other fg (black) components on some sides, use 0 for those sides.
* (4) The input pix, as extended by the extra pixels on selected sides,
* can optionally be returned. For debugging, call
* pixDisplayHitMissSel() to visualize the hit-miss sel superimposed
* on the generating bitmap.
*/
SEL *
pixGenerateSelRandom(PIX *pixs,
l_float32 hitfract,
l_float32 missfract,
l_int32 distance,
l_int32 toppix,
l_int32 botpix,
l_int32 leftpix,
l_int32 rightpix,
PIX **ppixe)
{
l_int32 ws, hs, w, h, x, y, i, j, thresh;
l_uint32 val;
PIX *pixt1, *pixt2, *pixfg, *pixbg;
SEL *seld, *sel;
PROCNAME("pixGenerateSelRandom");
if (ppixe) *ppixe = NULL;
if (!pixs)
return (SEL *)ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (SEL *)ERROR_PTR("pixs not 1 bpp", procName, NULL);
if (hitfract <= 0.0 && missfract <= 0.0)
return (SEL *)ERROR_PTR("no hits or misses", procName, NULL);
if (hitfract > 1.0 || missfract > 1.0)
return (SEL *)ERROR_PTR("fraction can't be > 1.0", procName, NULL);
if (distance <= 0)
distance = DEFAULT_DISTANCE_TO_BOUNDARY;
if (distance > MAX_DISTANCE_TO_BOUNDARY) {
L_WARNING("distance too large; setting to max value", procName);
distance = MAX_DISTANCE_TO_BOUNDARY;
}
/* Locate the foreground */
pixClipToForeground(pixs, &pixt1, NULL);
if (!pixt1)
return (SEL *)ERROR_PTR("pixt1 not made", procName, NULL);
ws = pixGetWidth(pixt1);
hs = pixGetHeight(pixt1);
w = ws;
h = hs;
/* Crop out a region including the foreground, and add pixels
* on sides depending on the side flags */
if (toppix || botpix || leftpix || rightpix) {
x = y = 0;
if (toppix) {
h += toppix;
y = toppix;
}
if (botpix)
h += botpix;
if (leftpix) {
w += leftpix;
x = leftpix;
}
if (rightpix)
w += rightpix;
pixt2 = pixCreate(w, h, 1);
pixRasterop(pixt2, x, y, ws, hs, PIX_SRC, pixt1, 0, 0);
}
else
pixt2 = pixClone(pixt1);
if (ppixe)
*ppixe = pixClone(pixt2);
pixDestroy(&pixt1);
/* Identify fg and bg pixels that are at least 'distance' pixels
* away from the boundary pixels in their set */
seld = selCreateBrick(2 * distance + 1, 2 * distance + 1,
distance, distance, SEL_HIT);
pixfg = pixErode(NULL, pixt2, seld);
pixbg = pixDilate(NULL, pixt2, seld);
pixInvert(pixbg, pixbg);
selDestroy(&seld);
pixDestroy(&pixt2);
/* Generate the sel from a random selection of these points */
sel = selCreateBrick(h, w, h / 2, w / 2, SEL_DONT_CARE);
if (hitfract > 0.0) {
thresh = (l_int32)(hitfract * (l_float64)RAND_MAX);
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
pixGetPixel(pixfg, j, i, &val);
if (val) {
if (rand() < thresh)
selSetElement(sel, i, j, SEL_HIT);
}
}
}
}
if (missfract > 0.0) {
thresh = (l_int32)(missfract * (l_float64)RAND_MAX);
for (i = 0; i < h; i++) {
for (j = 0; j < w; j++) {
pixGetPixel(pixbg, j, i, &val);
if (val) {
if (rand() < thresh)
selSetElement(sel, i, j, SEL_MISS);
}
}
}
}
pixDestroy(&pixfg);
pixDestroy(&pixbg);
return sel;
}
/*!
* pixOtsuAdaptiveThreshold()
*
* Input: pixs (8 bpp)
* sx, sy (desired tile dimensions; actual size may vary)
* smoothx, smoothy (half-width of convolution kernel applied to
* threshold array: use 0 for no smoothing)
* scorefract (fraction of the max Otsu score; typ. 0.1;
* use 0.0 for standard Otsu)
* &pixth (<optional return> array of threshold values
* found for each tile)
* &pixd (<optional return> thresholded input pixs, based on
* the threshold array)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) The Otsu method finds a single global threshold for an image.
* This function allows a locally adapted threshold to be
* found for each tile into which the image is broken up.
* (2) The array of threshold values, one for each tile, constitutes
* a highly downscaled image. This array is optionally
* smoothed using a convolution. The full width and height of the
* convolution kernel are (2 * @smoothx + 1) and (2 * @smoothy + 1).
* (3) The minimum tile dimension allowed is 16. If such small
* tiles are used, it is recommended to use smoothing, because
* without smoothing, each small tile determines the splitting
* threshold independently. A tile that is entirely in the
* image bg will then hallucinate fg, resulting in a very noisy
* binarization. The smoothing should be large enough that no
* tile is only influenced by one type (fg or bg) of pixels,
* because it will force a split of its pixels.
* (4) To get a single global threshold for the entire image, use
* input values of @sx and @sy that are larger than the image.
* For this situation, the smoothing parameters are ignored.
* (5) The threshold values partition the image pixels into two classes:
* one whose values are less than the threshold and another
* whose values are greater than or equal to the threshold.
* This is the same use of 'threshold' as in pixThresholdToBinary().
* (6) The scorefract is the fraction of the maximum Otsu score, which
* is used to determine the range over which the histogram minimum
* is searched. See numaSplitDistribution() for details on the
* underlying method of choosing a threshold.
* (7) This uses enables a modified version of the Otsu criterion for
* splitting the distribution of pixels in each tile into a
* fg and bg part. The modification consists of searching for
* a minimum in the histogram over a range of pixel values where
* the Otsu score is within a defined fraction, @scorefract,
* of the max score. To get the original Otsu algorithm, set
* @scorefract == 0.
*/
l_int32
pixOtsuAdaptiveThreshold(PIX *pixs,
l_int32 sx,
l_int32 sy,
l_int32 smoothx,
l_int32 smoothy,
l_float32 scorefract,
PIX **ppixth,
PIX **ppixd)
{
l_int32 w, h, nx, ny, i, j, thresh;
l_uint32 val;
PIX *pixt, *pixb, *pixthresh, *pixth, *pixd;
PIXTILING *pt;
PROCNAME("pixOtsuAdaptiveThreshold");
if (!ppixth && !ppixd){
return ERROR_INT("neither &pixth nor &pixd defined", procName, 1);
LOGE("neither &pixth nor &pixd defined");
}
if (ppixth) *ppixth = NULL;
if (ppixd) *ppixd = NULL;
if (!pixs || pixGetDepth(pixs) != 8){
return ERROR_INT("pixs not defined or not 8 bpp", procName, 1);
LOGE("pixs not defined or not 8 bpp");
}
if (sx < 16 || sy < 16){
return ERROR_INT("sx and sy must be >= 16", procName, 1);
LOGE("sx and sy must be >= 16");
}
/* Compute the threshold array for the tiles */
pixGetDimensions(pixs, &w, &h, NULL);
nx = L_MAX(1, w / sx);
ny = L_MAX(1, h / sy);
smoothx = L_MIN(smoothx, (nx - 1) / 2);
smoothy = L_MIN(smoothy, (ny - 1) / 2);
pt = pixTilingCreate(pixs, nx, ny, 0, 0, 0, 0);
pixthresh = pixCreate(nx, ny, 8);
for (i = 0; i < ny; i++) {
for (j = 0; j < nx; j++) {
pixt = pixTilingGetTile(pt, i, j);
pixSplitDistributionFgBg(pixt, scorefract, 1, &thresh,
NULL, NULL, 0);
pixSetPixel(pixthresh, j, i, thresh); /* see note (4) */
pixDestroy(&pixt);
}
}
/* Optionally smooth the threshold array */
if (smoothx > 0 || smoothy > 0)
pixth = pixBlockconv(pixthresh, smoothx, smoothy);
else
pixth = pixClone(pixthresh);
pixDestroy(&pixthresh);
/* Optionally apply the threshold array to binarize pixs */
if (ppixd) {
pixd = pixCreate(w, h, 1);
for (i = 0; i < ny; i++) {
for (j = 0; j < nx; j++) {
pixt = pixTilingGetTile(pt, i, j);
pixGetPixel(pixth, j, i, &val);
pixb = pixThresholdToBinary(pixt, val);
pixTilingPaintTile(pixd, i, j, pixb, pt);
pixDestroy(&pixt);
pixDestroy(&pixb);
}
}
*ppixd = pixd;
}
if (ppixth)
*ppixth = pixth;
else
pixDestroy(&pixth);
pixTilingDestroy(&pt);
return 0;
}
/*!
* \brief pixSaveTiledOutline()
*
* \param[in] pixs 1, 2, 4, 8, 32 bpp
* \param[in] pixa the pix are accumulated here
* \param[in] scalefactor 0.0 to disable; otherwise this is a scale factor
* \param[in] newrow 0 if placed on the same row as previous; 1 otherwise
* \param[in] space horizontal and vertical spacing, in pixels
* \param[in] linewidth width of added outline for image; 0 for no outline
* \param[in] dp depth of pixa; 8 or 32 bpp; only used on first call
* \return 0 if OK, 1 on error.
*
* <pre>
* Notes:
* (1) Before calling this function for the first time, use
* pixaCreate() to make the %pixa that will accumulate the pix.
* This is passed in each time pixSaveTiled() is called.
* (2) %scalefactor scales the input image. After scaling and
* possible depth conversion, the image is saved in the input
* pixa, along with a box that specifies the location to
* place it when tiled later. Disable saving the pix by
* setting %scalefactor == 0.0.
* (3) %newrow and %space specify the location of the new pix
* with respect to the last one(s) that were entered.
* (4) %dp specifies the depth at which all pix are saved. It can
* be only 8 or 32 bpp. Any colormap is removed. This is only
* used at the first invocation.
* (5) This function uses two variables from call to call.
* If they were static, the function would not be .so or thread
* safe, and furthermore, there would be interference with two or
* more pixa accumulating images at a time. Consequently,
* we use the first pix in the pixa to store and obtain both
* the depth and the current position of the bottom (one pixel
* below the lowest image raster line when laid out using
* the boxa). The bottom variable is stored in the input format
* field, which is the only field available for storing an int.
* </pre>
*/
l_int32
pixSaveTiledOutline(PIX *pixs,
PIXA *pixa,
l_float32 scalefactor,
l_int32 newrow,
l_int32 space,
l_int32 linewidth,
l_int32 dp)
{
l_int32 n, top, left, bx, by, bw, w, h, depth, bottom;
BOX *box;
PIX *pix1, *pix2, *pix3, *pix4;
PROCNAME("pixSaveTiledOutline");
if (scalefactor == 0.0) return 0;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (!pixa)
return ERROR_INT("pixa not defined", procName, 1);
n = pixaGetCount(pixa);
if (n == 0) {
bottom = 0;
if (dp != 8 && dp != 32) {
L_WARNING("dp not 8 or 32 bpp; using 32\n", procName);
depth = 32;
} else {
depth = dp;
}
} else { /* extract the depth and bottom params from the first pix */
pix1 = pixaGetPix(pixa, 0, L_CLONE);
depth = pixGetDepth(pix1);
bottom = pixGetInputFormat(pix1); /* not typical usage! */
pixDestroy(&pix1);
}
/* Remove colormap if it exists; otherwise a copy. This
* guarantees that pix4 is not a clone of pixs. */
pix1 = pixRemoveColormapGeneral(pixs, REMOVE_CMAP_BASED_ON_SRC, L_COPY);
/* Scale and convert to output depth */
if (scalefactor == 1.0) {
pix2 = pixClone(pix1);
} else if (scalefactor > 1.0) {
pix2 = pixScale(pix1, scalefactor, scalefactor);
} else if (scalefactor < 1.0) {
if (pixGetDepth(pix1) == 1)
pix2 = pixScaleToGray(pix1, scalefactor);
else
pix2 = pixScale(pix1, scalefactor, scalefactor);
}
pixDestroy(&pix1);
if (depth == 8)
pix3 = pixConvertTo8(pix2, 0);
else
pix3 = pixConvertTo32(pix2);
pixDestroy(&pix2);
/* Add black outline */
if (linewidth > 0)
pix4 = pixAddBorder(pix3, linewidth, 0);
else
pix4 = pixClone(pix3);
pixDestroy(&pix3);
/* Find position of current pix (UL corner plus size) */
if (n == 0) {
top = 0;
left = 0;
} else if (newrow == 1) {
top = bottom + space;
left = 0;
} else if (n > 0) {
pixaGetBoxGeometry(pixa, n - 1, &bx, &by, &bw, NULL);
top = by;
left = bx + bw + space;
}
pixGetDimensions(pix4, &w, &h, NULL);
bottom = L_MAX(bottom, top + h);
box = boxCreate(left, top, w, h);
pixaAddPix(pixa, pix4, L_INSERT);
pixaAddBox(pixa, box, L_INSERT);
/* Save the new bottom value */
pix1 = pixaGetPix(pixa, 0, L_CLONE);
pixSetInputFormat(pix1, bottom); /* not typical usage! */
pixDestroy(&pix1);
return 0;
}
/*!
* dewarpShowResults()
*
* Input: dewa
* sarray (of indexed input images)
* boxa (crop boxes for input images; can be null)
* firstpage, lastpage
* fontdir (for text bitmap fonts)
* pdfout (filename)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This generates a pdf of image pairs (before, after) for
* the designated set of input pages.
* (2) If the boxa exists, its elements are aligned with numbers
* in the filenames in @sa. It is used to crop the input images.
* It is assumed that the dewa was generated from the cropped
* images. No undercropping is applied before rendering.
