Exemple #1
0
/*!
 *  boxaIntersectsBox()
 *
 *      Input:  boxas
 *              box (for intersecting)
 *      Return  boxad (boxa with all boxes in boxas that intersect box),
 *                     or null on error
 *
 *  Notes:
 *      (1) All boxes in boxa that intersect with box (i.e., are completely
 *          or partially contained in box) are retained.
 */
BOXA *
boxaIntersectsBox(BOXA  *boxas,
                  BOX   *box)
{
l_int32  i, n, val;
BOX     *boxt;
BOXA    *boxad;

    PROCNAME("boxaIntersectsBox");

    if (!boxas)
        return (BOXA *)ERROR_PTR("boxas not defined", procName, NULL);
    if (!box)
        return (BOXA *)ERROR_PTR("box not defined", procName, NULL);
    if ((n = boxaGetCount(boxas)) == 0)
        return boxaCreate(1);  /* empty */

    boxad = boxaCreate(0);
    for (i = 0; i < n; i++) {
        boxt = boxaGetBox(boxas, i, L_CLONE);
        boxIntersects(box, boxt, &val);
        if (val == 1)
            boxaAddBox(boxad, boxt, L_COPY);
        boxDestroy(&boxt);  /* destroy the clone */
    }

    return boxad;
}
Exemple #2
0
/*!
 *  boxaClipToBox()
 *
 *      Input:  boxas
 *              box (for clipping)
 *      Return  boxad (boxa with boxes in boxas clipped to box),
 *                     or null on error
 *
 *  Notes:
 *      (1) All boxes in boxa not intersecting with box are removed, and
 *          the remaining boxes are clipped to box.
 */
BOXA *
boxaClipToBox(BOXA  *boxas,
              BOX   *box)
{
l_int32  i, n;
BOX     *boxt, *boxo;
BOXA    *boxad;

    PROCNAME("boxaClipToBox");

    if (!boxas)
        return (BOXA *)ERROR_PTR("boxas not defined", procName, NULL);
    if (!box)
        return (BOXA *)ERROR_PTR("box not defined", procName, NULL);
    if ((n = boxaGetCount(boxas)) == 0)
        return boxaCreate(1);  /* empty */

    boxad = boxaCreate(0);
    for (i = 0; i < n; i++) {
        boxt = boxaGetBox(boxas, i, L_CLONE);
        if ((boxo = boxOverlapRegion(box, boxt)) != NULL)
            boxaAddBox(boxad, boxo, L_INSERT);
        boxDestroy(&boxt);
    }

    return boxad;
}
Exemple #3
0
/*!
 *  boxaSplitEvenOdd()
 *
 *      Input:  boxa
 *              &boxae, &boxao (<return> save even and odd boxes in their
 *                 separate boxa, setting the other type to invalid boxes.)
 *      Return: 0 if OK, 1 on error
 *
 *  Notes:
 *      (1) For example, boxae copies of the even boxes, in their original
 *          location, that are in boxa.  Invalid boxes are placed
 *          in the odd array locations.
 *
 */
l_int32
boxaSplitEvenOdd(BOXA   *boxa,
                 BOXA  **pboxae,
                 BOXA  **pboxao)
{
l_int32  i, n;
BOX     *box, *boxt;

    PROCNAME("boxaSplitEvenOdd");

    if (!pboxae || !pboxao)
        return ERROR_INT("&boxae and &boxao not defined", procName, 1);
    *pboxae = *pboxao = NULL;
    if (!boxa)
        return ERROR_INT("boxa not defined", procName, 1);

    n = boxaGetCount(boxa);
    *pboxae = boxaCreate(n);
    *pboxao = boxaCreate(n);
    for (i = 0; i < n; i++) {
        box = boxaGetBox(boxa, i, L_COPY);
        boxt = boxCreate(0, 0, 0, 0);  /* empty placeholder */
        if ((i & 1) == 0) {
            boxaAddBox(*pboxae, box, L_INSERT);
            boxaAddBox(*pboxao, boxt, L_INSERT);
        }
        else {
            boxaAddBox(*pboxae, boxt, L_INSERT);
            boxaAddBox(*pboxao, box, L_INSERT);
        }
    }
    return 0;
}
/*!
 *  boxaReadStream()
 *
 *      Input:  stream
 *      Return: boxa, or null on error
 */
BOXA *
boxaReadStream(FILE  *fp)
{
l_int32  n, i, x, y, w, h, version;
l_int32  ignore;
BOX     *box;
BOXA    *boxa;

    PROCNAME("boxaReadStream");

    if (!fp)
        return (BOXA *)ERROR_PTR("stream not defined", procName, NULL);

    if (fscanf(fp, "\nBoxa Version %d\n", &version) != 1)
        return (BOXA *)ERROR_PTR("not a boxa file", procName, NULL);
    if (version != BOXA_VERSION_NUMBER)
        return (BOXA *)ERROR_PTR("invalid boxa version", procName, NULL);
    if (fscanf(fp, "Number of boxes = %d\n", &n) != 1)
        return (BOXA *)ERROR_PTR("not a boxa file", procName, NULL);

    if ((boxa = boxaCreate(n)) == NULL)
        return (BOXA *)ERROR_PTR("boxa not made", procName, NULL);

    for (i = 0; i < n; i++) {
        if (fscanf(fp, "  Box[%d]: x = %d, y = %d, w = %d, h = %d\n",
                &ignore, &x, &y, &w, &h) != 5)
            return (BOXA *)ERROR_PTR("box descr not valid", procName, NULL);
        if ((box = boxCreate(x, y, w, h)) == NULL)
            return (BOXA *)ERROR_PTR("box not made", procName, NULL);
        boxaAddBox(boxa, box, L_INSERT);
    }

    return boxa;
}
/*!
 *  boxaCopy()
 *
 *      Input:  boxa
 *              copyflag (L_COPY, L_CLONE, L_COPY_CLONE)
 *      Return: new boxa, or null on error
 *
 *  Notes:
 *      (1) See pix.h for description of the copyflag.
 *      (2) The copy-clone makes a new boxa that holds clones of each box.
 */
BOXA *
boxaCopy(BOXA    *boxa,
         l_int32  copyflag)
{
l_int32  i;
BOX     *boxc;
BOXA    *boxac;

    PROCNAME("boxaCopy");

    if (!boxa)
        return (BOXA *)ERROR_PTR("boxa not defined", procName, NULL);

    if (copyflag == L_CLONE) {
        boxa->refcount++;
        return boxa;
    }

    if (copyflag != L_COPY && copyflag != L_COPY_CLONE)
        return (BOXA *)ERROR_PTR("invalid copyflag", procName, NULL);

    if ((boxac = boxaCreate(boxa->nalloc)) == NULL)
        return (BOXA *)ERROR_PTR("boxac not made", procName, NULL);
    for (i = 0; i < boxa->n; i++) {
        if (copyflag == L_COPY)
            boxc = boxaGetBox(boxa, i, L_COPY);
        else   /* copy-clone */
            boxc = boxaGetBox(boxa, i, L_CLONE);
        boxaAddBox(boxac, boxc, L_INSERT);
    }
    return boxac;
}
Exemple #6
0
/*!
 *  boxaMergeEvenOdd()
 *
 *      Input:  boxae (boxes to go in even positions in merged boxa)
 *              boxao (boxes to go in odd positions in merged boxa)
 *      Return: boxad (merged), or null on error
 *
 *  Notes:
 *      (1) Boxes are alternatingly selected from boxae and boxao.
 *          Both boxae and boxao are of the same size.
 */
BOXA *
boxaMergeEvenOdd(BOXA  *boxae,
                 BOXA  *boxao)
{
l_int32  i, n;
BOX     *box;
BOXA    *boxad;

    PROCNAME("boxaMergeEvenOdd");

    if (!boxae || !boxao)
        return (BOXA *)ERROR_PTR("boxae and boxao not defined", procName, NULL);
    n = boxaGetCount(boxae);
    if (n != boxaGetCount(boxao))
        return (BOXA *)ERROR_PTR("boxa sizes differ", procName, NULL);

    boxad = boxaCreate(n);
    for (i = 0; i < n; i++) {
        if ((i & 1) == 0)
            box = boxaGetBox(boxae, i, L_COPY);
        else
            box = boxaGetBox(boxao, i, L_COPY);
        boxaAddBox(boxad, box, L_INSERT);
    }
    return boxad;
}
Exemple #7
0
/*!
 *  boxaTransform()
 *
 *      Input:  boxa
 *              shiftx, shifty
 *              scalex, scaley
 *      Return: boxad, or null on error
 *
 *  Notes:
 *      (1) This is a very simple function that first shifts, then scales.
 */
BOXA *
boxaTransform(BOXA      *boxas,
              l_int32    shiftx,
              l_int32    shifty,
              l_float32  scalex,
              l_float32  scaley)
{
l_int32  i, n;
BOX     *boxs, *boxd;
BOXA    *boxad;

    PROCNAME("boxaTransform");

    if (!boxas)
        return (BOXA *)ERROR_PTR("boxas not defined", procName, NULL);
    n = boxaGetCount(boxas);
    if ((boxad = boxaCreate(n)) == NULL)
        return (BOXA *)ERROR_PTR("boxad not made", procName, NULL);
    for (i = 0; i < n; i++) {
        if ((boxs = boxaGetBox(boxas, i, L_CLONE)) == NULL)
            return (BOXA *)ERROR_PTR("boxs not found", procName, NULL);
        boxd = boxTransform(boxs, shiftx, shifty, scalex, scaley);
        boxDestroy(&boxs);
        boxaAddBox(boxad, boxd, L_INSERT);
    }

    return boxad;
}
Exemple #8
0
/*!
 *  boxaSortByIndex()
 *
 *      Input:  boxas
 *              naindex (na that maps from the new boxa to the input boxa)
 *      Return: boxad (sorted), or null on error
 */
BOXA *
boxaSortByIndex(BOXA  *boxas,
                NUMA  *naindex)
{
l_int32  i, n, index;
BOX     *box;
BOXA    *boxad;

    PROCNAME("boxaSortByIndex");

    if (!boxas)
        return (BOXA *)ERROR_PTR("boxas not defined", procName, NULL);
    if (!naindex)
        return (BOXA *)ERROR_PTR("naindex not defined", procName, NULL);

    n = boxaGetCount(boxas);
    boxad = boxaCreate(n);
    for (i = 0; i < n; i++) {
        numaGetIValue(naindex, i, &index);
        box = boxaGetBox(boxas, index, L_COPY);
        boxaAddBox(boxad, box, L_INSERT);
    }

    return boxad;
}
Exemple #9
0
/*!
 *  boxaRotateOrth()
 *
 *      Input:  boxa
 *              w, h (of image in which the boxa is embedded)
 *              rotation (0 = noop, 1 = 90 deg, 2 = 180 deg, 3 = 270 deg;
 *                        all rotations are clockwise)
 *      Return: boxad, or null on error
 *
 *  Notes:
 *      (1) See boxRotateOrth() for details.
 */
BOXA *
boxaRotateOrth(BOXA    *boxas,
               l_int32  w,
               l_int32  h,
               l_int32  rotation)
{
l_int32  i, n;
BOX     *boxs, *boxd;
BOXA    *boxad;

    PROCNAME("boxaRotateOrth");

    if (!boxas)
        return (BOXA *)ERROR_PTR("boxas not defined", procName, NULL);
    if (rotation == 0)
        return boxaCopy(boxas, L_COPY);
    if (rotation < 1 || rotation > 3)
        return (BOXA *)ERROR_PTR("rotation not in {0,1,2,3}", procName, NULL);

    n = boxaGetCount(boxas);
    if ((boxad = boxaCreate(n)) == NULL)
        return (BOXA *)ERROR_PTR("boxad not made", procName, NULL);
    for (i = 0; i < n; i++) {
        if ((boxs = boxaGetBox(boxas, i, L_CLONE)) == NULL)
            return (BOXA *)ERROR_PTR("boxs not found", procName, NULL);
        boxd = boxRotateOrth(boxs, w, h, rotation);
        boxDestroy(&boxs);
        boxaAddBox(boxad, boxd, L_INSERT);
    }

    return boxad;
}
Exemple #10
0
/*!
 *  boxaEncapsulateAligned()
 *
 *      Input:  boxa
 *              num (number put into each boxa in the baa)
 *              copyflag  (L_COPY or L_CLONE)
 *      Return: boxaa, or null on error
 *
 *  Notes:
 *      (1) This puts @num boxes from the input @boxa into each of a
 *          set of boxa within an output boxaa.
 *      (2) This assumes that the boxes in @boxa are in sets of @num each.
 */
BOXAA *
boxaEncapsulateAligned(BOXA    *boxa,
                       l_int32  num,
                       l_int32  copyflag)
{
l_int32  i, j, n, nbaa, index;
BOX     *box;
BOXA    *boxat;
BOXAA   *baa;

    PROCNAME("boxaEncapsulateAligned");

    if (!boxa)
        return (BOXAA *)ERROR_PTR("boxa not defined", procName, NULL);
    if (copyflag != L_COPY && copyflag != L_CLONE)
        return (BOXAA *)ERROR_PTR("invalid copyflag", procName, NULL);

    n = boxaGetCount(boxa);
    nbaa = (n + num - 1) / num;
    if (n / num != nbaa)
        L_ERROR("inconsistent alignment: n / num not an integer", procName);
    baa = boxaaCreate(nbaa);
    for (i = 0, index = 0; i < nbaa; i++) {
        boxat = boxaCreate(num);
        for (j = 0; j < num; j++, index++) {
            box = boxaGetBox(boxa, index, copyflag);
            boxaAddBox(boxat, box, L_INSERT);
        }
        boxaaAddBoxa(baa, boxat, L_INSERT);
    }

    return baa;
}
Exemple #11
0
/*!
 * \brief   boxaaQuadtreeRegions()
 *
 * \param[in]    w, h     size of pix that is being quadtree-ized
 * \param[in]    nlevels  number of levels in quadtree
 * \return  baa for quadtree regions at each level, or NULL on error
 *
 * <pre>
 * Notes:
 *      (1) The returned boxaa has %nlevels of boxa, each containing
 *          the set of rectangles at that level.  The rectangle at
 *          level 0 is the entire region; at level 1 the region is
 *          divided into 4 rectangles, and at level n there are n^4
 *          rectangles.
 *      (2) At each level, the rectangles in the boxa are in "raster"
 *          order, with LR (fast scan) and TB (slow scan).
 * </pre>
 */
BOXAA *
boxaaQuadtreeRegions(l_int32  w,
                     l_int32  h,
                     l_int32  nlevels)
{
l_int32   i, j, k, maxpts, nside, nbox, bw, bh;
l_int32  *xstart, *xend, *ystart, *yend;
BOX      *box;
BOXA     *boxa;
BOXAA    *baa;

    PROCNAME("boxaaQuadtreeRegions");

    if (nlevels < 1)
        return (BOXAA *)ERROR_PTR("nlevels must be >= 1", procName, NULL);
    if (w < (1 << (nlevels - 1)))
        return (BOXAA *)ERROR_PTR("w doesn't support nlevels", procName, NULL);
    if (h < (1 << (nlevels - 1)))
        return (BOXAA *)ERROR_PTR("h doesn't support nlevels", procName, NULL);

    baa = boxaaCreate(nlevels);
    maxpts = 1 << (nlevels - 1);
    xstart = (l_int32 *)LEPT_CALLOC(maxpts, sizeof(l_int32));
    xend = (l_int32 *)LEPT_CALLOC(maxpts, sizeof(l_int32));
    ystart = (l_int32 *)LEPT_CALLOC(maxpts, sizeof(l_int32));
    yend = (l_int32 *)LEPT_CALLOC(maxpts, sizeof(l_int32));
    for (k = 0; k < nlevels; k++) {
        nside = 1 << k;  /* number of boxes in each direction */
        for (i = 0; i < nside; i++) {
            xstart[i] = (w - 1) * i / nside;
            if (i > 0) xstart[i]++;
            xend[i] = (w - 1) * (i + 1) / nside;
            ystart[i] = (h - 1) * i / nside;
            if (i > 0) ystart[i]++;
            yend[i] = (h - 1) * (i + 1) / nside;
#if DEBUG_BOXES
            fprintf(stderr,
               "k = %d, xs[%d] = %d, xe[%d] = %d, ys[%d] = %d, ye[%d] = %d\n",
                    k, i, xstart[i], i, xend[i], i, ystart[i], i, yend[i]);
#endif  /* DEBUG_BOXES */
        }
        nbox = 1 << (2 * k);
        boxa = boxaCreate(nbox);
        for (i = 0; i < nside; i++) {
            bh = yend[i] - ystart[i] + 1;
            for (j = 0; j < nside; j++) {
                bw = xend[j] - xstart[j] + 1;
                box = boxCreate(xstart[j], ystart[i], bw, bh);
                boxaAddBox(boxa, box, L_INSERT);
            }
        }
        boxaaAddBoxa(baa, boxa, L_INSERT);
    }