*/
l_int32
dewarpShowResults(L_DEWARPA *dewa,
SARRAY *sa,
BOXA *boxa,
l_int32 firstpage,
l_int32 lastpage,
const char *fontdir,
const char *pdfout)
{
char bufstr[256];
char *outpath;
l_int32 i, modelpage;
L_BMF *bmf;
BOX *box;
L_DEWARP *dew;
PIX *pixs, *pixc, *pixd, *pixt1, *pixt2;
PIXA *pixa;
PROCNAME("dewarpShowResults");
if (!dewa)
return ERROR_INT("dewa not defined", procName, 1);
if (!sa)
return ERROR_INT("sa not defined", procName, 1);
if (!pdfout)
return ERROR_INT("pdfout not defined", procName, 1);
if (firstpage > lastpage)
return ERROR_INT("invalid first/last page numbers", procName, 1);
lept_rmdir("dewarp_pdfout");
lept_mkdir("dewarp_pdfout");
if ((bmf = bmfCreate(fontdir, 6)) == NULL)
L_ERROR("bmf not made; page info not displayed", procName);
fprintf(stderr, "Dewarping and generating s/by/s view\n");
for (i = firstpage; i <= lastpage; i++) {
if (i && (i % 10 == 0)) fprintf(stderr, ".. %d ", i);
pixs = pixReadIndexed(sa, i);
if (boxa) {
box = boxaGetBox(boxa, i, L_CLONE);
pixc = pixClipRectangle(pixs, box, NULL);
boxDestroy(&box);
}
else
pixc = pixClone(pixs);
dew = dewarpaGetDewarp(dewa, i);
pixd = NULL;
if (dew) {
dewarpaApplyDisparity(dewa, dew->pageno, pixc,
GRAYIN_VALUE, 0, 0, &pixd, NULL);
dewarpMinimize(dew);
}
pixa = pixaCreate(2);
pixaAddPix(pixa, pixc, L_INSERT);
if (pixd)
pixaAddPix(pixa, pixd, L_INSERT);
pixt1 = pixaDisplayTiledAndScaled(pixa, 32, 500, 2, 0, 35, 2);
if (dew) {
modelpage = (dew->hasref) ? dew->refpage : dew->pageno;
snprintf(bufstr, sizeof(bufstr), "Page %d; using %d\n",
i, modelpage);
}
else
snprintf(bufstr, sizeof(bufstr), "Page %d; no dewarp\n", i);
pixt2 = pixAddSingleTextblock(pixt1, bmf, bufstr, 0x0000ff00,
L_ADD_BELOW, 0);
snprintf(bufstr, sizeof(bufstr), "/tmp/dewarp_pdfout/%05d", i);
pixWrite(bufstr, pixt2, IFF_JFIF_JPEG);
pixaDestroy(&pixa);
pixDestroy(&pixs);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
}
fprintf(stderr, "\n");
fprintf(stderr, "Generating pdf of result\n");
convertFilesToPdf("/tmp/dewarp_pdfout", NULL, 100, 1.0, L_JPEG_ENCODE,
0, "Dewarp sequence", pdfout);
outpath = genPathname(pdfout, NULL);
fprintf(stderr, "Output written to: %s\n", outpath);
FREE(outpath);
bmfDestroy(&bmf);
return 0;
}
/*!
* pixaDisplayTiledAndScaled()
*
* Input: pixa
* outdepth (output depth: 1, 8 or 32 bpp)
* tilewidth (each pix is scaled to this width)
* ncols (number of tiles in each row)
* background (0 for white, 1 for black; this is the color
* of the spacing between the images)
* spacing (between images, and on outside)
* border (width of additional black border on each image;
* use 0 for no border)
* Return: pix of tiled images, or null on error
*
* Notes:
* (1) This can be used to tile a number of renderings of
* an image that are at different scales and depths.
* (2) Each image, after scaling and optionally adding the
* black border, has width 'tilewidth'. Thus, the border does
* not affect the spacing between the image tiles. The
* maximum allowed border width is tilewidth / 5.
*/
PIX *
pixaDisplayTiledAndScaled(PIXA *pixa,
l_int32 outdepth,
l_int32 tilewidth,
l_int32 ncols,
l_int32 background,
l_int32 spacing,
l_int32 border)
{
l_int32 x, y, w, h, wd, hd, d;
l_int32 i, n, nrows, maxht, ninrow, irow, bordval;
l_int32 *rowht;
l_float32 scalefact;
PIX *pix, *pixn, *pixt, *pixb, *pixd;
PIXA *pixan;
PROCNAME("pixaDisplayTiledAndScaled");
if (!pixa)
return (PIX *)ERROR_PTR("pixa not defined", procName, NULL);
if (outdepth != 1 && outdepth != 8 && outdepth != 32)
return (PIX *)ERROR_PTR("outdepth not in {1, 8, 32}", procName, NULL);
if (border < 0 || border > tilewidth / 5)
border = 0;
if ((n = pixaGetCount(pixa)) == 0)
return (PIX *)ERROR_PTR("no components", procName, NULL);
/* Normalize scale and depth for each pix; optionally add border */
pixan = pixaCreate(n);
bordval = (outdepth == 1) ? 1 : 0;
for (i = 0; i < n; i++) {
if ((pix = pixaGetPix(pixa, i, L_CLONE)) == NULL)
continue;
pixGetDimensions(pix, &w, &h, &d);
scalefact = (l_float32)(tilewidth - 2 * border) / (l_float32)w;
if (d == 1 && outdepth > 1 && scalefact < 1.0)
pixt = pixScaleToGray(pix, scalefact);
else
pixt = pixScale(pix, scalefact, scalefact);
if (outdepth == 1)
pixn = pixConvertTo1(pixt, 128);
else if (outdepth == 8)
pixn = pixConvertTo8(pixt, FALSE);
else /* outdepth == 32 */
pixn = pixConvertTo32(pixt);
pixDestroy(&pixt);
if (border)
pixb = pixAddBorder(pixn, border, bordval);
else
pixb = pixClone(pixn);
pixaAddPix(pixan, pixb, L_INSERT);
pixDestroy(&pix);
pixDestroy(&pixn);
}
if ((n = pixaGetCount(pixan)) == 0) { /* should not have changed! */
pixaDestroy(&pixan);
return (PIX *)ERROR_PTR("no components", procName, NULL);
}
/* Determine the size of each row and of pixd */
wd = tilewidth * ncols + spacing * (ncols + 1);
nrows = (n + ncols - 1) / ncols;
if ((rowht = (l_int32 *)CALLOC(nrows, sizeof(l_int32))) == NULL)
return (PIX *)ERROR_PTR("rowht array not made", procName, NULL);
maxht = 0;
ninrow = 0;
irow = 0;
for (i = 0; i < n; i++) {
pix = pixaGetPix(pixan, i, L_CLONE);
ninrow++;
pixGetDimensions(pix, &w, &h, NULL);
maxht = L_MAX(h, maxht);
if (ninrow == ncols) {
rowht[irow] = maxht;
maxht = ninrow = 0; /* reset */
irow++;
}
pixDestroy(&pix);
}
if (ninrow > 0) { /* last fencepost */
rowht[irow] = maxht;
irow++; /* total number of rows */
}
nrows = irow;
hd = spacing * (nrows + 1);
for (i = 0; i < nrows; i++)
hd += rowht[i];
pixd = pixCreate(wd, hd, outdepth);
if ((background == 1 && outdepth == 1) ||
(background == 0 && outdepth != 1))
pixSetAll(pixd);
/* Now blit images to pixd */
x = y = spacing;
irow = 0;
for (i = 0; i < n; i++) {
pix = pixaGetPix(pixan, i, L_CLONE);
pixGetDimensions(pix, &w, &h, NULL);
if (i && ((i % ncols) == 0)) { /* start new row */
x = spacing;
y += spacing + rowht[irow];
irow++;
}
pixRasterop(pixd, x, y, w, h, PIX_SRC, pix, 0, 0);
x += tilewidth + spacing;
pixDestroy(&pix);
}
pixaDestroy(&pixan);
FREE(rowht);
return pixd;
}
/*!
* ioFormatTest()
*
* Input: filename (input file)
* Return: 0 if OK; 1 on error or if the test fails
*
* Notes:
* (1) This writes and reads a set of output files losslessly
* in different formats to /tmp/format/, and tests that the
* result before and after is unchanged.
* (2) This should work properly on input images of any depth,
* with and without colormaps.
* (3) All supported formats are tested for bmp, png, tiff and
* non-ascii pnm. Ascii pnm also works (but who'd ever want
* to use it?) We allow 2 bpp bmp, although it's not
* supported elsewhere. And we don't support reading
* 16 bpp png, although this can be turned on in pngio.c.
* (4) This silently skips png or tiff testing if HAVE_LIBPNG
* or HAVE_LIBTIFF are 0, respectively.
*/
l_int32
ioFormatTest(const char *filename)
{
l_int32 d, equal, problems;
PIX *pixs, *pixc, *pix1, *pix2;
PIXCMAP *cmap;
PROCNAME("ioFormatTest");
if (!filename)
return ERROR_INT("filename not defined", procName, 1);
if ((pixs = pixRead(filename)) == NULL)
return ERROR_INT("pixs not made", procName, 1);
lept_mkdir("lept");
/* Note that the reader automatically removes colormaps
* from 1 bpp BMP images, but not from 8 bpp BMP images.
* Therefore, if our 8 bpp image initially doesn't have a
* colormap, we are going to need to remove it from any
* pix read from a BMP file. */
pixc = pixClone(pixs); /* laziness */
/* This does not test the alpha layer pixels, because most
* formats don't support it. Remove any alpha. */
if (pixGetSpp(pixc) == 4)
pixSetSpp(pixc, 3);
cmap = pixGetColormap(pixc); /* colormap; can be NULL */
d = pixGetDepth(pixc);
problems = FALSE;
/* ----------------------- BMP -------------------------- */
/* BMP works for 1, 2, 4, 8 and 32 bpp images.
* It always writes colormaps for 1 and 8 bpp, so we must
* remove it after readback if the input image doesn't have
* a colormap. Although we can write/read 2 bpp BMP, nobody
* else can read them! */
if (d == 1 || d == 8) {
L_INFO("write/read bmp\n", procName);
pixWrite(FILE_BMP, pixc, IFF_BMP);
pix1 = pixRead(FILE_BMP);
if (!cmap)
pix2 = pixRemoveColormap(pix1, REMOVE_CMAP_BASED_ON_SRC);
else
pix2 = pixClone(pix1);
pixEqual(pixc, pix2, &equal);
if (!equal) {
L_INFO(" **** bad bmp image: d = %d ****\n", procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
pixDestroy(&pix2);
}
if (d == 2 || d == 4 || d == 32) {
L_INFO("write/read bmp\n", procName);
pixWrite(FILE_BMP, pixc, IFF_BMP);
pix1 = pixRead(FILE_BMP);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad bmp image: d = %d ****\n", procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
}
/* ----------------------- PNG -------------------------- */
#if HAVE_LIBPNG
/* PNG works for all depths, but here, because we strip
* 16 --> 8 bpp on reading, we don't test png for 16 bpp. */
if (d != 16) {
L_INFO("write/read png\n", procName);
pixWrite(FILE_PNG, pixc, IFF_PNG);
pix1 = pixRead(FILE_PNG);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad png image: d = %d ****\n", procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
}
#endif /* HAVE_LIBPNG */
/* ----------------------- TIFF -------------------------- */
#if HAVE_LIBTIFF
/* TIFF works for 1, 2, 4, 8, 16 and 32 bpp images.
* Because 8 bpp tiff always writes 256 entry colormaps, the
* colormap sizes may be different for 8 bpp images with
* colormap; we are testing if the image content is the same.
* Likewise, the 2 and 4 bpp tiff images with colormaps
* have colormap sizes 4 and 16, rsp. This test should
* work properly on the content, regardless of the number
* of color entries in pixc. */
/* tiff uncompressed works for all pixel depths */
L_INFO("write/read uncompressed tiff\n", procName);
pixWrite(FILE_TIFF, pixc, IFF_TIFF);
pix1 = pixRead(FILE_TIFF);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad tiff uncompressed image: d = %d ****\n",
procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
/* tiff lzw works for all pixel depths */
L_INFO("write/read lzw compressed tiff\n", procName);
pixWrite(FILE_LZW, pixc, IFF_TIFF_LZW);
pix1 = pixRead(FILE_LZW);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad tiff lzw compressed image: d = %d ****\n",
procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
/* tiff adobe deflate (zip) works for all pixel depths */
L_INFO("write/read zip compressed tiff\n", procName);
pixWrite(FILE_ZIP, pixc, IFF_TIFF_ZIP);
pix1 = pixRead(FILE_ZIP);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad tiff zip compressed image: d = %d ****\n",
procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
/* tiff g4, g3, rle and packbits work for 1 bpp */
if (d == 1) {
L_INFO("write/read g4 compressed tiff\n", procName);
pixWrite(FILE_G4, pixc, IFF_TIFF_G4);
pix1 = pixRead(FILE_G4);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad tiff g4 image ****\n", procName);
problems = TRUE;
}
pixDestroy(&pix1);
L_INFO("write/read g3 compressed tiff\n", procName);
pixWrite(FILE_G3, pixc, IFF_TIFF_G3);
pix1 = pixRead(FILE_G3);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad tiff g3 image ****\n", procName);
problems = TRUE;
}
pixDestroy(&pix1);
L_INFO("write/read rle compressed tiff\n", procName);
pixWrite(FILE_RLE, pixc, IFF_TIFF_RLE);
pix1 = pixRead(FILE_RLE);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad tiff rle image: d = %d ****\n", procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
L_INFO("write/read packbits compressed tiff\n", procName);
pixWrite(FILE_PB, pixc, IFF_TIFF_PACKBITS);
pix1 = pixRead(FILE_PB);
pixEqual(pixc, pix1, &equal);
if (!equal) {
L_INFO(" **** bad tiff packbits image: d = %d ****\n",
procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
}
#endif /* HAVE_LIBTIFF */
/* ----------------------- PNM -------------------------- */
/* pnm works for 1, 2, 4, 8, 16 and 32 bpp.
* pnm doesn't have colormaps, so when we write colormapped
* pix out as pnm, the colormap is removed. Thus for the test,
* we must remove the colormap from pixc before testing. */
L_INFO("write/read pnm\n", procName);
pixWrite(FILE_PNM, pixc, IFF_PNM);
pix1 = pixRead(FILE_PNM);
if (cmap)
pix2 = pixRemoveColormap(pixc, REMOVE_CMAP_BASED_ON_SRC);
else
pix2 = pixClone(pixc);
pixEqual(pix1, pix2, &equal);
if (!equal) {
L_INFO(" **** bad pnm image: d = %d ****\n", procName, d);
problems = TRUE;
}
pixDestroy(&pix1);
pixDestroy(&pix2);
if (problems == FALSE)
L_INFO("All formats read and written OK!\n", procName);
pixDestroy(&pixc);
pixDestroy(&pixs);
return problems;
}
/*!
* \brief jbWordsInTextlines()
*
* \param[in] dirin directory of input pages
* \param[in] reduction 1 for full res; 2 for half-res
* \param[in] maxwidth of word mask components, to be kept
* \param[in] maxheight of word mask components, to be kept
* \param[in] thresh on correlation; 0.80 is reasonable
* \param[in] weight for handling thick text; 0.6 is reasonable
* \param[out] pnatl numa with textline index for each component
* \param[in] firstpage 0-based
* \param[in] npages use 0 for all pages in dirin
* \return classer for the set of pages
*
* <pre>
* Notes:
* (1) This is a high-level function. See prog/jbwords for example
* of usage.