    LEPT_FREE(xstart);
    LEPT_FREE(xend);
    LEPT_FREE(ystart);
    LEPT_FREE(yend);
    return baa;
}
Exemple #12
0
main(int    argc,
     char **argv)
{
char       *filein, *fileout;
l_int32     d;
BOX        *box1, *box2, *box3, *box4;
BOXA       *boxa;
PIX        *pixs, *pixt1, *pixt2, *pixt3;
PTA        *pta;
static char     mainName[] = "graphicstest";

    if (argc != 3)
        exit(ERROR_INT(" Syntax: graphicstest filein fileout", mainName, 1));

    filein = argv[1];
    fileout = argv[2];
    if ((pixs = pixRead(filein)) == NULL)
        exit(ERROR_INT(" Syntax: pixs not made", mainName, 1));
    d = pixGetDepth(pixs);
    if (d <= 8)
        pixt1 = pixConvertTo32(pixs);
    else
        pixt1 = pixClone(pixs);

        /* Paint on RGB */
    pixRenderLineArb(pixt1, 450, 20, 850, 320, 5, 200, 50, 125);
    pixRenderLineArb(pixt1, 30, 40, 440, 40, 5, 100, 200, 25);
    pixRenderLineBlend(pixt1, 30, 60, 440, 70, 5, 115, 200, 120, 0.3);
    pixRenderLineBlend(pixt1, 30, 600, 440, 670, 9, 215, 115, 30, 0.5);
    pixRenderLineBlend(pixt1, 130, 700, 540, 770, 9, 255, 255, 250, 0.4);
    pixRenderLineBlend(pixt1, 130, 800, 540, 870, 9, 0, 0, 0, 0.4);
    box1 = boxCreate(70, 80, 300, 245);
    box2 = boxCreate(470, 180, 150, 205);
    box3 = boxCreate(520, 220, 160, 220);
    box4 = boxCreate(570, 260, 160, 220);
    boxa = boxaCreate(3);
    boxaAddBox(boxa, box2, L_INSERT);
    boxaAddBox(boxa, box3, L_INSERT);
    boxaAddBox(boxa, box4, L_INSERT);
    pixRenderBoxArb(pixt1, box1, 3, 200, 200, 25);
    pixRenderBoxaBlend(pixt1, boxa, 17, 200, 200, 25, 0.4, 1);
    pta = ptaCreate(5);
    ptaAddPt(pta, 250, 300);
    ptaAddPt(pta, 350, 450);
    ptaAddPt(pta, 400, 600);
    ptaAddPt(pta, 212, 512);
    ptaAddPt(pta, 180, 375);
    pixRenderPolylineBlend(pixt1, pta, 17, 25, 200, 200, 0.5, 1, 1);
    pixWrite(fileout, pixt1, IFF_JFIF_JPEG);
    pixDisplay(pixt1, 200, 200);

    pixDestroy(&pixs);
    pixDestroy(&pixt1);
    boxDestroy(&box1);
    boxaDestroy(&boxa);
    ptaDestroy(&pta);
    pixDestroy(&pixs);
    return 0;
}
Exemple #13
0
main(int    argc,
     char **argv)
{
    l_int32       i, k, x, y, w, h;
    BOX          *box;
    BOXA         *boxa1, *boxa2;
    PIX          *pix1, *pix2, *pixd;
    PIXA         *pixa;
    L_REGPARAMS  *rp;

    if (regTestSetup(argc, argv, &rp))
        return 1;

    for (k = 0; k < 7; k++) {
        srand(45617);
        pixa = pixaCreate(2);
        boxa1 = boxaCreate(0);
        for (i = 0; i < 500; i++) {
            x = (l_int32)(600.0 * (l_float64)rand() / (l_float64)RAND_MAX);
            y = (l_int32)(600.0 * (l_float64)rand() / (l_float64)RAND_MAX);
            w = (l_int32)
                (1.0 + maxsize[k] * (l_float64)rand() / (l_float64)RAND_MAX);
            h = (l_int32)
                (1.0 + maxsize[k] * (l_float64)rand() / (l_float64)RAND_MAX);
            box = boxCreate(x, y, w, h);
            boxaAddBox(boxa1, box, L_INSERT);
        }

        pix1 = pixCreate(660, 660, 1);
        pixRenderBoxa(pix1, boxa1, 1, L_SET_PIXELS);
        pixaAddPix(pixa, pix1, L_INSERT);
        boxa2 = boxaCombineOverlaps(boxa1);
        pix2 = pixCreate(660, 660, 1);
        pixRenderBoxa(pix2, boxa2, 1, L_SET_PIXELS);
        pixaAddPix(pixa, pix2, L_INSERT);

        pixd = pixaDisplayTiledInRows(pixa, 1, 1500, 1.0, 0, 50, 2);
        pixDisplayWithTitle(pixd, 100, 100 + 100 * k, NULL, rp->display);
        regTestWritePixAndCheck(rp, pixd, IFF_PNG);
        fprintf(stderr, "%d: n_init = %d, n_final = %d\n",
                k, boxaGetCount(boxa1), boxaGetCount(boxa2));
        pixDestroy(&pixd);
        boxaDestroy(&boxa1);
        boxaDestroy(&boxa2);
        pixaDestroy(&pixa);
    }

    regTestCleanup(rp);
    return 0;
}
Exemple #14
0
/*!
 *  boxaaFlattenToBoxa()
 *
 *      Input:  boxaa
 *              &naindex  (<optional return> the boxa index in the boxaa)
 *              copyflag  (L_COPY or L_CLONE)
 *      Return: boxa, or null on error
 *
 *  Notes:
 *      (1) This 'flattens' the boxaa to a boxa, taking the boxes in
 *          order in the first boxa, then the second, etc.
 *      (2) If a boxa is empty, we generate an invalid, placeholder box
 *          of zero size.  This is useful when converting from a boxaa
 *          where each boxa has either 0 or 1 boxes, and it is necessary
 *          to maintain a 1:1 correspondence between the initial
 *          boxa array and the resulting box array.
 *      (3) If &naindex is defined, we generate a Numa that gives, for
 *          each box in the boxaa, the index of the boxa to which it belongs.
 */
BOXA *
boxaaFlattenToBoxa(BOXAA   *baa,
                   NUMA   **pnaindex,
                   l_int32  copyflag)
{
l_int32  i, j, m, n;
BOXA    *boxa, *boxat;
BOX     *box;
NUMA    *naindex;

    PROCNAME("boxaaFlattenToBoxa");

    if (pnaindex) *pnaindex = NULL;
    if (!baa)
        return (BOXA *)ERROR_PTR("baa not defined", procName, NULL);
    if (copyflag != L_COPY && copyflag != L_CLONE)
        return (BOXA *)ERROR_PTR("invalid copyflag", procName, NULL);
    if (pnaindex) {
        naindex = numaCreate(0);
        *pnaindex = naindex;
    }

    n = boxaaGetCount(baa);
    boxa = boxaCreate(n);
    for (i = 0; i < n; i++) {
        boxat = boxaaGetBoxa(baa, i, L_CLONE);
        m = boxaGetCount(boxat);
        if (m == 0) {  /* placeholder box */
            box = boxCreate(0, 0, 0, 0);
            boxaAddBox(boxa, box, L_INSERT);
            if (pnaindex)
                numaAddNumber(naindex, i);  /* save 'row' number */
        }
        else {
            for (j = 0; j < m; j++) {
                box = boxaGetBox(boxat, j, copyflag);
                boxaAddBox(boxa, box, L_INSERT);
                if (pnaindex)
                    numaAddNumber(naindex, i);  /* save 'row' number */
            }
        }
        boxaDestroy(&boxat);
    }

    return boxa;
}
/* static */
void BoxChar::RotateBoxes(float rotation, int xcenter, int ycenter,
                          int start_box, int end_box, vector<BoxChar*>* boxes) {
  Boxa* orig = boxaCreate(0);
  for (int i = start_box; i < end_box; ++i) {
    BOX* box = (*boxes)[i]->box_;
    if (box) boxaAddBox(orig, box, L_CLONE);
  }
  Boxa* rotated = boxaRotate(orig, xcenter, ycenter, rotation);
  boxaDestroy(&orig);
  for (int i = start_box, box_ind = 0; i < end_box; ++i) {
    if ((*boxes)[i]->box_) {
      boxDestroy(&((*boxes)[i]->box_));
      (*boxes)[i]->box_ = boxaGetBox(rotated, box_ind++, L_CLONE);
    }
  }
  boxaDestroy(&rotated);
}
/*!
 *  pixColorGrayCmap()
 *
 *      Input:  pixs (2, 4 or 8 bpp, with colormap)
 *              box (<optional> region to set color; can be NULL)
 *              type (L_PAINT_LIGHT, L_PAINT_DARK)
 *              rval, gval, bval (target color)
 *      Return: 0 if OK, 1 on error
 *
 *  Notes:
 *      (1) This is an in-place operation.
 *      (2) If type == L_PAINT_LIGHT, it colorizes non-black pixels,
 *          preserving antialiasing.
 *          If type == L_PAINT_DARK, it colorizes non-white pixels,
 *          preserving antialiasing.
 *      (3) box gives the region to apply color; if NULL, this
 *          colorizes the entire image.
 *      (4) If the cmap is only 2 or 4 bpp, pixs is converted in-place
 *          to an 8 bpp cmap.  A 1 bpp cmap is not a valid input pix.
 *      (5) This can also be called through pixColorGray().
 *      (6) This operation increases the colormap size by the number of
 *          different gray (non-black or non-white) colors in the
 *          input colormap.  If there is not enough room in the colormap
 *          for this expansion, it returns 1 (error), and the caller
 *          should check the return value.
 *      (7) Using the darkness of each original pixel in the rect,
 *          it generates a new color (based on the input rgb values).
 *          If type == L_PAINT_LIGHT, the new color is a (generally)
 *          darken-to-black version of the  input rgb color, where the
 *          amount of darkening increases with the darkness of the
 *          original pixel color.
 *          If type == L_PAINT_DARK, the new color is a (generally)
 *          faded-to-white version of the  input rgb color, where the
 *          amount of fading increases with the brightness of the
 *          original pixel color.
 */
l_int32
pixColorGrayCmap(PIX     *pixs,
                 BOX     *box,
                 l_int32  type,
                 l_int32  rval,
                 l_int32  gval,
                 l_int32  bval)
{
l_int32   w, h, d, ret;
PIX      *pixt;
BOXA     *boxa;
PIXCMAP  *cmap;

    PROCNAME("pixColorGrayCmap");

    if (!pixs)
        return ERROR_INT("pixs not defined", procName, 1);
    if ((cmap = pixGetColormap(pixs)) == NULL)
        return ERROR_INT("no colormap", procName, 1);
    pixGetDimensions(pixs, &w, &h, &d);
    if (d != 2 && d != 4 && d != 8)
        return ERROR_INT("depth not in {2, 4, 8}", procName, 1);
    if (type != L_PAINT_DARK && type != L_PAINT_LIGHT)
        return ERROR_INT("invalid type", procName, 1);

        /* If 2 bpp or 4 bpp, convert in-place to 8 bpp. */
    if (d == 2 || d == 4) {
        pixt = pixConvertTo8(pixs, 1);
        pixTransferAllData(pixs, &pixt, 0, 0);
    }

        /* If box == NULL, color the entire image */
    boxa = boxaCreate(1);
    if (box) {
        boxaAddBox(boxa, box, L_COPY);
    } else {
        box = boxCreate(0, 0, w, h);
        boxaAddBox(boxa, box, L_INSERT);
    }
    ret = pixColorGrayRegionsCmap(pixs, boxa, type, rval, gval, bval);

    boxaDestroy(&boxa);
    return ret;
}
/*!
 *  pixaaCreate()
 *
 *      Input:  n  (initial number of pixa ptrs)
 *      Return: pixaa, or null on error
 *
 *  Notes:
 *      (1) A pixaa provides a 2-level hierarchy of images.
 *          A common use is for segmentation masks, which are
 *          inexpensive to store in png format.
 *      (2) For example, suppose you want a mask for each textline
 *          in a two-column page.  The textline masks for each column
 *          can be represented by a pixa, of which there are 2 in the pixaa.
 *          The boxes for the textline mask components within a column
 *          can have their origin referred to the column rather than the page.
 *          Then the boxa field can be used to represent the two box (regions)
 *          for the columns, and the (x,y) components of each box can
 *          be used to get the absolute position of the textlines on
 *          the page.
 */
PIXAA *
pixaaCreate(l_int32  n)
{
PIXAA  *pixaa;

    PROCNAME("pixaaCreate");

    if (n <= 0)
        n = INITIAL_PTR_ARRAYSIZE;

    if ((pixaa = (PIXAA *)CALLOC(1, sizeof(PIXAA))) == NULL)
        return (PIXAA *)ERROR_PTR("pixaa not made", procName, NULL);
    pixaa->n = 0;
    pixaa->nalloc = n;
    
    if ((pixaa->pixa = (PIXA **)CALLOC(n, sizeof(PIXA *))) == NULL)
        return (PIXAA *)ERROR_PTR("pixa ptrs not made", procName, NULL);
    pixaa->boxa = boxaCreate(n);

    return pixaa;
}
Exemple #18
0
/*!
 * \brief   ptaConvertToBoxa()
 *
 * \param[in]    pta
 * \param[in]    ncorners   2 or 4 for the representation of each box
 * \return  boxa with one box for each 2 or 4 points in the pta,
 *                    or NULL on error
 *
 * <pre>
 * Notes:
 *      (1) For 2 corners, the order of the 2 points is UL, LR.
 *          For 4 corners, the order of points is UL, UR, LL, LR.
 *      (2) Each derived box is the minimum size containing all corners.
 * </pre>
 */
BOXA *
ptaConvertToBoxa(PTA     *pta,
                 l_int32  ncorners)
{
l_int32  i, n, nbox, x1, y1, x2, y2, x3, y3, x4, y4, x, y, xmax, ymax;
BOX     *box;
BOXA    *boxa;

    PROCNAME("ptaConvertToBoxa");

    if (!pta)
        return (BOXA *)ERROR_PTR("pta not defined", procName, NULL);
    if (ncorners != 2 && ncorners != 4)
        return (BOXA *)ERROR_PTR("ncorners not 2 or 4", procName, NULL);
    n = ptaGetCount(pta);
    if (n % ncorners != 0)
        return (BOXA *)ERROR_PTR("size % ncorners != 0", procName, NULL);
    nbox = n / ncorners;
    if ((boxa = boxaCreate(nbox)) == NULL)
        return (BOXA *)ERROR_PTR("boxa not made", procName, NULL);
    for (i = 0; i < n; i += ncorners) {
        ptaGetIPt(pta, i, &x1, &y1);
        ptaGetIPt(pta, i + 1, &x2, &y2);
        if (ncorners == 2) {
            box = boxCreate(x1, y1, x2 - x1 + 1, y2 - y1 + 1);
            boxaAddBox(boxa, box, L_INSERT);
            continue;
        }
        ptaGetIPt(pta, i + 2, &x3, &y3);
        ptaGetIPt(pta, i + 3, &x4, &y4);
        x = L_MIN(x1, x3);
        y = L_MIN(y1, y2);
        xmax = L_MAX(x2, x4);
        ymax = L_MAX(y3, y4);
        box = boxCreate(x, y, xmax - x + 1, ymax - y + 1);
        boxaAddBox(boxa, box, L_INSERT);
    }

    return boxa;
}
Exemple #19
0
/*!
 *  boxaSort2dByIndex()
 *
 *      Input:  boxas
 *              naa (numaa that maps from the new baa to the input boxa)
 *      Return: baa (sorted boxaa), or null on error
 */
BOXAA *
boxaSort2dByIndex(BOXA   *boxas,
                  NUMAA  *naa)
{
l_int32  ntot, boxtot, i, j, n, nn, index;
BOX     *box;
BOXA    *boxa;
BOXAA   *baa;
NUMA    *na;

    PROCNAME("boxaSort2dByIndex");

    if (!boxas)
        return (BOXAA *)ERROR_PTR("boxas not defined", procName, NULL);
    if (!naa)
        return (BOXAA *)ERROR_PTR("naindex not defined", procName, NULL);

        /* Check counts */
    ntot = numaaGetNumberCount(naa);
    boxtot = boxaGetCount(boxas);
    if (ntot != boxtot)
        return (BOXAA *)ERROR_PTR("element count mismatch", procName, NULL);

    n = numaaGetCount(naa);
    baa = boxaaCreate(n);
    for (i = 0; i < n; i++) {
        na = numaaGetNuma(naa, i, L_CLONE);
        nn = numaGetCount(na);
        boxa = boxaCreate(nn);
        for (j = 0; j < nn; j++) {
            numaGetIValue(na, i, &index);
            box = boxaGetBox(boxas, index, L_COPY);
            boxaAddBox(boxa, box, L_INSERT);
        }
        boxaaAddBoxa(baa, boxa, L_INSERT);
        numaDestroy(&na);
    }

    return baa;
}
Exemple #20
0
/*!
 * \brief   boxaSelectRange()
 *
 * \param[in]    boxas
 * \param[in]    first      use 0 to select from the beginning
 * \param[in]    last       use -1 to select to the end
 * \param[in]    copyflag   L_COPY, L_CLONE
 * \return  boxad, or NULL on error
 *
 * <pre>
 * Notes:
 *      (1) The copyflag specifies what we do with each box from boxas.
 *          Specifically, L_CLONE inserts a clone into boxad of each
 *          selected box from boxas.
 * </pre>
 */
BOXA *
boxaSelectRange(BOXA    *boxas,
                l_int32  first,
                l_int32  last,
                l_int32  copyflag)
{
l_int32  n, nbox, i;
BOX     *box;
BOXA    *boxad;