* (2) Typically, use input of 75 - 150 ppi for finding words.
* </pre>
*/
JBCLASSER *
jbWordsInTextlines(const char *dirin,
l_int32 reduction,
l_int32 maxwidth,
l_int32 maxheight,
l_float32 thresh,
l_float32 weight,
NUMA **pnatl,
l_int32 firstpage,
l_int32 npages)
{
char *fname;
l_int32 nfiles, i, w, h;
BOXA *boxa;
JBCLASSER *classer;
NUMA *nai, *natl;
PIX *pix1, *pix2;
PIXA *pixa;
SARRAY *safiles;
PROCNAME("jbWordsInTextlines");
if (!pnatl)
return (JBCLASSER *)ERROR_PTR("&natl not defined", procName, NULL);
*pnatl = NULL;
if (!dirin)
return (JBCLASSER *)ERROR_PTR("dirin not defined", procName, NULL);
if (reduction != 1 && reduction != 2)
return (JBCLASSER *)ERROR_PTR("reduction not in {1,2}", procName, NULL);
safiles = getSortedPathnamesInDirectory(dirin, NULL, firstpage, npages);
nfiles = sarrayGetCount(safiles);
/* Classify components */
classer = jbCorrelationInit(JB_WORDS, maxwidth, maxheight, thresh, weight);
classer->safiles = sarrayCopy(safiles);
natl = numaCreate(0);
*pnatl = natl;
for (i = 0; i < nfiles; i++) {
fname = sarrayGetString(safiles, i, L_NOCOPY);
if ((pix1 = pixRead(fname)) == NULL) {
L_WARNING("image file %d not read\n", procName, i);
continue;
}
if (reduction == 1)
pix2 = pixClone(pix1);
else /* reduction == 2 */
pix2 = pixReduceRankBinaryCascade(pix1, 1, 0, 0, 0);
pixGetWordsInTextlines(pix2, JB_WORDS_MIN_WIDTH,
JB_WORDS_MIN_HEIGHT, maxwidth, maxheight,
&boxa, &pixa, &nai);
pixGetDimensions(pix2, &w, &h, NULL);
classer->w = w;
classer->h = h;
jbAddPageComponents(classer, pix2, boxa, pixa);
numaJoin(natl, nai, 0, -1);
pixDestroy(&pix1);
pixDestroy(&pix2);
numaDestroy(&nai);
boxaDestroy(&boxa);
pixaDestroy(&pixa);
}
sarrayDestroy(&safiles);
return classer;
}
/*!
* pixaDisplayTiledInRows()
*
* Input: pixa
* outdepth (output depth: 1, 8 or 32 bpp)
* maxwidth (of output image)
* scalefactor (applied to every pix; use 1.0 for no scaling)
* background (0 for white, 1 for black; this is the color
* of the spacing between the images)
* spacing (between images, and on outside)
* border (width of black border added to each image;
* use 0 for no border)
* Return: pixd (of tiled images), or null on error
*
* Notes:
* (1) This saves a pixa to a single image file of width not to
* exceed maxwidth, with background color either white or black,
* and with each row tiled such that the top of each pix is
* aligned and separated by 'spacing' from the next one.
* A black border can be added to each pix.
* (2) All pix are converted to outdepth; existing colormaps are removed.
* (3) This does a reasonably spacewise-efficient job of laying
* out the individual pix images into a tiled composite.
*/
PIX *
pixaDisplayTiledInRows(PIXA *pixa,
l_int32 outdepth,
l_int32 maxwidth,
l_float32 scalefactor,
l_int32 background,
l_int32 spacing,
l_int32 border)
{
l_int32 h; /* cumulative height over all the rows */
l_int32 w; /* cumulative height in the current row */
l_int32 bordval, wtry, wt, ht;
l_int32 irow; /* index of current pix in current row */
l_int32 wmaxrow; /* width of the largest row */
l_int32 maxh; /* max height in row */
l_int32 i, j, index, n, x, y, nrows, ninrow;
NUMA *nainrow; /* number of pix in the row */
NUMA *namaxh; /* height of max pix in the row */
PIX *pix, *pixn, *pixt, *pixd;
PIXA *pixan;
PROCNAME("pixaDisplayTiledInRows");
if (!pixa)
return (PIX *)ERROR_PTR("pixa not defined", procName, NULL);
if (outdepth != 1 && outdepth != 8 && outdepth != 32)
return (PIX *)ERROR_PTR("outdepth not in {1, 8, 32}", procName, NULL);
if (border < 0)
border = 0;
if (scalefactor <= 0.0) scalefactor = 1.0;
if ((n = pixaGetCount(pixa)) == 0)
return (PIX *)ERROR_PTR("no components", procName, NULL);
/* Normalize depths, scale, remove colormaps; optionally add border */
pixan = pixaCreate(n);
bordval = (outdepth == 1) ? 1 : 0;
for (i = 0; i < n; i++) {
if ((pix = pixaGetPix(pixa, i, L_CLONE)) == NULL)
continue;
if (outdepth == 1)
pixn = pixConvertTo1(pix, 128);
else if (outdepth == 8)
pixn = pixConvertTo8(pix, FALSE);
else /* outdepth == 32 */
pixn = pixConvertTo32(pix);
pixDestroy(&pix);
if (scalefactor != 1.0)
pixt = pixScale(pixn, scalefactor, scalefactor);
else
pixt = pixClone(pixn);
if (border)
pixd = pixAddBorder(pixt, border, bordval);
else
pixd = pixClone(pixt);
pixDestroy(&pixn);
pixDestroy(&pixt);
pixaAddPix(pixan, pixd, L_INSERT);
}
if (pixaGetCount(pixan) != n) {
n = pixaGetCount(pixan);
L_WARNING_INT("only got %d components", procName, n);
if (n == 0) {
pixaDestroy(&pixan);
return (PIX *)ERROR_PTR("no components", procName, NULL);
}
}
/* Compute parameters for layout */
nainrow = numaCreate(0);
namaxh = numaCreate(0);
wmaxrow = 0;
w = h = spacing;
maxh = 0; /* max height in row */
for (i = 0, irow = 0; i < n; i++, irow++) {
pixaGetPixDimensions(pixan, i, &wt, &ht, NULL);
wtry = w + wt + spacing;
if (wtry > maxwidth) { /* end the current row and start next one */
numaAddNumber(nainrow, irow);
numaAddNumber(namaxh, maxh);
wmaxrow = L_MAX(wmaxrow, w);
h += maxh + spacing;
irow = 0;
w = wt + 2 * spacing;
maxh = ht;
} else {
w = wtry;
maxh = L_MAX(maxh, ht);
}
}
/* Enter the parameters for the last row */
numaAddNumber(nainrow, irow);
numaAddNumber(namaxh, maxh);
wmaxrow = L_MAX(wmaxrow, w);
h += maxh + spacing;
if ((pixd = pixCreate(wmaxrow, h, outdepth)) == NULL) {
numaDestroy(&nainrow);
numaDestroy(&namaxh);
pixaDestroy(&pixan);
return (PIX *)ERROR_PTR("pixd not made", procName, NULL);
}
/* Reset the background color if necessary */
if ((background == 1 && outdepth == 1) ||
(background == 0 && outdepth != 1))
pixSetAll(pixd);
/* Blit the images to the dest */
nrows = numaGetCount(nainrow);
y = spacing;
for (i = 0, index = 0; i < nrows; i++) { /* over rows */
numaGetIValue(nainrow, i, &ninrow);
numaGetIValue(namaxh, i, &maxh);
x = spacing;
for (j = 0; j < ninrow; j++, index++) { /* over pix in row */
pix = pixaGetPix(pixan, index, L_CLONE);
pixGetDimensions(pix, &wt, &ht, NULL);
pixRasterop(pixd, x, y, wt, ht, PIX_SRC, pix, 0, 0);
pixDestroy(&pix);
x += wt + spacing;
}
y += maxh + spacing;
}
numaDestroy(&nainrow);
numaDestroy(&namaxh);
pixaDestroy(&pixan);
return pixd;
}
void
RotateTest(PIX *pixs,
l_int32 reduction,
L_REGPARAMS *rp)
{
l_int32 w, h, d, outformat;
PIX *pixt1, *pixt2, *pixt3, *pixd;
PIXA *pixa;
pixGetDimensions(pixs, &w, &h, &d);
outformat = (d == 8 || d == 32) ? IFF_JFIF_JPEG : IFF_PNG;
pixa = pixaCreate(0);
pixt1 = pixRotate(pixs, ANGLE1, L_ROTATE_SHEAR, L_BRING_IN_WHITE, w, h);
pixSaveTiled(pixt1, pixa, reduction, 1, 20, 32);
pixt2 = pixRotate(pixs, ANGLE1, L_ROTATE_SHEAR, L_BRING_IN_BLACK, w, h);
pixSaveTiled(pixt2, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixt1 = pixRotate(pixs, ANGLE1, L_ROTATE_SHEAR, L_BRING_IN_WHITE, 0, 0);
pixSaveTiled(pixt1, pixa, reduction, 1, 20, 0);
pixt2 = pixRotate(pixs, ANGLE1, L_ROTATE_SHEAR, L_BRING_IN_BLACK, 0, 0);
pixSaveTiled(pixt2, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixt1 = pixRotate(pixs, ANGLE2, L_ROTATE_SHEAR, L_BRING_IN_WHITE, w, h);
pixSaveTiled(pixt1, pixa, reduction, 1, 20, 0);
pixt2 = pixRotate(pixs, ANGLE2, L_ROTATE_SHEAR, L_BRING_IN_BLACK, w, h);
pixSaveTiled(pixt2, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixt1 = pixRotate(pixs, ANGLE2, L_ROTATE_SHEAR, L_BRING_IN_WHITE, 0, 0);
pixSaveTiled(pixt1, pixa, reduction, 1, 20, 0);
pixt2 = pixRotate(pixs, ANGLE2, L_ROTATE_SHEAR, L_BRING_IN_BLACK, 0, 0);
pixSaveTiled(pixt2, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixd = pixaDisplay(pixa, 0, 0);
regTestWritePixAndCheck(rp, pixd, outformat);
pixDisplayWithTitle(pixd, 100, 100, NULL, rp->display);
pixDestroy(&pixd);
pixaDestroy(&pixa);
pixa = pixaCreate(0);
pixt1 = pixRotate(pixs, ANGLE2, L_ROTATE_SAMPLING, L_BRING_IN_WHITE, w, h);
pixSaveTiled(pixt1, pixa, reduction, 1, 20, 32);
pixt2 = pixRotate(pixs, ANGLE2, L_ROTATE_SAMPLING, L_BRING_IN_BLACK, w, h);
pixSaveTiled(pixt2, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
pixt1 = pixRotate(pixs, ANGLE2, L_ROTATE_SAMPLING, L_BRING_IN_WHITE, 0, 0);
pixSaveTiled(pixt1, pixa, reduction, 1, 20, 0);
pixt2 = pixRotate(pixs, ANGLE2, L_ROTATE_SAMPLING, L_BRING_IN_BLACK, 0, 0);
pixSaveTiled(pixt2, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
if (pixGetDepth(pixs) == 1)
pixt1 = pixScaleToGray2(pixs);
else
pixt1 = pixClone(pixs);
pixt2 = pixRotate(pixt1, ANGLE2, L_ROTATE_AREA_MAP, L_BRING_IN_WHITE, w, h);
pixSaveTiled(pixt2, pixa, reduction, 1, 20, 0);
pixt3 = pixRotate(pixt1, ANGLE2, L_ROTATE_AREA_MAP, L_BRING_IN_BLACK, w, h);
pixSaveTiled(pixt3, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
pixt2 = pixRotate(pixt1, ANGLE2, L_ROTATE_AREA_MAP, L_BRING_IN_WHITE, 0, 0);
pixSaveTiled(pixt2, pixa, reduction, 1, 20, 0);
pixt3 = pixRotate(pixt1, ANGLE2, L_ROTATE_AREA_MAP, L_BRING_IN_BLACK, 0, 0);
pixSaveTiled(pixt3, pixa, reduction, 0, 20, 0);
pixDestroy(&pixt2);
pixDestroy(&pixt3);
pixDestroy(&pixt1);
pixd = pixaDisplay(pixa, 0, 0);
regTestWritePixAndCheck(rp, pixd, outformat);
pixDisplayWithTitle(pixd, 100, 100, NULL, rp->display);
pixDestroy(&pixd);
pixaDestroy(&pixa);
return;
}
/*
* pixWriteSegmentedPageToPS()
*
* Input: pixs (all depths; colormap ok)
* pixm (<optional> 1 bpp segmentation mask over image region)
* textscale (scale of text output relative to pixs)
* imagescale (scale of image output relative to pixs)
* threshold (threshold for binarization; typ. 190)
* pageno (page number in set; use 1 for new output file)
* fileout (output ps file)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) This generates the PS string for a mixed text/image page,
* and adds it to an existing file if @pageno > 1.
* The PS output is determined by fitting the result to
* a letter-size (8.5 x 11 inch) page.
* (2) The two images (pixs and pixm) are at the same resolution
* (typically 300 ppi). They are used to generate two compressed
* images, pixb and pixc, that are put directly into the output
* PS file.
* (3) pixb is the text component. In the PostScript world, we think of
* it as a mask through which we paint black. It is produced by
* scaling pixs by @textscale, and thresholding to 1 bpp.
* (4) pixc is the image component, which is that part of pixs under
* the mask pixm. It is scaled from pixs by @imagescale.
* (5) Typical values are textscale = 2.0 and imagescale = 0.5.
* (6) If pixm == NULL, the page has only text. If it is all black,
* the page is all image and has no text.
* (7) This can be used to write a multi-page PS file, by using
* sequential page numbers with the same output file. It can
* also be used to write separate PS files for each page,
* by using different output files with @pageno = 0 or 1.
*/
l_int32
pixWriteSegmentedPageToPS(PIX *pixs,
PIX *pixm,
l_float32 textscale,
l_float32 imagescale,
l_int32 threshold,
l_int32 pageno,
const char *fileout)
{
l_int32 alltext, notext, d, ret;
l_uint32 val;
l_float32 scaleratio;
PIX *pixmi, *pixmis, *pixt, *pixg, *pixsc, *pixb, *pixc;
PROCNAME("pixWriteSegmentedPageToPS");
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (!fileout)
return ERROR_INT("fileout not defined", procName, 1);
if (imagescale <= 0.0 || textscale <= 0.0)
return ERROR_INT("relative scales must be > 0.0", procName, 1);
/* Analyze the page. Determine the ratio by which the
* binary text mask is scaled relative to the image part.