    PROCNAME("boxaSelectRange");

    if (!boxas)
        return (BOXA *)ERROR_PTR("boxas not defined", procName, NULL);
    if (copyflag != L_COPY && copyflag != L_CLONE)
        return (BOXA *)ERROR_PTR("invalid copyflag", procName, NULL);
    if ((n = boxaGetCount(boxas)) == 0) {
        L_WARNING("boxas is empty\n", procName);
        return boxaCopy(boxas, copyflag);
    }
    first = L_MAX(0, first);
    if (last < 0) last = n - 1;
    if (first >= n)
        return (BOXA *)ERROR_PTR("invalid first", procName, NULL);
    if (last >= n) {
        L_WARNING("last = %d is beyond max index = %d; adjusting\n",
                  procName, last, n - 1);
        last = n - 1;
    }
    if (first > last)
        return (BOXA *)ERROR_PTR("first > last", procName, NULL);

    nbox = last - first + 1;
    boxad = boxaCreate(nbox);
    for (i = first; i <= last; i++) {
        box = boxaGetBox(boxas, i, copyflag);
        boxaAddBox(boxad, box, L_INSERT);
    }
    return boxad;
}
Exemple #21
0
/*!
 * \brief   boxaSelectWithIndicator()
 *
 * \param[in]    boxas
 * \param[in]    na         indicator numa
 * \param[out]   pchanged   [optional] 1 if changed; 0 if clone returned
 * \return  boxad, or NULL on error
 *
 * <pre>
 * Notes:
 *      (1) Returns a copy of the boxa if no components are removed.
 *      (2) Uses box copies in the new boxa.
 *      (3) The indicator numa has values 0 (ignore) and 1 (accept).
 *      (4) If all indicator values are 0, the returned boxa is empty.
 * </pre>
 */
BOXA *
boxaSelectWithIndicator(BOXA     *boxas,
                        NUMA     *na,
                        l_int32  *pchanged)
{
l_int32  i, n, ival, nsave;
BOX     *box;
BOXA    *boxad;

    PROCNAME("boxaSelectWithIndicator");

    if (pchanged) *pchanged = FALSE;
    if (!boxas)
        return (BOXA *)ERROR_PTR("boxas not defined", procName, NULL);
    if (!na)
        return (BOXA *)ERROR_PTR("na not defined", procName, NULL);

    nsave = 0;
    n = numaGetCount(na);
    for (i = 0; i < n; i++) {
        numaGetIValue(na, i, &ival);
        if (ival == 1) nsave++;
    }

    if (nsave == n) {
        if (pchanged) *pchanged = FALSE;
        return boxaCopy(boxas, L_COPY);
    }
    if (pchanged) *pchanged = TRUE;
    boxad = boxaCreate(nsave);
    for (i = 0; i < n; i++) {
        numaGetIValue(na, i, &ival);
        if (ival == 0) continue;
        box = boxaGetBox(boxas, i, L_COPY);
        boxaAddBox(boxad, box, L_INSERT);
    }

    return boxad;
}
Exemple #22
0
/*!
 *  boxaaFlattenAligned()
 *
 *      Input:  boxaa
 *              num (number extracted from each)
 *              copyflag  (L_COPY or L_CLONE)
 *      Return: boxa, or null on error
 *
 *  Notes:
 *      (1) This 'flattens' the boxaa to a boxa, taking the first @num
 *          boxes from each boxa.
 *      (2) If less than @num boxes are in a boxa, we add invalid placeholder
 *          boxes to preserve the alignment between the input boxaa
 *          and the output boxa.
 */
BOXA *
boxaaFlattenAligned(BOXAA   *baa,
                    l_int32  num,
                    l_int32  copyflag)
{
l_int32  i, j, m, n, mval, nshort;
BOXA    *boxat, *boxad;
BOX     *box;

    PROCNAME("boxaaFlattenAligned");

    if (!baa)
        return (BOXA *)ERROR_PTR("baa not defined", procName, NULL);
    if (copyflag != L_COPY && copyflag != L_CLONE)
        return (BOXA *)ERROR_PTR("invalid copyflag", procName, NULL);

    n = boxaaGetCount(baa);
    boxad = boxaCreate(n);
    for (i = 0; i < n; i++) {
        boxat = boxaaGetBoxa(baa, i, L_CLONE);
        m = boxaGetCount(boxat);
        mval = L_MIN(m, num);
        nshort = num - mval;
        for (j = 0; j < mval; j++) {  /* take the first @num if possible */
            box = boxaGetBox(boxat, j, copyflag);
            boxaAddBox(boxad, box, L_INSERT);
        }
        for (j = 0; j < nshort; j++) {  /* add placeholders if necessary */
            box = boxCreate(0, 0, 0, 0);
            boxaAddBox(boxad, box, L_INSERT);
        }
        boxaDestroy(&boxat);
    }

    return boxad;
}
Exemple #23
0
/*!
 * \brief   pixConnCompBB()
 *
 * \param[in]    pixs 1 bpp
 * \param[in]    connectivity 4 or 8
 * \return  boxa, or NULL on error
 *
 * <pre>
 * Notes:
 *     (1) Finds bounding boxes of 4- or 8-connected components
 *         in a binary image.
 *     (2) This works on a copy of the input pix.  The c.c. are located
 *         in raster order and erased one at a time.  In the process,
 *         the b.b. is computed and saved.
 * </pre>
 */
BOXA *
pixConnCompBB(PIX     *pixs,
              l_int32  connectivity)
{
    l_int32   h, iszero;
    l_int32   x, y, xstart, ystart;
    PIX      *pixt;
    BOX      *box;
    BOXA     *boxa;
    L_STACK  *stack, *auxstack;

    PROCNAME("pixConnCompBB");

    if (!pixs || pixGetDepth(pixs) != 1)
        return (BOXA *)ERROR_PTR("pixs undefined or not 1 bpp", procName, NULL);
    if (connectivity != 4 && connectivity != 8)
        return (BOXA *)ERROR_PTR("connectivity not 4 or 8", procName, NULL);

    boxa = NULL;
    pixt = NULL;
    stack = NULL;

    pixZero(pixs, &iszero);
    if (iszero)
        return boxaCreate(1);  /* return empty boxa */

    if ((pixt = pixCopy(NULL, pixs)) == NULL)
        return (BOXA *)ERROR_PTR("pixt not made", procName, NULL);

    h = pixGetHeight(pixs);
    if ((stack = lstackCreate(h)) == NULL) {
        L_ERROR("stack not made\n", procName);
        goto cleanup;
    }
    auxstack = lstackCreate(0);
    stack->auxstack = auxstack;
    boxa = boxaCreate(0);

    xstart = 0;
    ystart = 0;
    while (1) {
        if (!nextOnPixelInRaster(pixt, xstart, ystart, &x, &y))
            break;

        if ((box = pixSeedfillBB(pixt, stack, x, y, connectivity)) == NULL) {
            L_ERROR("box not made\n", procName);
            boxaDestroy(&boxa);
            goto cleanup;
        }
        boxaAddBox(boxa, box, L_INSERT);

        xstart = x;
        ystart = y;
    }

#if  DEBUG
    pixCountPixels(pixt, &iszero, NULL);
    fprintf(stderr, "Number of remaining pixels = %d\n", iszero);
    pixWrite("junkremain", pixt1, IFF_PNG);
#endif  /* DEBUG */

    /* Cleanup, freeing the fillsegs on each stack */
cleanup:
    lstackDestroy(&stack, TRUE);
    pixDestroy(&pixt);
    return boxa;
}
Exemple #24
0
/*!
 * \brief   pixConnCompPixa()
 *
 * \param[in]    pixs 1 bpp
 * \param[out]   ppixa pixa of each c.c.
 * \param[in]    connectivity 4 or 8
 * \return  boxa, or NULL on error
 *
 * <pre>
 * Notes:
 *      (1) This finds bounding boxes of 4- or 8-connected components
 *          in a binary image, and saves images of each c.c
 *          in a pixa array.
 *      (2) It sets up 2 temporary pix, and for each c.c. that is
 *          located in raster order, it erases the c.c. from one pix,
 *          then uses the b.b. to extract the c.c. from the two pix using
 *          an XOR, and finally erases the c.c. from the second pix.
 *      (3) A clone of the returned boxa (where all boxes in the array
 *          are clones) is inserted into the pixa.
 *      (4) If the input is valid, this always returns a boxa and a pixa.
 *          If pixs is empty, the boxa and pixa will be empty.
 * </pre>
 */
BOXA *
pixConnCompPixa(PIX     *pixs,
                PIXA   **ppixa,
                l_int32  connectivity)
{
    l_int32   h, iszero;
    l_int32   x, y, xstart, ystart;
    PIX      *pix1, *pix2, *pix3, *pix4;
    PIXA     *pixa;
    BOX      *box;
    BOXA     *boxa;
    L_STACK  *stack, *auxstack;