* If there is no image region (alltext == TRUE), the
* text mask will be rendered directly to fit the page,
* and scaleratio = 1.0. */
alltext = TRUE;
notext = FALSE;
scaleratio = 1.0;
if (pixm) {
pixZero(pixm, &alltext); /* pixm empty: all text */
if (alltext)
pixm = NULL; /* treat it as not existing here */
else {
pixmi = pixInvert(NULL, pixm);
pixZero(pixmi, ¬ext); /* pixm full; no text */
pixDestroy(&pixmi);
scaleratio = textscale / imagescale;
}
}
if (pixGetDepth(pixs) == 1) { /* render tiff g4 */
pixb = pixClone(pixs);
pixc = NULL;
}
else {
pixt = pixConvertTo8Or32(pixs, 0, 0); /* this can be a clone of pixs */
/* Get the binary text mask. Note that pixg cannot be a
* clone of pixs, because it may be altered by pixSetMasked(). */
pixb = NULL;
if (notext == FALSE) {
d = pixGetDepth(pixt);
if (d == 8)
pixg = pixCopy(NULL, pixt);
else /* d == 32 */
pixg = pixConvertRGBToLuminance(pixt);
if (pixm) /* clear out the image parts */
pixSetMasked(pixg, pixm, 255);
if (textscale == 1.0)
pixsc = pixClone(pixg);
else if (textscale >= 0.7)
pixsc = pixScaleGrayLI(pixg, textscale, textscale);
else
pixsc = pixScaleAreaMap(pixg, textscale, textscale);
pixb = pixThresholdToBinary(pixsc, threshold);
pixDestroy(&pixg);
pixDestroy(&pixsc);
}
/* Get the scaled image region */
pixc = NULL;
if (pixm) {
if (imagescale == 1.0)
pixsc = pixClone(pixt); /* can possibly be a clone of pixs */
else
pixsc = pixScale(pixt, imagescale, imagescale);
/* If pixm is not full, clear the pixels in pixsc
* corresponding to bg in pixm, where there can be text
* that is written through the mask pixb. Note that
* we could skip this and use pixsc directly in
* pixWriteMixedToPS(); however, clearing these
* non-image regions to a white background will reduce
* the size of pixc (relative to pixsc), and hence
* reduce the size of the PS file that is generated.
* Use a copy so that we don't accidentally alter pixs. */
if (notext == FALSE) {
pixmis = pixScale(pixm, imagescale, imagescale);
pixmi = pixInvert(NULL, pixmis);
val = (d == 8) ? 0xff : 0xffffff00;
pixc = pixCopy(NULL, pixsc);
pixSetMasked(pixc, pixmi, val); /* clear non-image part */
pixDestroy(&pixmis);
pixDestroy(&pixmi);
}
else
pixc = pixClone(pixsc);
pixDestroy(&pixsc);
}
pixDestroy(&pixt);
}
ret = pixWriteMixedToPS(pixb, pixc, scaleratio, pageno, fileout);
pixDestroy(&pixb);
pixDestroy(&pixc);
return ret;
}
main(int argc,
char **argv)
{
char *filein, *fileout;
l_int32 w, h, d, w2, h2, i, ncols;
l_float32 angle, conf;
BOX *box;
BOXA *boxa, *boxas, *boxad, *boxa2;
NUMA *numa;
PIX *pixs, *pixt, *pixb, *pixb2, *pixd;
PIX *pixtlm, *pixvws;
PIX *pixt1, *pixt2, *pixt3, *pixt4, *pixt5, *pixt6;
PIXA *pixam, *pixac, *pixad, *pixat;
PIXAA *pixaa, *pixaa2;
PTA *pta;
SEL *selsplit;
static char mainName[] = "textlinemask";
if (argc != 3)
exit(ERROR_INT(" Syntax: textlinemask filein fileout", mainName, 1));
filein = argv[1];
fileout = argv[2];
pixDisplayWrite(NULL, -1); /* init debug output */
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
pixGetDimensions(pixs, &w, &h, &d);
/* Binarize input */
if (d == 8)
pixt = pixThresholdToBinary(pixs, 128);
else if (d == 1)
pixt = pixClone(pixs);
else {
fprintf(stderr, "depth is %d\n", d);
exit(1);
}
/* Deskew */
pixb = pixFindSkewAndDeskew(pixt, 1, &angle, &conf);
pixDestroy(&pixt);
fprintf(stderr, "Skew angle: %7.2f degrees; %6.2f conf\n", angle, conf);
pixDisplayWrite(pixb, DEBUG_OUTPUT);
#if 1
/* Use full image morphology to find columns, at 2x reduction.
* This only works for very simple layouts where each column
* of text extends the full height of the input image.
* pixam has a pix component over each column. */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixt1 = pixMorphCompSequence(pixb2, "c5.500", 0);
boxa = pixConnComp(pixt1, &pixam, 8);
ncols = boxaGetCount(boxa);
fprintf(stderr, "Num columns: %d\n", ncols);
pixDisplayWrite(pixt1, DEBUG_OUTPUT);
/* Use selective region-based morphology to get the textline mask. */
pixad = pixaMorphSequenceByRegion(pixb2, pixam, "c100.3", 0, 0);
pixGetDimensions(pixb2, &w2, &h2, NULL);
if (DEBUG_OUTPUT) {
pixt2 = pixaDisplay(pixad, w2, h2);
pixDisplayWrite(pixt2, DEBUG_OUTPUT);
pixDestroy(&pixt2);
}
/* Some of the lines may be touching, so use a HMT to split the
* lines in each column, and use a pixaa to save the results. */
selsplit = selCreateFromString(seltext, 17, 7, "selsplit");
pixaa = pixaaCreate(ncols);
for (i = 0; i < ncols; i++) {
pixt3 = pixaGetPix(pixad, i, L_CLONE);
box = pixaGetBox(pixad, i, L_COPY);
pixt4 = pixHMT(NULL, pixt3, selsplit);
pixXor(pixt4, pixt4, pixt3);
boxa2 = pixConnComp(pixt4, &pixac, 8);
pixaaAddPixa(pixaa, pixac, L_INSERT);
pixaaAddBox(pixaa, box, L_INSERT);
if (DEBUG_OUTPUT) {
pixt5 = pixaDisplayRandomCmap(pixac, 0, 0);
pixDisplayWrite(pixt5, DEBUG_OUTPUT);
fprintf(stderr, "Num textlines in col %d: %d\n", i,
boxaGetCount(boxa2));
pixDestroy(&pixt5);
}
pixDestroy(&pixt3);
pixDestroy(&pixt4);
boxaDestroy(&boxa2);
}
/* Visual output */
if (DEBUG_OUTPUT) {
pixDisplayMultiple("/tmp/junk_write_display*");
pixat = pixaReadFiles("/tmp", "junk_write_display");
pixt5 = selDisplayInPix(selsplit, 31, 2);
pixaAddPix(pixat, pixt5, L_INSERT);
pixt6 = pixaDisplayTiledAndScaled(pixat, 32, 400, 3, 0, 35, 3);
pixWrite(fileout, pixt6, IFF_PNG);
pixaDestroy(&pixat);
pixDestroy(&pixt6);
}
/* Test pixaa I/O */
pixaaWrite("/tmp/junkpixaa", pixaa);
pixaa2 = pixaaRead("/tmp/junkpixaa");
pixaaWrite("/tmp/junkpixaa2", pixaa2);
/* Test pixaa display */
pixd = pixaaDisplay(pixaa, w2, h2);
pixWrite("/tmp/junkdisplay", pixd, IFF_PNG);
pixDestroy(&pixd);
/* Cleanup */
pixDestroy(&pixb2);
pixDestroy(&pixt1);
pixaDestroy(&pixam);
pixaDestroy(&pixad);
pixaaDestroy(&pixaa);
pixaaDestroy(&pixaa2);
boxaDestroy(&boxa);
selDestroy(&selsplit);
#endif
#if 0
/* Use the baseline finder; not really what is needed */
numa = pixFindBaselines(pixb, &pta, 1);
#endif
#if 0
/* Use the textline mask function; parameters are not quite right */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixtlm = pixGenTextlineMask(pixb2, &pixvws, NULL, 1);
pixDisplay(pixtlm, 0, 100);
pixDisplay(pixvws, 500, 100);
pixDestroy(&pixb2);
pixDestroy(&pixtlm);
pixDestroy(&pixvws);
#endif
#if 0
/* Use the Breuel whitespace partition method; slow and we would
* still need to work to extract the fg regions. */
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
boxas = pixConnComp(pixb2, NULL, 8);
boxad = boxaGetWhiteblocks(boxas, NULL, L_SORT_BY_HEIGHT,
3, 0.1, 200, 0.2, 0);
pixd = pixDrawBoxa(pixb2, boxad, 7, 0xe0708000);
pixDisplay(pixd, 100, 500);
pixDestroy(&pixb2);
pixDestroy(&pixd);
boxaDestroy(&boxas);
boxaDestroy(&boxad);
#endif
#if 0
/* Use morphology to find columns and then selective
* region-based morphology to get the textline mask.
* This is for display; we really want to get a pixa of the
* specific textline masks. */
startTimer();
pixb2 = pixReduceRankBinary2(pixb, 2, NULL);
pixt1 = pixMorphCompSequence(pixb2, "c5.500", 0); /* column mask */
pixt2 = pixMorphSequenceByRegion(pixb2, pixt1, "c100.3", 8, 0, 0, &boxa);
fprintf(stderr, "time = %7.3f sec\n", stopTimer());
pixDisplay(pixt1, 100, 500);
pixDisplay(pixt2, 800, 500);
pixDestroy(&pixb2);
pixDestroy(&pixt1);
pixDestroy(&pixt2);
boxaDestroy(&boxa);
#endif
pixDestroy(&pixs);
pixDestroy(&pixb);
exit(0);
}
/*!
* \brief pixGetWordsInTextlines()
*
* \param[in] pixs 1 bpp, typ. 300 ppi
* \param[in] reduction 1 for input res; 2 for 2x reduction of input res
* \param[in] minwidth, minheight of saved components; smaller are discarded
* \param[in] maxwidth, maxheight of saved components; larger are discarded
* \param[out] pboxad word boxes sorted in textline line order
* \param[out] ppixad word images sorted in textline line order
* \param[out] pnai index of textline for each word
* \return 0 if OK, 1 on error
*
* <pre>
* Notes:
* (1) The input should be at a resolution of about 300 ppi.
* The word masks and word images can be computed at either
* 150 ppi or 300 ppi. For the former, set reduction = 2.
* (2) The four size constraints on saved components are all
* scaled by %reduction.
* (3) The result are word images (and their b.b.), extracted in
* textline order, at either full res or 2x reduction,
* and with a numa giving the textline index for each word.
* (4) The pixa and boxa interfaces should make this type of
* application simple to put together. The steps are:
* ~ optionally reduce by 2x
* ~ generate first estimate of word masks
* ~ get b.b. of these, and remove the small and big ones
* ~ extract pixa of the word images, using the b.b.
* ~ sort actual word images in textline order (2d)
* ~ flatten them to a pixa (1d), saving the textline index
* for each pix
* (5) In an actual application, it may be desirable to pre-filter
* the input image to remove large components, to extract
* single columns of text, and to deskew them. For example,
* to remove both large components and small noisy components
* that can interfere with the statistics used to estimate
* parameters for segmenting by words, but still retain text lines,
* the following image preprocessing can be done:
* Pix *pixt = pixMorphSequence(pixs, "c40.1", 0);
* Pix *pixf = pixSelectBySize(pixt, 0, 60, 8,
* L_SELECT_HEIGHT, L_SELECT_IF_LT, NULL);
* pixAnd(pixf, pixf, pixs); // the filtered image
* The closing turns text lines into long blobs, but does not
* significantly increase their height. But if there are many
* small connected components in a dense texture, this is likely
* to generate tall components that will be eliminated in pixf.
* </pre>
*/
l_int32
pixGetWordsInTextlines(PIX *pixs,
l_int32 reduction,
l_int32 minwidth,
l_int32 minheight,
l_int32 maxwidth,
l_int32 maxheight,
BOXA **pboxad,
PIXA **ppixad,
NUMA **pnai)
{
l_int32 maxdil;
BOXA *boxa1, *boxad;
BOXAA *baa;
NUMA *nai;
NUMAA *naa;
PIXA *pixa1, *pixad;
PIX *pix1;
PIXAA *paa;
PROCNAME("pixGetWordsInTextlines");
if (!pboxad || !ppixad || !pnai)
return ERROR_INT("&boxad, &pixad, &nai not all defined", procName, 1);
*pboxad = NULL;
*ppixad = NULL;
*pnai = NULL;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (reduction != 1 && reduction != 2)
return ERROR_INT("reduction not in {1,2}", procName, 1);
if (reduction == 1) {
pix1 = pixClone(pixs);
maxdil = 18;
} else { /* reduction == 2 */
pix1 = pixReduceRankBinaryCascade(pixs, 1, 0, 0, 0);
maxdil = 9;
}
/* Get the bounding boxes of the words from the word mask. */
pixWordBoxesByDilation(pix1, maxdil, minwidth, minheight,
maxwidth, maxheight, &boxa1, NULL);
/* Generate a pixa of the word images */
pixa1 = pixaCreateFromBoxa(pix1, boxa1, NULL); /* mask over each word */
/* Sort the bounding boxes of these words by line. We use the
* index mapping to allow identical sorting of the pixa. */
baa = boxaSort2d(boxa1, &naa, -1, -1, 4);
paa = pixaSort2dByIndex(pixa1, naa, L_CLONE);
/* Flatten the word paa */
pixad = pixaaFlattenToPixa(paa, &nai, L_CLONE);
boxad = pixaGetBoxa(pixad, L_COPY);
*pnai = nai;
*pboxad = boxad;
*ppixad = pixad;
pixDestroy(&pix1);
pixaDestroy(&pixa1);
boxaDestroy(&boxa1);
boxaaDestroy(&baa);
pixaaDestroy(&paa);
numaaDestroy(&naa);
return 0;
}
l_int32 main(int argc,
char **argv)
{
char buf[256], dirname[256];
char *dirin, *pattern, *subdirout, *fname, *tail, *basename;
l_int32 thresh, i, n;
l_float32 scalefactor;
PIX *pix1, *pix2, *pix3, *pix4;
SARRAY *sa;
static char mainName[] = "binarizefiles.c";
if (argc != 6) {
fprintf(stderr,
"Syntax: binarizefiles dirin pattern thresh scalefact dirout\n"
" dirin: input directory for image files\n"
" pattern: use 'allfiles' to convert all files\n"
" in the directory\n"
" thresh: 0 for adaptive; > 0 for global thresh (e.g., 128)\n"
" scalefactor: in (0.0 ... 4.0]; use 1.0 to prevent scaling\n"
" subdirout: subdirectory of /tmp for output files\n");
return 1;
}
dirin = argv[1];
pattern = argv[2];
thresh = atoi(argv[3]);
scalefactor = atof(argv[4]);
subdirout = argv[5];
if (!strcmp(pattern, "allfiles"))
pattern = NULL;
if (scalefactor <= 0.0 || scalefactor > 4.0) {
L_WARNING("invalid scalefactor: setting to 1.0\n", mainName);
scalefactor = 1.0;
}
/* Get the input filenames */
sa = getSortedPathnamesInDirectory(dirin, pattern, 0, 0);
sarrayWriteStream(stderr, sa);
n = sarrayGetCount(sa);
/* Write the output files */
makeTempDirname(dirname, 256, subdirout);
fprintf(stderr, "dirname: %s\n", dirname);
lept_mkdir(subdirout);
for (i = 0; i < n; i++) {
fname = sarrayGetString(sa, i, L_NOCOPY);
if ((pix1 = pixRead(fname)) == NULL) {
L_ERROR("file %s not read as image", mainName, fname);
continue;
}
splitPathAtDirectory(fname, NULL, &tail);
splitPathAtExtension(tail, &basename, NULL);
snprintf(buf, sizeof(buf), "%s/%s.tif", dirname, basename);
FREE(tail);
FREE(basename);
fprintf(stderr, "fileout: %s\n", buf);
if (scalefactor != 1.0)
pix2 = pixScale(pix1, scalefactor, scalefactor);
else
pix2 = pixClone(pix1);
if (thresh == 0) {
pix4 = pixConvertTo8(pix2, 0);
pix3 = pixAdaptThresholdToBinary(pix4, NULL, 1.0);
pixDestroy(&pix4);
} else {
pix3 = pixConvertTo1(pix2, thresh);
}
pixWrite(buf, pix3, IFF_TIFF_G4);
pixDestroy(&pix1);
pixDestroy(&pix2);
pixDestroy(&pix3);
}
sarrayDestroy(&sa);
return 0;
}
/*!