    PROCNAME("pixConnCompPixa");

    if (!ppixa)
        return (BOXA *)ERROR_PTR("&pixa not defined", procName, NULL);
    *ppixa = NULL;
    if (!pixs || pixGetDepth(pixs) != 1)
        return (BOXA *)ERROR_PTR("pixs undefined or not 1 bpp", procName, NULL);
    if (connectivity != 4 && connectivity != 8)
        return (BOXA *)ERROR_PTR("connectivity not 4 or 8", procName, NULL);

    boxa = NULL;
    pix1 = pix2 = pix3 = pix4 = NULL;
    stack = NULL;

    pixZero(pixs, &iszero);
    if (iszero)
        return boxaCreate(1);  /* return empty boxa */

    pix1 = pixCopy(NULL, pixs);
    pix2 = pixCopy(NULL, pixs);
    if (!pix1 || !pix2) {
        L_ERROR("pix1 or pix2 not made\n", procName);
        goto cleanup;
    }

    h = pixGetHeight(pixs);
    if ((stack = lstackCreate(h)) == NULL) {
        L_ERROR("stack not made\n", procName);
        goto cleanup;
    }
    auxstack = lstackCreate(0);
    stack->auxstack = auxstack;
    pixa = pixaCreate(0);
    boxa = boxaCreate(0);

    xstart = 0;
    ystart = 0;
    while (1) {
        if (!nextOnPixelInRaster(pix1, xstart, ystart, &x, &y))
            break;

        if ((box = pixSeedfillBB(pix1, stack, x, y, connectivity)) == NULL) {
            L_ERROR("box not made\n", procName);
            pixaDestroy(&pixa);
            boxaDestroy(&boxa);
            goto cleanup;
        }
        boxaAddBox(boxa, box, L_INSERT);

        /* Save the c.c. and remove from pix2 as well */
        pix3 = pixClipRectangle(pix1, box, NULL);
        pix4 = pixClipRectangle(pix2, box, NULL);
        pixXor(pix3, pix3, pix4);
        pixRasterop(pix2, box->x, box->y, box->w, box->h, PIX_SRC ^ PIX_DST,
                    pix3, 0, 0);
        pixaAddPix(pixa, pix3, L_INSERT);
        pixDestroy(&pix4);

        xstart = x;
        ystart = y;
    }

#if  DEBUG
    pixCountPixels(pix1, &iszero, NULL);
    fprintf(stderr, "Number of remaining pixels = %d\n", iszero);
    pixWrite("junkremain", pix1, IFF_PNG);
#endif  /* DEBUG */

    /* Remove old boxa of pixa and replace with a clone copy */
    boxaDestroy(&pixa->boxa);
    pixa->boxa = boxaCopy(boxa, L_CLONE);
    *ppixa = pixa;

    /* Cleanup, freeing the fillsegs on each stack */
cleanup:
    lstackDestroy(&stack, TRUE);
    pixDestroy(&pix1);
    pixDestroy(&pix2);
    return boxa;
}
// 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(&regions_to_clear);
  pixaDestroy(&ccs);
  if (devanagari_split_debugimage) {
    DumpDebugImage(split_for_pageseg ? "pageseg_split_debug.png" :
                   "ocr_split_debug.png");
  }
  return true;
}
Exemple #26
0
int main(int    argc,
         char **argv)
{
l_uint8     *data1, *data2;
l_int32      i, same, w, h, width, success, nba;
size_t       size1, size2;
l_float32    diffarea, diffxor, scalefact;
BOX         *box;
BOXA        *boxa1, *boxa2, *boxa3;
BOXAA       *baa1, *baa2, *baa3;
PIX         *pix1, *pixdb;
PIXA        *pixa1, *pixa2;
L_REGPARAMS  *rp;

    if (regTestSetup(argc, argv, &rp))
        return 1;

    lept_mkdir("lept/boxa");

        /* Make a boxa and display its contents */
    boxa1 = boxaCreate(6);
    box = boxCreate(60, 60, 40, 20);
    boxaAddBox(boxa1, box, L_INSERT);
    box = boxCreate(120, 50, 20, 50);
    boxaAddBox(boxa1, box, L_INSERT);
    box = boxCreate(50, 140, 46, 60);
    boxaAddBox(boxa1, box, L_INSERT);
    box = boxCreate(166, 130, 64, 28);
    boxaAddBox(boxa1, box, L_INSERT);
    box = boxCreate(64, 224, 44, 34);
    boxaAddBox(boxa1, box, L_INSERT);
    box = boxCreate(117, 206, 26, 74);
    boxaAddBox(boxa1, box, L_INSERT);
    pix1 = DisplayBoxa(boxa1);
    regTestWritePixAndCheck(rp, pix1, IFF_PNG);  /* 0 */
    pixDisplayWithTitle(pix1, 0, 0, NULL, rp->display);
    pixDestroy(&pix1);

    boxaCompareRegions(boxa1, boxa1, 100, &same, &diffarea, &diffxor, NULL);
    regTestCompareValues(rp, 1, same, 0.0);  /* 1 */
    regTestCompareValues(rp, 0.0, diffarea, 0.0);  /* 2 */
    regTestCompareValues(rp, 0.0, diffxor, 0.0);  /* 3 */

    boxa2 = boxaTransform(boxa1, -13, -13, 1.0, 1.0);
    boxaCompareRegions(boxa1, boxa2, 10, &same, &diffarea, &diffxor, NULL);
    regTestCompareValues(rp, 1, same, 0.0);  /* 4 */
    regTestCompareValues(rp, 0.0, diffarea, 0.0);  /* 5 */
    regTestCompareValues(rp, 0.0, diffxor, 0.0);  /* 6 */
    boxaDestroy(&boxa2);

    boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP_AND_BOT, 6,
                                       L_ADJUST_CHOOSE_MIN, 1.0, 0);
    pix1 = DisplayBoxa(boxa2);
    regTestWritePixAndCheck(rp, pix1, IFF_PNG);  /* 7 */
    pixDisplayWithTitle(pix1, 200, 0, NULL, rp->display);
    pixDestroy(&pix1);

    boxaCompareRegions(boxa1, boxa2, 10, &same, &diffarea, &diffxor, &pixdb);
    regTestCompareValues(rp, 1, same, 0.0);  /* 8 */
    regTestCompareValues(rp, 0.053, diffarea, 0.002);  /* 9 */
    regTestCompareValues(rp, 0.240, diffxor, 0.002);  /* 10 */
    regTestWritePixAndCheck(rp, pixdb, IFF_PNG);  /* 11 */
    pixDisplayWithTitle(pixdb, 400, 0, NULL, rp->display);
    pixDestroy(&pixdb);
    boxaDestroy(&boxa1);
    boxaDestroy(&boxa2);

        /* Input is a fairly clean boxa */
    boxa1 = boxaRead("boxa1.ba");
    boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
                                       L_ADJUST_CHOOSE_MIN, 1.05, 1);
    width = 100;
    boxaGetExtent(boxa2, &w, &h, NULL);
    scalefact = (l_float32)width / (l_float32)w;
    boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
    pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
    regTestWritePixAndCheck(rp, pix1, IFF_PNG);  /* 12 */
    pixDisplayWithTitle(pix1, 600, 0, NULL, rp->display);
    pixDestroy(&pix1);
    boxaDestroy(&boxa1);
    boxaDestroy(&boxa2);
    boxaDestroy(&boxa3);

        /* Input is an unsmoothed and noisy boxa */
    boxa1 = boxaRead("boxa2.ba");
    boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
                                       L_ADJUST_CHOOSE_MIN, 1.05, 1);
    width = 100;
    boxaGetExtent(boxa2, &w, &h, NULL);
    scalefact = (l_float32)width / (l_float32)w;
    boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
    pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
    regTestWritePixAndCheck(rp, pix1, IFF_PNG);  /* 13 */
    pixDisplayWithTitle(pix1, 800, 0, NULL, rp->display);
    pixDestroy(&pix1);
    boxaDestroy(&boxa1);
    boxaDestroy(&boxa2);
    boxaDestroy(&boxa3);

        /* Input is a boxa smoothed with a median window filter */
    boxa1 = boxaRead("boxa3.ba");
    boxa2 = boxaReconcileEvenOddHeight(boxa1, L_ADJUST_TOP, 80,
                                       L_ADJUST_CHOOSE_MIN, 1.05, 1);
    width = 100;
    boxaGetExtent(boxa2, &w, &h, NULL);
    scalefact = (l_float32)width / (l_float32)w;
    boxa3 = boxaTransform(boxa2, 0, 0, scalefact, scalefact);
    pix1 = boxaDisplayTiled(boxa3, NULL, 1500, 2, 1.0, 0, 3, 2);
    regTestWritePixAndCheck(rp, pix1, IFF_PNG);  /* 14 */
    pixDisplayWithTitle(pix1, 1000, 0, NULL, rp->display);
    pixDestroy(&pix1);
    boxaDestroy(&boxa1);
    boxaDestroy(&boxa2);
    boxaDestroy(&boxa3);

        /* Test serialized boxa I/O to and from memory */
    data1 = l_binaryRead("boxa2.ba", &size1);
    boxa1 = boxaReadMem(data1, size1);
    boxaWriteMem(&data2, &size2, boxa1);
    boxa2 = boxaReadMem(data2, size2);
    boxaWrite("/tmp/lept/boxa/boxa1.ba", boxa1);
    boxaWrite("/tmp/lept/boxa/boxa2.ba", boxa2);
    filesAreIdentical("/tmp/lept/boxa/boxa1.ba", "/tmp/lept/boxa/boxa2.ba",
                      &same); 
    regTestCompareValues(rp, 1, same, 0.0);  /* 15 */
    boxaDestroy(&boxa1);
    boxaDestroy(&boxa2);
    lept_free(data1);
    lept_free(data2);

        /* ----------- Test pixaDisplayBoxaa() ------------ */
    pixa1 = pixaReadBoth("showboxes.pac");
    baa1 = boxaaRead("showboxes1.baa");
    baa2 = boxaaTranspose(baa1);
    baa3 = boxaaTranspose(baa2);
    nba = boxaaGetCount(baa1);
    success = TRUE;
    for (i = 0; i < nba; i++) {
        boxa1 = boxaaGetBoxa(baa1, i, L_CLONE);
        boxa2 = boxaaGetBoxa(baa3, i, L_CLONE);
        boxaEqual(boxa1, boxa2, 0, NULL, &same);
        boxaDestroy(&boxa1);
        boxaDestroy(&boxa2);
        if (!same) success = FALSE;
    }
        /* Check that the transpose is reversible */
    regTestCompareValues(rp, 1, success, 0.0);  /* 16 */
    pixa2 = pixaDisplayBoxaa(pixa1, baa2, L_DRAW_RGB, 2);
    pix1 = pixaDisplayTiledInRows(pixa2, 32, 1400, 1.0, 0, 10, 0);
    regTestWritePixAndCheck(rp, pix1, IFF_PNG);  /* 17 */
    pixDisplayWithTitle(pix1, 0, 600, NULL, rp->display);
    fprintf(stderr, "Writing to: /tmp/lept/boxa/show.pdf\n");
    l_pdfSetDateAndVersion(FALSE);
    pixaConvertToPdf(pixa2, 75, 1.0, 0, 0, NULL, "/tmp/lept/boxa/show.pdf");
    regTestCheckFile(rp, "/tmp/lept/boxa/show.pdf");  /* 18 */
    pixDestroy(&pix1);
    pixaDestroy(&pixa1);
    pixaDestroy(&pixa2);
    boxaaDestroy(&baa1);
    boxaaDestroy(&baa2);
    boxaaDestroy(&baa3);

    return regTestCleanup(rp);
}
/*!
 *  pixaGenerateFont()
 *
 *      Input:  pix (of 95 characters in 3 rows)
 *              fontsize (4, 6, 8, ... , 20, in pts at 300 ppi)
 *              &bl1 (<return> baseline of row 1)
 *              &bl2 (<return> baseline of row 2)
 *              &bl3 (<return> baseline of row 3)
 *      Return: pixa of font bitmaps for 95 characters, or null on error
 *
 *  Notes:
 *      (1) This does all the work.  See pixaGenerateFontFromFile()
 *          for an overview.
 *      (2) The pix is for one of the 9 fonts.  @fontsize is only
 *          used here for debugging.
 */
PIXA *
pixaGenerateFont(PIX      *pixs,
                 l_int32   fontsize,
                 l_int32  *pbl0,
                 l_int32  *pbl1,
                 l_int32  *pbl2)
{
l_int32   i, j, nrows, nrowchars, nchars, h, yval;
l_int32   width, height;
l_int32   baseline[3];
l_int32  *tab = NULL;
BOX      *box, *box1, *box2;
BOXA     *boxar, *boxac, *boxacs;
PIX      *pix1, *pix2, *pixr, *pixrc, *pixc;
PIXA     *pixa;
l_int32   n, w, inrow, top;
l_int32  *ia;
NUMA     *na;

    PROCNAME("pixaGenerateFont");

    if (!pbl0 || !pbl1 || !pbl2)
        return (PIXA *)ERROR_PTR("&bl not all defined", procName, NULL);
    *pbl0 = *pbl1 = *pbl2 = 0;
    if (!pixs)
        return (PIXA *)ERROR_PTR("pixs not defined", procName, NULL);

        /* Locate the 3 rows of characters */
    w = pixGetWidth(pixs);
    na = pixCountPixelsByRow(pixs, NULL);
    boxar = boxaCreate(0);
    n = numaGetCount(na);
    ia = numaGetIArray(na);
    inrow = 0;
    for (i = 0; i < n; i++) {
        if (!inrow && ia[i] > 0) {
            inrow = 1;
            top = i;
        } else if (inrow && ia[i] == 0) {
            inrow = 0;
            box = boxCreate(0, top, w, i - top);
            boxaAddBox(boxar, box, L_INSERT);
        }
    }
    FREE(ia);
    numaDestroy(&na);
    nrows = boxaGetCount(boxar);
#if  DEBUG_FONT_GEN
    L_INFO("For fontsize %s, have %d rows\n", procName, fontsize, nrows);
#endif  /* DEBUG_FONT_GEN */
    if (nrows != 3) {
        L_INFO("nrows = %d; skipping fontsize %d\n", procName, nrows, fontsize);
        return (PIXA *)ERROR_PTR("3 rows not generated", procName, NULL);
    }

        /* Grab the character images and baseline data */
#if DEBUG_BASELINE
    lept_rmdir("baseline");
    lept_mkdir("baseline");
#endif  /* DEBUG_BASELINE */
    tab = makePixelSumTab8();
    pixa = pixaCreate(95);
    for (i = 0; i < nrows; i++) {
        box = boxaGetBox(boxar, i, L_CLONE);
        pixr = pixClipRectangle(pixs, box, NULL);  /* row of chars */
        pixGetTextBaseline(pixr, tab, &yval);
        baseline[i] = yval;

#if DEBUG_BASELINE
        L_INFO("Baseline info: row %d, yval = %d, h = %d\n", procName,
               i, yval, pixGetHeight(pixr));
        pix1 = pixCopy(NULL, pixr);
        pixRenderLine(pix1, 0, yval, pixGetWidth(pix1), yval, 1,
                      L_FLIP_PIXELS);
        if (i == 0 )
            pixWrite("/tmp/baseline/row0.png", pix1, IFF_PNG);
        else if (i == 1)
            pixWrite("/tmp/baseline/row1.png", pix1, IFF_PNG);
        else
            pixWrite("/tmp/baseline/row2.png", pix1, IFF_PNG);
        pixDestroy(&pix1);
#endif  /* DEBUG_BASELINE */

        boxDestroy(&box);
        pixrc = pixCloseSafeBrick(NULL, pixr, 1, 35);
        boxac = pixConnComp(pixrc, NULL, 8);
        boxacs = boxaSort(boxac, L_SORT_BY_X, L_SORT_INCREASING, NULL);
        if (i == 0) {  /* consolidate the two components of '"' */
            box1 = boxaGetBox(boxacs, 1, L_CLONE);
            box2 = boxaGetBox(boxacs, 2, L_CLONE);
            box1->w = box2->x + box2->w - box1->x;  /* increase width */
            boxDestroy(&box1);
            boxDestroy(&box2);
            boxaRemoveBox(boxacs, 2);
        }
        h = pixGetHeight(pixr);
        nrowchars = boxaGetCount(boxacs);
        for (j = 0; j < nrowchars; j++) {
            box = boxaGetBox(boxacs, j, L_COPY);
            if (box->w <= 2 && box->h == 1) {  /* skip 1x1, 2x1 components */
                boxDestroy(&box);
                continue;
            }
            box->y = 0;
            box->h = h - 1;
            pixc = pixClipRectangle(pixr, box, NULL);
            boxDestroy(&box);
            if (i == 0 && j == 0)  /* add a pix for the space; change later */
                pixaAddPix(pixa, pixc, L_COPY);
            if (i == 2 && j == 0)  /* add a pix for the '\'; change later */
                pixaAddPix(pixa, pixc, L_COPY);
            pixaAddPix(pixa, pixc, L_INSERT);
        }
        pixDestroy(&pixr);
        pixDestroy(&pixrc);
        boxaDestroy(&boxac);
        boxaDestroy(&boxacs);
    }
    FREE(tab);

    nchars = pixaGetCount(pixa);
    if (nchars != 95)
        return (PIXA *)ERROR_PTR("95 chars not generated", procName, NULL);