* \brief pixDisplayWriteFormat()
*
* \param[in] pix 1, 2, 4, 8, 16, 32 bpp
* \param[in] reduction -1 to erase and reset; 0 to disable;
* otherwise this is a reduction factor
* \param[in] format IFF_DEFAULT or IFF_PNG
* \return 0 if OK; 1 on error
*
* <pre>
* Notes:
* (1) This writes files with pathnames "/tmp/lept/display/file.*"
* if reduction \> 0. These can be collected into a pdf using
* pixDisplayMultiple();
* (2) Before writing a set of files, call
* pixDisplayWrite(NULL, -1);
* This erases any previously written files in that directory.
* (3) If reduction \> 1 and depth == 1, this does a scale-to-gray
* reduction.
* (4) This function uses a static internal variable to number
* output files written by a single process. Behavior
* with a shared library may be unpredictable.
* (5) Output file format is as follows:
* format == IFF_DEFAULT:
* png if d \< 8 or d == 16 or if the output pix
* has a colormap. Otherwise, output is jpg.
* format == IFF_PNG:
* png (lossless) on all images.
* (6) For 16 bpp, the choice of full dynamic range with log scale
* is the best for displaying these images. Alternative outputs are
* pix8 = pixMaxDynamicRange(pixt, L_LINEAR_SCALE);
* pix8 = pixConvert16To8(pixt, 0); // low order byte
* pix8 = pixConvert16To8(pixt, 1); // high order byte
* </pre>
*/
l_int32
pixDisplayWriteFormat(PIX *pixs,
l_int32 reduction,
l_int32 format)
{
char buf[L_BUF_SIZE];
char *fname;
l_float32 scale;
PIX *pix1, *pix2;
static l_int32 index = 0; /* caution: not .so or thread safe */
PROCNAME("pixDisplayWriteFormat");
if (reduction == 0) return 0;
if (reduction < 0) { /* initialize */
lept_rmdir("lept/display");
index = 0;
return 0;
}
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (format != IFF_DEFAULT && format != IFF_PNG) {
L_INFO("invalid format; using default\n", procName);
format = IFF_DEFAULT;
}
if (index == 0)
lept_mkdir("lept/display");
index++;
if (reduction == 1) {
pix1 = pixClone(pixs);
} else {
scale = 1. / (l_float32)reduction;
if (pixGetDepth(pixs) == 1)
pix1 = pixScaleToGray(pixs, scale);
else
pix1 = pixScale(pixs, scale, scale);
}
if (pixGetDepth(pix1) == 16) {
pix2 = pixMaxDynamicRange(pix1, L_LOG_SCALE);
snprintf(buf, L_BUF_SIZE, "file.%03d.png", index);
fname = genPathname("/tmp/lept/display", buf);
pixWrite(fname, pix2, IFF_PNG);
pixDestroy(&pix2);
} else if (pixGetDepth(pix1) < 8 || pixGetColormap(pix1) ||
format == IFF_PNG) {
snprintf(buf, L_BUF_SIZE, "file.%03d.png", index);
fname = genPathname("/tmp/lept/display", buf);
pixWrite(fname, pix1, IFF_PNG);
} else {
snprintf(buf, L_BUF_SIZE, "file.%03d.jpg", index);
fname = genPathname("/tmp/lept/display", buf);
pixWrite(fname, pix1, format);
}
LEPT_FREE(fname);
pixDestroy(&pix1);
return 0;
}
// On setting the input image, a clone of it is owned by this class.
void ShiroRekhaSplitter::set_orig_pix(Pix* pix) {
if (orig_pix_) {
pixDestroy(&orig_pix_);
}
orig_pix_ = pixClone(pix);
}
/*!
* \brief pixDisplayWithTitle()
*
* \param[in] pix 1, 2, 4, 8, 16, 32 bpp
* \param[in] x, y location of display frame
* \param[in] title [optional] on frame; can be NULL;
* \param[in] dispflag 1 to write, else disabled
* \return 0 if OK; 1 on error
*
* <pre>
* Notes:
* (1) See notes for pixDisplay().
* (2) This displays the image if dispflag == 1.
* </pre>
*/
l_int32
pixDisplayWithTitle(PIX *pixs,
l_int32 x,
l_int32 y,
const char *title,
l_int32 dispflag)
{
char *tempname;
char buffer[L_BUF_SIZE];
static l_int32 index = 0; /* caution: not .so or thread safe */
l_int32 w, h, d, spp, maxheight, opaque, threeviews, ignore;
l_float32 ratw, rath, ratmin;
PIX *pix0, *pix1, *pix2;
PIXCMAP *cmap;
#ifndef _WIN32
l_int32 wt, ht;
#else
char *pathname;
char fullpath[_MAX_PATH];
#endif /* _WIN32 */
PROCNAME("pixDisplayWithTitle");
if (dispflag != 1) return 0;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (var_DISPLAY_PROG != L_DISPLAY_WITH_XZGV &&
var_DISPLAY_PROG != L_DISPLAY_WITH_XLI &&
var_DISPLAY_PROG != L_DISPLAY_WITH_XV &&
var_DISPLAY_PROG != L_DISPLAY_WITH_IV &&
var_DISPLAY_PROG != L_DISPLAY_WITH_OPEN) {
return ERROR_INT("no program chosen for display", procName, 1);
}
/* Display with three views if either spp = 4 or if colormapped
* and the alpha component is not fully opaque */
opaque = TRUE;
if ((cmap = pixGetColormap(pixs)) != NULL)
pixcmapIsOpaque(cmap, &opaque);
spp = pixGetSpp(pixs);
threeviews = (spp == 4 || !opaque) ? TRUE : FALSE;
/* If colormapped and not opaque, remove the colormap to RGBA */
if (!opaque)
pix0 = pixRemoveColormap(pixs, REMOVE_CMAP_WITH_ALPHA);
else
pix0 = pixClone(pixs);
/* Scale if necessary; this will also remove a colormap */
pixGetDimensions(pix0, &w, &h, &d);
maxheight = (threeviews) ? MAX_DISPLAY_HEIGHT / 3 : MAX_DISPLAY_HEIGHT;
if (w <= MAX_DISPLAY_WIDTH && h <= maxheight) {
if (d == 16) /* take MSB */
pix1 = pixConvert16To8(pix0, 1);
else
pix1 = pixClone(pix0);
} else {
ratw = (l_float32)MAX_DISPLAY_WIDTH / (l_float32)w;
rath = (l_float32)maxheight / (l_float32)h;
ratmin = L_MIN(ratw, rath);
if (ratmin < 0.125 && d == 1)
pix1 = pixScaleToGray8(pix0);
else if (ratmin < 0.25 && d == 1)
pix1 = pixScaleToGray4(pix0);
else if (ratmin < 0.33 && d == 1)
pix1 = pixScaleToGray3(pix0);
else if (ratmin < 0.5 && d == 1)
pix1 = pixScaleToGray2(pix0);
else
pix1 = pixScale(pix0, ratmin, ratmin);
}
pixDestroy(&pix0);
if (!pix1)
return ERROR_INT("pix1 not made", procName, 1);
/* Generate the three views if required */
if (threeviews)
pix2 = pixDisplayLayersRGBA(pix1, 0xffffff00, 0);
else
pix2 = pixClone(pix1);
if (index == 0) {
lept_rmdir("lept/disp");
lept_mkdir("lept/disp");
}
index++;
if (pixGetDepth(pix2) < 8 ||
(w < MAX_SIZE_FOR_PNG && h < MAX_SIZE_FOR_PNG)) {
snprintf(buffer, L_BUF_SIZE, "/tmp/lept/disp/write.%03d.png", index);
pixWrite(buffer, pix2, IFF_PNG);
} else {
snprintf(buffer, L_BUF_SIZE, "/tmp/lept/disp/write.%03d.jpg", index);
pixWrite(buffer, pix2, IFF_JFIF_JPEG);
}
tempname = genPathname(buffer, NULL);
#ifndef _WIN32
/* Unix */
if (var_DISPLAY_PROG == L_DISPLAY_WITH_XZGV) {
/* no way to display title */
pixGetDimensions(pix2, &wt, &ht, NULL);
snprintf(buffer, L_BUF_SIZE,
"xzgv --geometry %dx%d+%d+%d %s &", wt + 10, ht + 10,
x, y, tempname);
} else if (var_DISPLAY_PROG == L_DISPLAY_WITH_XLI) {
if (title) {
snprintf(buffer, L_BUF_SIZE,
"xli -dispgamma 1.0 -quiet -geometry +%d+%d -title \"%s\" %s &",
x, y, title, tempname);
} else {
snprintf(buffer, L_BUF_SIZE,
"xli -dispgamma 1.0 -quiet -geometry +%d+%d %s &",
x, y, tempname);
}
} else if (var_DISPLAY_PROG == L_DISPLAY_WITH_XV) {
if (title) {
snprintf(buffer, L_BUF_SIZE,
"xv -quit -geometry +%d+%d -name \"%s\" %s &",
x, y, title, tempname);
} else {
snprintf(buffer, L_BUF_SIZE,
"xv -quit -geometry +%d+%d %s &", x, y, tempname);
}
} else if (var_DISPLAY_PROG == L_DISPLAY_WITH_OPEN) {
snprintf(buffer, L_BUF_SIZE, "open %s &", tempname);
}
ignore = system(buffer);
#else /* _WIN32 */
/* Windows: L_DISPLAY_WITH_IV */
pathname = genPathname(tempname, NULL);
_fullpath(fullpath, pathname, sizeof(fullpath));
if (title) {
snprintf(buffer, L_BUF_SIZE,
"i_view32.exe \"%s\" /pos=(%d,%d) /title=\"%s\"",
fullpath, x, y, title);
} else {
snprintf(buffer, L_BUF_SIZE, "i_view32.exe \"%s\" /pos=(%d,%d)",
fullpath, x, y);
}
ignore = system(buffer);
LEPT_FREE(pathname);
#endif /* _WIN32 */
pixDestroy(&pix1);
pixDestroy(&pix2);
LEPT_FREE(tempname);
return 0;
}
// Top-level method to perform splitting based on current settings.
// Returns true if a split was actually performed.
// split_for_pageseg should be true if the splitting is being done prior to
// page segmentation. This mode uses the flag
// pageseg_devanagari_split_strategy to determine the splitting strategy.
bool ShiroRekhaSplitter::Split(bool split_for_pageseg) {
SplitStrategy split_strategy = split_for_pageseg ? pageseg_split_strategy_ :
ocr_split_strategy_;
if (split_strategy == NO_SPLIT) {
return false; // Nothing to do.
}
ASSERT_HOST(split_strategy == MINIMAL_SPLIT ||
split_strategy == MAXIMAL_SPLIT);
ASSERT_HOST(orig_pix_);
if (devanagari_split_debuglevel > 0) {
tprintf("Splitting shiro-rekha ...\n");
tprintf("Split strategy = %s\n",
split_strategy == MINIMAL_SPLIT ? "Minimal" : "Maximal");
tprintf("Initial pageseg available = %s\n",
segmentation_block_list_ ? "yes" : "no");
}
// Create a copy of original image to store the splitting output.
pixDestroy(&splitted_image_);
splitted_image_ = pixCopy(NULL, orig_pix_);
// Initialize debug image if required.
if (devanagari_split_debugimage) {
pixDestroy(&debug_image_);
debug_image_ = pixConvertTo32(orig_pix_);
}
// Determine all connected components in the input image. A close operation
// may be required prior to this, depending on the current settings.
Pix* pix_for_ccs = pixClone(orig_pix_);
if (perform_close_ && global_xheight_ != kUnspecifiedXheight &&
!segmentation_block_list_) {
if (devanagari_split_debuglevel > 0) {
tprintf("Performing a global close operation..\n");
}
// A global measure is available for xheight, but no local information
// exists.
pixDestroy(&pix_for_ccs);
pix_for_ccs = pixCopy(NULL, orig_pix_);
PerformClose(pix_for_ccs, global_xheight_);
}
Pixa* ccs;
Boxa* tmp_boxa = pixConnComp(pix_for_ccs, &ccs, 8);
boxaDestroy(&tmp_boxa);
pixDestroy(&pix_for_ccs);
// Iterate over all connected components. Get their bounding boxes and clip
// out the image regions corresponding to these boxes from the original image.
// Conditionally run splitting on each of them.
Boxa* regions_to_clear = boxaCreate(0);
for (int i = 0; i < pixaGetCount(ccs); ++i) {
Box* box = ccs->boxa->box[i];
Pix* word_pix = pixClipRectangle(orig_pix_, box, NULL);
ASSERT_HOST(word_pix);
int xheight = GetXheightForCC(box);
if (xheight == kUnspecifiedXheight && segmentation_block_list_ &&
devanagari_split_debugimage) {
pixRenderBoxArb(debug_image_, box, 1, 255, 0, 0);
}
// If some xheight measure is available, attempt to pre-eliminate small
// blobs from the shiro-rekha process. This is primarily to save the CCs
// corresponding to punctuation marks/small dots etc which are part of
// larger graphemes.
if (xheight == kUnspecifiedXheight ||
(box->w > xheight / 3 && box->h > xheight / 2)) {
SplitWordShiroRekha(split_strategy, word_pix, xheight,
box->x, box->y, regions_to_clear);
} else if (devanagari_split_debuglevel > 0) {
tprintf("CC dropped from splitting: %d,%d (%d, %d)\n",
box->x, box->y, box->w, box->h);
}
pixDestroy(&word_pix);
}
// Actually clear the boxes now.
for (int i = 0; i < boxaGetCount(regions_to_clear); ++i) {
Box* box = boxaGetBox(regions_to_clear, i, L_CLONE);
pixClearInRect(splitted_image_, box);
boxDestroy(&box);
}
boxaDestroy(®ions_to_clear);
pixaDestroy(&ccs);
if (devanagari_split_debugimage) {
DumpDebugImage(split_for_pageseg ? "pageseg_split_debug.png" :
"ocr_split_debug.png");
}
return true;
}
/*!
* pixGenerateSelWithRuns()
*
* Input: pix (1 bpp, typically small, to be used as a pattern)
* nhlines (number of hor lines along which elements are found)
* nvlines (number of vert lines along which elements are found)
* distance (min distance from boundary pixel; use 0 for default)
* minlength (min runlength to set hit or miss; use 0 for default)
* toppix (number of extra pixels of bg added above)
* botpix (number of extra pixels of bg added below)
* leftpix (number of extra pixels of bg added to left)
* rightpix (number of extra pixels of bg added to right)
* &pixe (<optional return> input pix expanded by extra pixels)
* Return: sel (hit-miss for input pattern), or null on error
*
* Notes:
* (1) The horizontal and vertical lines along which elements are
* selected are roughly equally spaced. The actual locations of
* the hits and misses are the centers of respective run-lengths.
* (2) No elements are selected that are less than 'distance' pixels away
* from a boundary pixel of the same color. This makes the
* match much more robust to edge noise. Valid inputs of
* 'distance' are 0, 1, 2, 3 and 4. If distance is either 0 or
* greater than 4, we reset it to the default value.
* (3) The 4 numbers for adding rectangles of pixels outside the fg
* can be use if the pattern is expected to be surrounded by bg
* (white) pixels. On the other hand, if the pattern may be near
* other fg (black) components on some sides, use 0 for those sides.
* (4) The pixels added to a side allow you to have miss elements there.
* There is a constraint between distance, minlength, and
* the added pixels for this to work. We illustrate using the
* default values. If you add 5 pixels to the top, and use a
* distance of 1, then you end up with a vertical run of at least
* 4 bg pixels along the top edge of the image. If you use a
* minimum runlength of 3, each vertical line will always find
* a miss near the center of its run. However, if you use a
* minimum runlength of 5, you will not get a miss on every vertical
* line. As another example, if you have 7 added pixels and a
* distance of 2, you can use a runlength up to 5 to guarantee
* that the miss element is recorded. We give a warning if the
* contraint does not guarantee a miss element outside the
* image proper.
* (5) The input pix, as extended by the extra pixels on selected sides,
* can optionally be returned. For debugging, call
* pixDisplayHitMissSel() to visualize the hit-miss sel superimposed
* on the generating bitmap.
*/
SEL *
pixGenerateSelWithRuns(PIX *pixs,
l_int32 nhlines,
l_int32 nvlines,
l_int32 distance,
l_int32 minlength,
l_int32 toppix,
l_int32 botpix,
l_int32 leftpix,
l_int32 rightpix,
PIX **ppixe)
{
l_int32 ws, hs, w, h, x, y, xval, yval, i, j, nh, nm;
l_float32 delh, delw;
NUMA *nah, *nam;
PIX *pixt1, *pixt2, *pixfg, *pixbg;
PTA *ptah, *ptam;
SEL *seld, *sel;
PROCNAME("pixGenerateSelWithRuns");
if (ppixe) *ppixe = NULL;
if (!pixs)
return (SEL *)ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (SEL *)ERROR_PTR("pixs not 1 bpp", procName, NULL);
if (nhlines < 1 && nvlines < 1)
return (SEL *)ERROR_PTR("nvlines and nhlines both < 1", procName, NULL);
if (distance <= 0)
distance = DEFAULT_DISTANCE_TO_BOUNDARY;
if (minlength <= 0)
minlength = DEFAULT_MIN_RUNLENGTH;
if (distance > MAX_DISTANCE_TO_BOUNDARY) {
L_WARNING("distance too large; setting to max value", procName);
distance = MAX_DISTANCE_TO_BOUNDARY;
}
/* Locate the foreground */
pixClipToForeground(pixs, &pixt1, NULL);
if (!pixt1)
return (SEL *)ERROR_PTR("pixt1 not made", procName, NULL);
ws = pixGetWidth(pixt1);
hs = pixGetHeight(pixt1);
w = ws;
h = hs;
/* Crop out a region including the foreground, and add pixels
* on sides depending on the side flags */
if (toppix || botpix || leftpix || rightpix) {
x = y = 0;
if (toppix) {
h += toppix;
y = toppix;
if (toppix < distance + minlength)
L_WARNING("no miss elements in added top pixels", procName);
}
if (botpix) {
h += botpix;
if (botpix < distance + minlength)
L_WARNING("no miss elements in added bot pixels", procName);
}
if (leftpix) {
w += leftpix;
x = leftpix;
if (leftpix < distance + minlength)
L_WARNING("no miss elements in added left pixels", procName);
}
if (rightpix) {
w += rightpix;
if (rightpix < distance + minlength)
L_WARNING("no miss elements in added right pixels", procName);
}
pixt2 = pixCreate(w, h, 1);
pixRasterop(pixt2, x, y, ws, hs, PIX_SRC, pixt1, 0, 0);
}
else
pixt2 = pixClone(pixt1);
if (ppixe)
*ppixe = pixClone(pixt2);
pixDestroy(&pixt1);
/* Identify fg and bg pixels that are at least 'distance' pixels
* away from the boundary pixels in their set */
seld = selCreateBrick(2 * distance + 1, 2 * distance + 1,
distance, distance, SEL_HIT);
pixfg = pixErode(NULL, pixt2, seld);
pixbg = pixDilate(NULL, pixt2, seld);
pixInvert(pixbg, pixbg);
selDestroy(&seld);
pixDestroy(&pixt2);
/* Accumulate hit and miss points */
ptah = ptaCreate(0);
ptam = ptaCreate(0);
if (nhlines >= 1) {
delh = (l_float32)h / (l_float32)(nhlines + 1);
for (i = 0, y = 0; i < nhlines; i++) {
y += (l_int32)(delh + 0.5);
nah = pixGetRunCentersOnLine(pixfg, -1, y, minlength);
nam = pixGetRunCentersOnLine(pixbg, -1, y, minlength);
nh = numaGetCount(nah);
nm = numaGetCount(nam);
for (j = 0; j < nh; j++) {
numaGetIValue(nah, j, &xval);
ptaAddPt(ptah, xval, y);
}
for (j = 0; j < nm; j++) {
numaGetIValue(nam, j, &xval);
ptaAddPt(ptam, xval, y);
}
numaDestroy(&nah);
numaDestroy(&nam);
}
}
if (nvlines >= 1) {
delw = (l_float32)w / (l_float32)(nvlines + 1);
for (i = 0, x = 0; i < nvlines; i++) {
x += (l_int32)(delw + 0.5);
nah = pixGetRunCentersOnLine(pixfg, x, -1, minlength);
nam = pixGetRunCentersOnLine(pixbg, x, -1, minlength);
nh = numaGetCount(nah);
nm = numaGetCount(nam);
for (j = 0; j < nh; j++) {
numaGetIValue(nah, j, &yval);
ptaAddPt(ptah, x, yval);
}
for (j = 0; j < nm; j++) {
numaGetIValue(nam, j, &yval);
ptaAddPt(ptam, x, yval);
}
numaDestroy(&nah);
numaDestroy(&nam);
}
}
/* Make the Sel with those points */
sel = selCreateBrick(h, w, h / 2, w / 2, SEL_DONT_CARE);
nh = ptaGetCount(ptah);
for (i = 0; i < nh; i++) {
ptaGetIPt(ptah, i, &x, &y);
selSetElement(sel, y, x, SEL_HIT);
}
nm = ptaGetCount(ptam);
for (i = 0; i < nm; i++) {
ptaGetIPt(ptam, i, &x, &y);
selSetElement(sel, y, x, SEL_MISS);
}
pixDestroy(&pixfg);
pixDestroy(&pixbg);
ptaDestroy(&ptah);
ptaDestroy(&ptam);
return sel;
}
/*!
* pixSaveTiledOutline()
*
* Input: pixs (1, 2, 4, 8, 32 bpp)
* pixa (the pix are accumulated here)
* reduction (0 to disable; otherwise this is a reduction factor)
* newrow (0 if placed on the same row as previous; 1 otherwise)
* space (horizontal and vertical spacing, in pixels)
* linewidth (width of added outline for image; 0 for no outline)
* dp (depth of pixa; 8 or 32 bpp; only used on first call)
* Return: 0 if OK, 1 on error.
*
* Notes:
* (1) Before calling this function for the first time, use
* pixaCreate() to make the @pixa that will accumulate the pix.
* This is passed in each time pixSaveTiled() is called.
* (2) @reduction is the integer reduction factor for the input
* image. After reduction and possible depth conversion,
* the image is saved in the input pixa, along with a box
* that specifies the location to place it when tiled later.
* Disable saving the pix by setting reduction == 0.
* (3) @newrow and @space specify the location of the new pix
* with respect to the last one(s) that were entered.
* (4) @dp specifies the depth at which all pix are saved. It can
* be only 8 or 32 bpp. Any colormap is removed. This is only
* used at the first invocation.
* (5) This function uses two variables from call to call.
* If they were static, the function would not be .so or thread
* safe, and furthermore, there would be interference with two or
* more pixa accumulating images at a time. Consequently,
* we use the first pix in the pixa to store and obtain both
* the depth and the current position of the bottom (one pixel
* below the lowest image raster line when laid out using
* the boxa). The bottom variable is stored in the input format
* field, which is the only field available for storing an int.
*/
l_int32
pixSaveTiledOutline(PIX *pixs,
PIXA *pixa,
l_int32 reduction,
l_int32 newrow,
l_int32 space,
l_int32 linewidth,
l_int32 dp)
{
l_int32 n, top, left, bx, by, bw, w, h, depth, bottom;
l_float32 scale;
BOX *box;
PIX *pix, *pixt1, *pixt2, *pixt3;
PROCNAME("pixSaveTiledOutline");
if (reduction == 0) return 0;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (!pixa)
return ERROR_INT("pixa not defined", procName, 1);
n = pixaGetCount(pixa);
if (n == 0) {
bottom = 0;
if (dp != 8 && dp != 32) {
L_WARNING("dp not 8 or 32 bpp; using 32", procName);
depth = 32;
} else
depth = dp;
}
else { /* extract the depth and bottom params from the first pix */
pix = pixaGetPix(pixa, 0, L_CLONE);
depth = pixGetDepth(pix);
bottom = pixGetInputFormat(pix); /* not typical usage! */
pixDestroy(&pix);
}
/* Scale and convert to output depth */
if (reduction == 1)
pixt1 = pixClone(pixs);
else {
scale = 1. / (l_float32)reduction;
if (pixGetDepth(pixs) == 1)
pixt1 = pixScaleToGray(pixs, scale);
else
pixt1 = pixScale(pixs, scale, scale);
}
if (depth == 8)
pixt2 = pixConvertTo8(pixt1, 0);
else
pixt2 = pixConvertTo32(pixt1);
pixDestroy(&pixt1);
/* Add black outline */
if (linewidth > 0)
pixt3 = pixAddBorder(pixt2, linewidth, 0);
else
pixt3 = pixClone(pixt2);
pixDestroy(&pixt2);
/* Find position of current pix (UL corner plus size) */
if (n == 0) {
top = 0;
left = 0;
}
else if (newrow == 1) {
top = bottom + space;
left = 0;
}
else if (n > 0) {
pixaGetBoxGeometry(pixa, n - 1, &bx, &by, &bw, NULL);
top = by;
left = bx + bw + space;
}
pixGetDimensions(pixt3, &w, &h, NULL);
bottom = L_MAX(bottom, top + h);
box = boxCreate(left, top, w, h);
pixaAddPix(pixa, pixt3, L_INSERT);
pixaAddBox(pixa, box, L_INSERT);
/* Save the new bottom value */
pix = pixaGetPix(pixa, 0, L_CLONE);
pixSetInputFormat(pix, bottom); /* not typical usage! */
pixDestroy(&pix);
return 0;
}
/*!
* pixGenerateSelBoundary()
*
* Input: pix (1 bpp, typically small, to be used as a pattern)
* hitdist (min distance from fg boundary pixel)
* missdist (min distance from bg boundary pixel)
* hitskip (number of boundary pixels skipped between hits)
* missskip (number of boundary pixels skipped between misses)
* topflag (flag for extra pixels of bg added above)
* botflag (flag for extra pixels of bg added below)
* leftflag (flag for extra pixels of bg added to left)
* rightflag (flag for extra pixels of bg added to right)
* &pixe (<optional return> input pix expanded by extra pixels)
* Return: sel (hit-miss for input pattern), or null on error
*
* Notes:
* (1) All fg elements selected are exactly hitdist pixels away from
* the nearest fg boundary pixel, and ditto for bg elements.
* Valid inputs of hitdist and missdist are 0, 1, 2, 3 and 4.
* For example, a hitdist of 0 puts the hits at the fg boundary.
* Usually, the distances should be > 0 avoid the effect of
* noise at the boundary.
* (2) Set hitskip < 0 if no hits are to be used. Ditto for missskip.
* If both hitskip and missskip are < 0, the sel would be empty,
* and NULL is returned.
* (3) The 4 flags determine whether the sel is increased on that side
* to allow bg misses to be placed all along that boundary.
* The increase in sel size on that side is the minimum necessary
* to allow the misses to be placed at mindist. For text characters,
* the topflag and botflag are typically set to 1, and the leftflag
* and rightflag to 0.
* (4) The input pix, as extended by the extra pixels on selected sides,
* can optionally be returned. For debugging, call
* pixDisplayHitMissSel() to visualize the hit-miss sel superimposed
* on the generating bitmap.
* (5) This is probably the best of the three sel generators, in the
* sense that you have the most flexibility with the smallest number
* of hits and misses.
*/
SEL *
pixGenerateSelBoundary(PIX *pixs,
l_int32 hitdist,
l_int32 missdist,
l_int32 hitskip,
l_int32 missskip,
l_int32 topflag,
l_int32 botflag,
l_int32 leftflag,
l_int32 rightflag,
PIX **ppixe)
{
l_int32 ws, hs, w, h, x, y, ix, iy, i, npt;
PIX *pixt1, *pixt2, *pixt3, *pixfg, *pixbg;
SEL *selh, *selm, *sel_3, *sel;
PTA *ptah, *ptam;
PROCNAME("pixGenerateSelBoundary");
if (ppixe) *ppixe = NULL;
if (!pixs)
return (SEL *)ERROR_PTR("pixs not defined", procName, NULL);
if (pixGetDepth(pixs) != 1)
return (SEL *)ERROR_PTR("pixs not 1 bpp", procName, NULL);
if (hitdist < 0 || hitdist > 4 || missdist < 0 || missdist > 4)
return (SEL *)ERROR_PTR("dist not in {0 .. 4}", procName, NULL);
if (hitskip < 0 && missskip < 0)
return (SEL *)ERROR_PTR("no hits or misses", procName, NULL);
/* Locate the foreground */
pixClipToForeground(pixs, &pixt1, NULL);
if (!pixt1)
return (SEL *)ERROR_PTR("pixt1 not made", procName, NULL);
ws = pixGetWidth(pixt1);
hs = pixGetHeight(pixt1);
w = ws;
h = hs;
/* Crop out a region including the foreground, and add pixels
* on sides depending on the side flags */
if (topflag || botflag || leftflag || rightflag) {
x = y = 0;
if (topflag) {
h += missdist + 1;
y = missdist + 1;
}
if (botflag)
h += missdist + 1;
if (leftflag) {
w += missdist + 1;
x = missdist + 1;
}
if (rightflag)
w += missdist + 1;
pixt2 = pixCreate(w, h, 1);
pixRasterop(pixt2, x, y, ws, hs, PIX_SRC, pixt1, 0, 0);
}
else {
pixt2 = pixClone(pixt1);
}
if (ppixe)
*ppixe = pixClone(pixt2);
pixDestroy(&pixt1);
/* Identify fg and bg pixels that are exactly hitdist and
* missdist (rsp) away from the boundary pixels in their set.
* Then get a subsampled set of these points. */
sel_3 = selCreateBrick(3, 3, 1, 1, SEL_HIT);
if (hitskip >= 0) {
selh = selCreateBrick(2 * hitdist + 1, 2 * hitdist + 1,
hitdist, hitdist, SEL_HIT);
pixt3 = pixErode(NULL, pixt2, selh);
pixfg = pixErode(NULL, pixt3, sel_3);
pixXor(pixfg, pixfg, pixt3);
ptah = pixSubsampleBoundaryPixels(pixfg, hitskip);
pixDestroy(&pixt3);
pixDestroy(&pixfg);
selDestroy(&selh);
}
if (missskip >= 0) {
selm = selCreateBrick(2 * missdist + 1, 2 * missdist + 1,
missdist, missdist, SEL_HIT);
pixt3 = pixDilate(NULL, pixt2, selm);
pixbg = pixDilate(NULL, pixt3, sel_3);
pixXor(pixbg, pixbg, pixt3);
ptam = pixSubsampleBoundaryPixels(pixbg, missskip);
pixDestroy(&pixt3);
pixDestroy(&pixbg);
selDestroy(&selm);
}
selDestroy(&sel_3);
pixDestroy(&pixt2);
/* Generate the hit-miss sel from these point */
sel = selCreateBrick(h, w, h / 2, w / 2, SEL_DONT_CARE);
if (hitskip >= 0) {
npt = ptaGetCount(ptah);
for (i = 0; i < npt; i++) {
ptaGetIPt(ptah, i, &ix, &iy);
selSetElement(sel, iy, ix, SEL_HIT);
}
}
if (missskip >= 0) {
npt = ptaGetCount(ptam);
for (i = 0; i < npt; i++) {
ptaGetIPt(ptam, i, &ix, &iy);
selSetElement(sel, iy, ix, SEL_MISS);
}
}
ptaDestroy(&ptah);
ptaDestroy(&ptam);
return sel;
}
/*!
* pixDisplayWithTitle()
*
* Input: pix (1, 2, 4, 8, 16, 32 bpp)
* x, y (location of display frame)
* title (<optional> on frame; can be NULL);
* dispflag (1 to write, else disabled)
* Return: 0 if OK; 1 on error
*
* Notes:
* (1) See notes for pixDisplay().
* (2) This displays the image if dispflag == 1.
*/
l_int32
pixDisplayWithTitle(PIX *pixs,
l_int32 x,
l_int32 y,
const char *title,
l_int32 dispflag)
{
char *tempname;
char buffer[L_BUF_SIZE];
static l_int32 index = 0; /* caution: not .so or thread safe */
l_int32 w, h, d, ignore;
l_float32 ratw, rath, ratmin;
PIX *pixt;
#ifndef _WIN32
l_int32 wt, ht;
#else
char *pathname;
char fullpath[_MAX_PATH];
#endif /* _WIN32 */
PROCNAME("pixDisplayWithTitle");
if (dispflag != 1) return 0;
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (var_DISPLAY_PROG != L_DISPLAY_WITH_XV &&
var_DISPLAY_PROG != L_DISPLAY_WITH_XLI &&
var_DISPLAY_PROG != L_DISPLAY_WITH_XZGV &&
var_DISPLAY_PROG != L_DISPLAY_WITH_IV)
return ERROR_INT("no program chosen for display", procName, 1);
pixGetDimensions(pixs, &w, &h, &d);
if (w <= MAX_DISPLAY_WIDTH && h <= MAX_DISPLAY_HEIGHT) {
if (d == 16) /* take MSB */
pixt = pixConvert16To8(pixs, 1);
else
pixt = pixClone(pixs);
}
else {
ratw = (l_float32)MAX_DISPLAY_WIDTH / (l_float32)w;
rath = (l_float32)MAX_DISPLAY_HEIGHT / (l_float32)h;
ratmin = L_MIN(ratw, rath);
if (ratmin < 0.125 && d == 1)
pixt = pixScaleToGray8(pixs);
else if (ratmin < 0.25 && d == 1)
pixt = pixScaleToGray4(pixs);
else if (ratmin < 0.33 && d == 1)
pixt = pixScaleToGray3(pixs);
else if (ratmin < 0.5 && d == 1)
pixt = pixScaleToGray2(pixs);
else
pixt = pixScale(pixs, ratmin, ratmin);
if (!pixt)
return ERROR_INT("pixt not made", procName, 1);
}
if (index == 0) {
lept_rmdir("display");
lept_mkdir("display");
}
index++;
if (pixGetDepth(pixt) < 8 ||
(w < MAX_SIZE_FOR_PNG && h < MAX_SIZE_FOR_PNG)) {
snprintf(buffer, L_BUF_SIZE, "/tmp/display/write.%03d.png", index);
pixWrite(buffer, pixt, IFF_PNG);
}
else {
snprintf(buffer, L_BUF_SIZE, "/tmp/display/write.%03d.jpg", index);
pixWrite(buffer, pixt, IFF_JFIF_JPEG);
}
tempname = stringNew(buffer);
#ifndef _WIN32
/* Unix */
if (var_DISPLAY_PROG == L_DISPLAY_WITH_XV) {
if (title)
snprintf(buffer, L_BUF_SIZE,
"xv -quit -geometry +%d+%d -name \"%s\" %s &",
x, y, title, tempname);
else
snprintf(buffer, L_BUF_SIZE,
"xv -quit -geometry +%d+%d %s &", x, y, tempname);
}
else if (var_DISPLAY_PROG == L_DISPLAY_WITH_XLI) {
if (title)
snprintf(buffer, L_BUF_SIZE,
"xli -dispgamma 1.0 -quiet -geometry +%d+%d -title \"%s\" %s &",
x, y, title, tempname);
else
snprintf(buffer, L_BUF_SIZE,
"xli -dispgamma 1.0 -quiet -geometry +%d+%d %s &",
x, y, tempname);
}
else if (var_DISPLAY_PROG == L_DISPLAY_WITH_XZGV) {
/* no way to display title */
pixGetDimensions(pixt, &wt, &ht, NULL);
snprintf(buffer, L_BUF_SIZE,
"xzgv --geometry %dx%d+%d+%d %s &", wt + 10, ht + 10,
x, y, tempname);
}
ignore = system(buffer);
#else /* _WIN32 */
/* Windows: L_DISPLAY_WITH_IV */
pathname = genPathname(tempname, NULL);
_fullpath(fullpath, pathname, sizeof(fullpath));
if (title)
snprintf(buffer, L_BUF_SIZE,
"i_view32.exe \"%s\" /pos=(%d,%d) /title=\"%s\"",
fullpath, x, y, title);
else
snprintf(buffer, L_BUF_SIZE, "i_view32.exe \"%s\" /pos=(%d,%d)",
fullpath, x, y);
ignore = system(buffer);
FREE(pathname);
#endif /* _WIN32 */
pixDestroy(&pixt);
FREE(tempname);
return 0;
}
/*!
* pixSauvolaBinarize()
*
* Input: pixs (8 bpp grayscale; not colormapped)
* whsize (window half-width for measuring local statistics)
* factor (factor for reducing threshold due to variance; >= 0)
* addborder (1 to add border of width (@whsize + 1) on all sides)
* &pixm (<optional return> local mean values)
* &pixsd (<optional return> local standard deviation values)
* &pixth (<optional return> threshold values)
* &pixd (<optional return> thresholded image)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) The window width and height are 2 * @whsize + 1. The minimum
* value for @whsize is 2; typically it is >= 7..
* (2) The local statistics, measured over the window, are the
* average and standard deviation.
* (3) The measurements of the mean and standard deviation are
* performed inside a border of (@whsize + 1) pixels. If pixs does
* not have these added border pixels, use @addborder = 1 to add
* it here; otherwise use @addborder = 0.
* (4) The Sauvola threshold is determined from the formula:
* t = m * (1 - k * (1 - s / 128))
* where:
* t = local threshold
* m = local mean
* k = @factor (>= 0) [ typ. 0.35 ]
* s = local standard deviation, which is maximized at
* 127.5 when half the samples are 0 and half are 255.
* (5) The basic idea of Niblack and Sauvola binarization is that
* the local threshold should be less than the median value,
* and the larger the variance, the closer to the median
* it should be chosen. Typical values for k are between
* 0.2 and 0.5.
*/
l_int32
pixSauvolaBinarize(PIX *pixs,
l_int32 whsize,
l_float32 factor,
l_int32 addborder,
PIX **ppixm,
PIX **ppixsd,
PIX **ppixth,
PIX **ppixd)
{
l_int32 w, h;
PIX *pixg, *pixsc, *pixm, *pixms, *pixth, *pixd;
PROCNAME("pixSauvolaBinarize");
if (!ppixm && !ppixsd && !ppixth && !ppixd)
return ERROR_INT("no outputs", procName, 1);
if (ppixm) *ppixm = NULL;
if (ppixsd) *ppixsd = NULL;
if (ppixth) *ppixth = NULL;
if (ppixd) *ppixd = NULL;
if (!pixs || pixGetDepth(pixs) != 8)
return ERROR_INT("pixs undefined or not 8 bpp", procName, 1);
if (pixGetColormap(pixs))
return ERROR_INT("pixs is cmapped", procName, 1);
pixGetDimensions(pixs, &w, &h, NULL);
if (whsize < 2)
return ERROR_INT("whsize must be >= 2", procName, 1);
if (w < 2 * whsize + 3 || h < 2 * whsize + 3)
return ERROR_INT("whsize too large for image", procName, 1);
if (factor < 0.0)
return ERROR_INT("factor must be >= 0", procName, 1);
if (addborder) {
pixg = pixAddMirroredBorder(pixs, whsize + 1, whsize + 1,
whsize + 1, whsize + 1);
pixsc = pixClone(pixs);
} else {
pixg = pixClone(pixs);
pixsc = pixRemoveBorder(pixs, whsize + 1);
}
if (!pixg || !pixsc)
return ERROR_INT("pixg and pixsc not made", procName, 1);
/* All these functions strip off the border pixels. */
if (ppixm || ppixth || ppixd)
pixm = pixWindowedMean(pixg, whsize, whsize, 1, 1);
if (ppixsd || ppixth || ppixd)
pixms = pixWindowedMeanSquare(pixg, whsize, whsize, 1);
if (ppixth || ppixd)
pixth = pixSauvolaGetThreshold(pixm, pixms, factor, ppixsd);
if (ppixd)
pixd = pixApplyLocalThreshold(pixsc, pixth, 1);
if (ppixm)
*ppixm = pixm;
else
pixDestroy(&pixm);
pixDestroy(&pixms);
if (ppixth)
*ppixth = pixth;
else
pixDestroy(&pixth);
if (ppixd)
*ppixd = pixd;
else
pixDestroy(&pixd);
pixDestroy(&pixg);
pixDestroy(&pixsc);
return 0;
}
/*!
* pixDisplayWriteFormat()
*
* Input: pix (1, 2, 4, 8, 16, 32 bpp)
* reduction (-1 to reset/erase; 0 to disable;
* otherwise this is a reduction factor)
* format (IFF_PNG or IFF_JFIF_JPEG)
* Return: 0 if OK; 1 on error
*
* Notes:
* (1) This writes files if reduction > 0. These can be displayed using
* pixDisplayMultiple("/tmp/junk_write_display*");
* (2) All previously written files can be erased by calling with
* reduction < 0; the value of pixs is ignored.
* (3) If reduction > 1 and depth == 1, this does a scale-to-gray
* reduction.
* (4) This function uses a static internal variable to number
* output files written by a single process. Behavior
* with a shared library may be unpredictable.
* (5) Output file format is as follows:
* format == IFF_JFIF_JPEG:
* png if d < 8 or d == 16 or if the output pix
* has a colormap. Otherwise, output is jpg.
* format == IFF_PNG:
* png (lossless) on all images.
* (6) For 16 bpp, the choice of full dynamic range with log scale
* is the best for displaying these images. Alternative outputs are
* pix8 = pixMaxDynamicRange(pixt, L_LINEAR_SCALE);
* pix8 = pixConvert16To8(pixt, 0); // low order byte
* pix8 = pixConvert16To8(pixt, 1); // high order byte
*/
l_int32
pixDisplayWriteFormat(PIX *pixs,
l_int32 reduction,
l_int32 format)
{
char buffer[L_BUF_SIZE];
l_int32 ignore;
l_float32 scale;
PIX *pixt, *pix8;
static l_int32 index = 0; /* caution: not .so or thread safe */
PROCNAME("pixDisplayWriteFormat");
if (reduction == 0) return 0;
if (reduction < 0) {
index = 0; /* reset; this will cause erasure at next call to write */
return 0;
}
if (format != IFF_JFIF_JPEG && format != IFF_PNG)
return ERROR_INT("invalid format", procName, 1);
if (!pixs)
return ERROR_INT("pixs not defined", procName, 1);
if (index == 0) {
snprintf(buffer, L_BUF_SIZE,
"rm -f /tmp/junk_write_display.*.png /tmp/junk_write_display.*.jpg");
ignore = system(buffer);
}
index++;
if (reduction == 1)
pixt = pixClone(pixs);
else {
scale = 1. / (l_float32)reduction;
if (pixGetDepth(pixs) == 1)
pixt = pixScaleToGray(pixs, scale);
else
pixt = pixScale(pixs, scale, scale);
}
if (pixGetDepth(pixt) == 16) {
pix8 = pixMaxDynamicRange(pixt, L_LOG_SCALE);
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_write_display.%03d.png", index);
pixWrite(buffer, pix8, IFF_PNG);
pixDestroy(&pix8);
}
else if (pixGetDepth(pixt) < 8 || pixGetColormap(pixt) ||
format == IFF_PNG) {
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_write_display.%03d.png", index);
pixWrite(buffer, pixt, IFF_PNG);
}
else {
snprintf(buffer, L_BUF_SIZE, "/tmp/junk_write_display.%03d.jpg", index);
pixWrite(buffer, pixt, format);
}
pixDestroy(&pixt);
return 0;
}
int main(int argc,
char **argv)
{
char *filein, *fileout, *fname;
char buffer[512];
const char *psfile = "/tmp/junk_split_image.ps";
l_int32 nx, ny, i, w, h, d, ws, hs, n, res, ignore;
l_float32 scale;
PIX *pixs, *pixt, *pixr;
PIXA *pixa;
static char mainName[] = "splitimage2pdf";
if (argc != 5)
return ERROR_INT(" Syntax: splitimage2pdf filein nx ny fileout",
mainName, 1);
filein = argv[1];
nx = atoi(argv[2]);
ny = atoi(argv[3]);
fileout = argv[4];
lept_rm(NULL, "junk_split_image.ps");
if ((pixs = pixRead(filein)) == NULL)
return ERROR_INT("pixs not made", mainName, 1);
d = pixGetDepth(pixs);
if (d == 1 )
lept_rm(NULL, "junk_split_image.tif");
else if (d == 8 || d == 32)
lept_rm(NULL, "junk_split_image.jpg");
else
return ERROR_INT("d not in {1,8,32} bpp", mainName, 1);
pixGetDimensions(pixs, &ws, &hs, NULL);
if (ny * ws > nx * hs)
pixr = pixRotate90(pixs, 1);
else
pixr = pixClone(pixs);
pixa = pixaSplitPix(pixr, nx, ny, 0, 0);
n = pixaGetCount(pixa);
res = 300;
for (i = 0; i < n; i++) {
pixt = pixaGetPix(pixa, i, L_CLONE);
pixGetDimensions(pixt, &w, &h, NULL);
scale = L_MIN(FILL_FACTOR * 2550 / w, FILL_FACTOR * 3300 / h);
fname = NULL;
if (d == 1) {
fname = genPathname("/tmp", "junk_split_image.tif");
pixWrite(fname, pixt, IFF_TIFF_G4);
if (i == 0) {
convertG4ToPS(fname, psfile, "w", 0, 0, 300,
scale, 1, FALSE, TRUE);
} else {
convertG4ToPS(fname, psfile, "a", 0, 0, 300,
scale, 1, FALSE, TRUE);
}
} else {
fname = genPathname("/tmp", "junk_split_image.jpg");
pixWrite(fname, pixt, IFF_JFIF_JPEG);
if (i == 0) {
convertJpegToPS(fname, psfile, "w", 0, 0, 300,
scale, 1, TRUE);
} else {
convertJpegToPS(fname, psfile, "a", 0, 0, 300,
scale, 1, TRUE);
}
}
lept_free(fname);
pixDestroy(&pixt);
}
snprintf(buffer, sizeof(buffer), "ps2pdf %s %s", psfile, fileout);
ignore = system(buffer); /* ps2pdf */
pixaDestroy(&pixa);
pixDestroy(&pixr);
pixDestroy(&pixs);
return 0;
}
/*!
* deskew()
*
* Input: pixs
* redsearch (for binary search: reduction factor = 1, 2 or 4)
* Return: deskewed pix, or NULL on error
*/
PIX *
deskew(PIX *pixs,
l_int32 redsearch)
{
l_float32 angle, conf, deg2rad;
PIX *pixg; /* gray version */
PIX *pixb; /* binary version */
PIX *pixd; /* destination image */
PROCNAME("deskew");
if (!pixs)
return (PIX *)ERROR_PTR("pixs not defined", procName, NULL);
/* Calculate a skew angle. We may need to make a binary version of the
* image for this calculation.
*/
if (pixGetDepth(pixs) != 1) {
/* FIX ME: We should probably pick a threshold value with more care. */
/* Create a grayscale image if we need one. */
if (pixGetDepth(pixs) >= 24) {
pixg = pixConvertRGBToGray(pixs, 0.0, 0.0, 0.0);
} else {
pixg = pixs;
}
pixb = pixThresholdToBinary(pixg, 127);
if (pixg != pixs) {
pixDestroy(&pixg);
}
/* Assert: We are done with any gray image. */
} else {
pixb = pixs;
}
/* Assert: We have a valid binary image. */
if (redsearch != 1 && redsearch != 2 && redsearch != 4)
return (PIX *)ERROR_PTR("redsearch not in {1,2,4}", procName, NULL);
deg2rad = 3.1415926535 / 180.;
if (pixFindSkewSweepAndSearch(pixb, &angle, &conf,
DEFAULT_SWEEP_REDUCTION, redsearch,
DEFAULT_SWEEP_RANGE, DEFAULT_SWEEP_DELTA,
DEFAULT_MINBS_DELTA)) {
pixd = pixClone(pixs);
goto finish;
}
if (L_ABS(angle) < MIN_DESKEW_ANGLE || conf < MIN_ALLOWED_CONFIDENCE) {
pixd = pixClone(pixs);
goto finish;
}
/* If the pixel depth of pixs is 1, we need to use a bit-depth
* independent rotate instead of the more accurate area mapping rotate.
*/
if (pixGetDepth(pixs) == 1) {
if ((pixd = pixRotateShear(pixs, 0, 0, deg2rad * angle, 0xffffff00)) == NULL) {
pixd = pixClone(pixs);
}
} else {
#if defined(COLOR_ROTATE)
if ((pixd = pixRotateAMColorFast(pixs, deg2rad * angle)) == NULL) {
pixd = pixClone(pixs);
}
#else
if ((pixd = pixRotateAM(pixs, deg2rad * angle, 0xffffff00)) == NULL) {
pixd = pixClone(pixs);
}
#endif
}
finish:
if (pixb != pixs) {
pixDestroy(&pixb);
}
return pixd;
}
/*!
* pixStrokeWidthTransform()
*
* Input: pixs (1 bpp)
* color (0 for white runs, 1 for black runs)
* depth (of pixd: 8 or 16 bpp)
* nangles (2, 4, 6 or 8)
* Return: pixd (8 or 16 bpp), or null on error
*
* Notes:
* (1) The dest Pix is 8 or 16 bpp, with the pixel values
* equal to the stroke width in which it is a member.
* The values are clipped to the max pixel value if necessary.
* (2) The color determines if we're labelling white or black strokes.
* (3) A pixel that is not a member of the chosen color gets
* value 0; it belongs to a width of length 0 of the
* chosen color.
* (4) This chooses, for each dest pixel, the minimum of sets
* of runlengths through each pixel. Here are the sets:
* nangles increment set
* ------- --------- --------------------------------
* 2 90 {0, 90}
* 4 45 {0, 45, 90, 135}
* 6 30 {0, 30, 60, 90, 120, 150}
* 8 22.5 {0, 22.5, 45, 67.5, 90, 112.5, 135, 157.5}
* (5) Runtime scales linearly with (nangles - 2).
*/
PIX *
pixStrokeWidthTransform(PIX *pixs,
l_int32 color,
l_int32 depth,
l_int32 nangles) {
l_float32 angle, pi;
PIX *pixh, *pixv, *pixt, *pixg1, *pixg2, *pixg3, *pixg4;
PROCNAME("pixStrokeWidthTransform");
if (!pixs || pixGetDepth(pixs) != 1)
return (PIX *) ERROR_PTR("pixs undefined or not 1 bpp", procName, NULL);
if (depth != 8 && depth != 16)
return (PIX *) ERROR_PTR("depth must be 8 or 16 bpp", procName, NULL);
if (nangles != 2 && nangles != 4 && nangles != 6 && nangles != 8)
return (PIX *) ERROR_PTR("nangles not in {2,4,6,8}", procName, NULL);
/* Use fg runs for evaluation */
if (color == 0)
pixt = pixInvert(NULL, pixs);
else
pixt = pixClone(pixs);
/* Find min length at 0 and 90 degrees */
pixh = pixRunlengthTransform(pixt, 1, L_HORIZONTAL_RUNS, depth);
pixv = pixRunlengthTransform(pixt, 1, L_VERTICAL_RUNS, depth);
pixg1 = pixMinOrMax(NULL, pixh, pixv, L_CHOOSE_MIN);
pixDestroy(&pixh);
pixDestroy(&pixv);
pixg2 = pixg3 = pixg4 = NULL;
pi = 3.1415926535;
if (nangles == 4 || nangles == 8) {
/* Find min length at +45 and -45 degrees */
angle = pi / 4.0;
pixg2 = pixFindMinRunsOrthogonal(pixt, angle, depth);
}
if (nangles == 6) {
/* Find min length at +30 and -60 degrees */
angle = pi / 6.0;
pixg2 = pixFindMinRunsOrthogonal(pixt, angle, depth);
/* Find min length at +60 and -30 degrees */
angle = pi / 3.0;
pixg3 = pixFindMinRunsOrthogonal(pixt, angle, depth);
}
if (nangles == 8) {
/* Find min length at +22.5 and -67.5 degrees */
angle = pi / 8.0;
pixg3 = pixFindMinRunsOrthogonal(pixt, angle, depth);
/* Find min length at +67.5 and -22.5 degrees */
angle = 3.0 * pi / 8.0;
pixg4 = pixFindMinRunsOrthogonal(pixt, angle, depth);
}
pixDestroy(&pixt);
if (nangles > 2)
pixMinOrMax(pixg1, pixg1, pixg2, L_CHOOSE_MIN);
if (nangles > 4)
pixMinOrMax(pixg1, pixg1, pixg3, L_CHOOSE_MIN);
if (nangles > 6)
pixMinOrMax(pixg1, pixg1, pixg4, L_CHOOSE_MIN);
pixDestroy(&pixg2);
pixDestroy(&pixg3);
pixDestroy(&pixg4);
return pixg1;
}
/*!
* pixQuadtreeVariance()
*
* Input: pixs (8 bpp, no colormap)
* nlevels (in quadtree)
* *pix_ma (input mean accumulator; can be null)
* *dpix_msa (input mean square accumulator; can be null)
* *pfpixa_v (<optional return> variance values in quadtree)
* *pfpixa_rv (<optional return> root variance values in quadtree)
* Return: 0 if OK, 1 on error
*
* Notes:
* (1) The returned fpixav and fpixarv have @nlevels of fpix,
* each containing at the respective levels the variance
* and root variance values.
*/
l_int32
pixQuadtreeVariance(PIX *pixs,
l_int32 nlevels,
PIX *pix_ma,
DPIX *dpix_msa,
FPIXA **pfpixa_v,
FPIXA **pfpixa_rv)
{
l_int32 i, j, w, h, size, n;
l_float32 var, rvar;
BOX *box;
BOXA *boxa;
BOXAA *baa;
FPIX *fpixv, *fpixrv;
PIX *pix_mac; /* copy of mean accumulator */
DPIX *dpix_msac; /* msa clone */
PROCNAME("pixQuadtreeVariance");
if (!pfpixa_v && !pfpixa_rv)
return ERROR_INT("neither &fpixav nor &fpixarv defined", procName, 1);
if (pfpixa_v) *pfpixa_v = NULL;
if (pfpixa_rv) *pfpixa_rv = NULL;
if (!pixs || pixGetDepth(pixs) != 8)
return ERROR_INT("pixs not defined or not 8 bpp", procName, 1);
pixGetDimensions(pixs, &w, &h, NULL);
if (nlevels > quadtreeMaxLevels(w, h))
return ERROR_INT("nlevels too large for image", procName, 1);
if (!pix_ma)
pix_mac = pixBlockconvAccum(pixs);
else
pix_mac = pixClone(pix_ma);
if (!pix_mac)
return ERROR_INT("pix_mac not made", procName, 1);
if (!dpix_msa)
dpix_msac = pixMeanSquareAccum(pixs);
else
dpix_msac = dpixClone(dpix_msa);
if (!dpix_msac)
return ERROR_INT("dpix_msac not made", procName, 1);
if ((baa = boxaaQuadtreeRegions(w, h, nlevels)) == NULL) {
pixDestroy(&pix_mac);
dpixDestroy(&dpix_msac);
return ERROR_INT("baa not made", procName, 1);
}
if (pfpixa_v) *pfpixa_v = fpixaCreate(nlevels);
if (pfpixa_rv) *pfpixa_rv = fpixaCreate(nlevels);
for (i = 0; i < nlevels; i++) {
boxa = boxaaGetBoxa(baa, i, L_CLONE);
size = 1 << i;
n = boxaGetCount(boxa); /* n == size * size */
if (pfpixa_v) fpixv = fpixCreate(size, size);
if (pfpixa_rv) fpixrv = fpixCreate(size, size);
for (j = 0; j < n; j++) {
box = boxaGetBox(boxa, j, L_CLONE);
pixVarianceInRectangle(pixs, box, pix_mac, dpix_msac, &var, &rvar);
if (pfpixa_v) fpixSetPixel(fpixv, j % size, j / size, var);
if (pfpixa_rv) fpixSetPixel(fpixrv, j % size, j / size, rvar);
boxDestroy(&box);
}
if (pfpixa_v) fpixaAddFPix(*pfpixa_v, fpixv, L_INSERT);
if (pfpixa_rv) fpixaAddFPix(*pfpixa_rv, fpixrv, L_INSERT);
boxaDestroy(&boxa);
}
pixDestroy(&pix_mac);
dpixDestroy(&dpix_msac);
boxaaDestroy(&baa);
return 0;
}