    *pbl0 = baseline[0];
    *pbl1 = baseline[1];
    *pbl2 = baseline[2];

        /* Fix the space character up; it should have no ON pixels,
         * and be about twice as wide as the '!' character.    */
    pix1 = pixaGetPix(pixa, 0, L_CLONE);
    width = 2 * pixGetWidth(pix1);
    height = pixGetHeight(pix1);
    pixDestroy(&pix1);
    pix1 = pixCreate(width, height, 1);
    pixaReplacePix(pixa, 0, pix1, NULL);

        /* Fix up the '\' character; use a LR flip of the '/' char */
    pix1 = pixaGetPix(pixa, 15, L_CLONE);
    pix2 = pixFlipLR(NULL, pix1);
    pixDestroy(&pix1);
    pixaReplacePix(pixa, 60, pix2, NULL);

#if DEBUG_CHARS
    pix1 = pixaDisplayTiled(pixa, 1500, 0, 10);
    pixDisplay(pix1, 100 * i, 200);
    pixDestroy(&pix1);
#endif  /* DEBUG_CHARS */

    boxaDestroy(&boxar);
    return pixa;
}
Exemple #28
0
int main_find_pattern(int    argc,
	char **argv)
{
	char        *filein, *fileout, *patternfile;
	l_int32      w, h, i, n;
	BOX         *box, *boxe;
	BOXA        *boxa1, *boxa2;
	PIX         *pixs, *pixp, *pixpe;
	PIX         *pixd, *pixt1, *pixt2, *pixhmt;
	SEL         *sel_2h, *sel;
	static char  mainName[] = "findpattern1";

	filein = "feyn.tif";
	patternfile = "char.tif";
	fileout = "result.findpattern1";

	if ((pixs = pixRead(filein)) == NULL)
		printf("pixs not made\n");
	if ((pixp = pixRead(patternfile)) == NULL)
		printf("pixp not made\n");
	w = pixGetWidth(pixp);
	h = pixGetHeight(pixp);

	/* generate the hit-miss Sel with runs */
	sel = pixGenerateSelWithRuns(pixp, NumHorLines, NumVertLines, 0,
		MinRunlength, 7, 7, 0, 0, &pixpe);

	/* display the Sel two ways */
	selWriteStream(stderr, sel);
	pixt1 = pixDisplayHitMissSel(pixpe, sel, 9, HitColor, MissColor);
	pixDisplay(pixt1, 200, 200);
	pixWrite("junkpixt", pixt1, IFF_PNG);

	/* use the Sel to find all instances in the page */
	startTimer();
	pixhmt = pixHMT(NULL, pixs, sel);
	fprintf(stderr, "Time to find patterns = %7.3f\n", stopTimer());

	/* small erosion to remove noise; typically not necessary if
	* there are enough elements in the Sel */
	sel_2h = selCreateBrick(1, 2, 0, 0, SEL_HIT);
	pixt2 = pixErode(NULL, pixhmt, sel_2h);

	/* display the result visually by placing the Sel at each
	* location found */
	pixd = pixDilate(NULL, pixt2, sel);
	pixWrite(fileout, pixd, IFF_TIFF_G4);

	/* display outut with an outline around each located pattern */
	boxa1 = pixConnCompBB(pixt2, 8);
	n = boxaGetCount(boxa1);
	boxa2 = boxaCreate(n);
	for (i = 0; i < n; i++) {
		box = boxaGetBox(boxa1, i, L_COPY);
		boxe = boxCreate(box->x - w / 2, box->y - h / 2, w + 4, h + 4);
		boxaAddBox(boxa2, boxe, L_INSERT);
		pixRenderBox(pixs, boxe, 4, L_FLIP_PIXELS);
		boxDestroy(&box);
	}
	pixWrite("junkoutline", pixs, IFF_TIFF_G4);
	//boxaWriteStream(stderr, boxa2); //TODO ???

	pixDestroy(&pixs);
	pixDestroy(&pixp);
	pixDestroy(&pixpe);
	pixDestroy(&pixt1);
	pixDestroy(&pixt2);
	pixDestroy(&pixhmt);
	pixDestroy(&pixd);
	selDestroy(&sel);
	selDestroy(&sel_2h);
	boxaDestroy(&boxa1);
	boxaDestroy(&boxa2);

	printf("\n---\nEND\n");
	getchar();
	return 0;
}
std::vector<Figure> extractFigures(PIX *original, PageRegions &pageRegions,
                                   DocumentStatistics &docStats, bool verbose,
                                   bool showSteps,
                                   std::vector<Figure> &errors) {
  BOXA *bodytext = pageRegions.bodytext;
  BOXA *graphics = pageRegions.graphics;
  BOXA *captions = pageRegions.getCaptionsBoxa();
  std::vector<Caption> unassigned_captions = pageRegions.captions;
  int total_captions = captions->n;

  PIXA *steps = showSteps ? pixaCreate(4) : NULL;

  // Add bodyText boxes to fill up the margin
  BOX *margin;
  BOX *foreground;
  pixClipToForeground(original, NULL, &foreground);
  BOX *extent;
  boxaGetExtent(graphics, NULL, NULL, &extent);
  margin = boxBoundingRegion(extent, foreground);
  boxDestroy(&extent);
  boxaGetExtent(bodytext, NULL, NULL, &extent);
  margin = boxBoundingRegion(margin, extent);
  boxDestroy(&extent);
  boxaGetExtent(pageRegions.other, NULL, NULL, &extent);
  margin = boxBoundingRegion(margin, extent);
  int x = margin->x - 2, y = margin->y - 2, h = margin->h + 4,
      w = margin->w + 4;
  x = std::max(x, 0);
  y = std::max(y, 0);
  h = std::min((int)original->h, h);
  w = std::min((int)original->w, w);
  boxDestroy(&margin);
  boxaAddBox(bodytext, boxCreate(0, 0, original->w, y), L_CLONE);
  boxaAddBox(bodytext, boxCreate(0, y + h, original->w, original->h - y - h),
             L_CLONE);
  boxaAddBox(bodytext, boxCreate(0, 0, x, original->h), L_CLONE);
  boxaAddBox(bodytext, boxCreate(x + w, 0, original->w - x - w, original->h),
             L_CLONE);

  // Add captions to body text
  boxaJoin(bodytext, captions, 0, captions->n);

  if (showSteps)
    pixaAddPix(steps, original, L_CLONE);

  // Generate proposed regions for each caption box
  double center = original->w / 2.0;
  BOXAA *allProposals = boxaaCreate(captions->n);
  BOXA *claimedImages = boxaCreate(captions->n);
  for (int i = 0; i < captions->n; i++) {
    BOX *captBox = boxaGetBox(captions, i, L_CLONE);
    BOXA *proposals = boxaCreate(4);
    for (int j = 0; j < bodytext->n; j++) {
      BOX *txtBox = boxaGetBox(bodytext, j, L_CLONE);
      BOX *proposal = NULL;
      int tolerance = 2;
      int horizontal = 0;
      int vertical = 0;

      boxAlignment(captBox, txtBox, tolerance, &horizontal, &vertical);
      if (vertical * horizontal != 0 or (vertical == 0 and horizontal == 0)) {
        continue;
      }

      if (vertical == 0) {
        if (horizontal == 1) {
          proposal = boxRelocateOneSide(NULL, captBox,
                                        txtBox->x + txtBox->w + 2, L_FROM_LEFT);
        } else if (horizontal == -1) {
          proposal =
              boxRelocateOneSide(NULL, captBox, txtBox->x - 2, L_FROM_RIGHT);
        }
        boxExpandUD(proposal, bodytext);
        if (horizontal == -1) {
          proposal->w -= captBox->w + 1;
          proposal->x = captBox->x + captBox->w + 1;
        } else if (horizontal == 1) {
          proposal->w -= captBox->w + 1;
        }
      } else {
        if (vertical == 1) {
          proposal = boxRelocateOneSide(NULL, captBox,
                                        txtBox->y + txtBox->h + 3, L_FROM_TOP);
        } else if (vertical == -1) {
          proposal =
              boxRelocateOneSide(NULL, captBox, txtBox->y - 3, L_FROM_BOT);
        }
        boxExpandLR(proposal, bodytext);
        if (vertical == -1) {
          proposal->h -= captBox->h + 1;
          proposal->y = captBox->y + captBox->h + 1;
        } else if (vertical == 1) {
          proposal->h -= captBox->h + 1;
        }
      }

      // For two columns document, captions that do not
      // cross the center should not have regions pass the center
      if (docStats.documentIsTwoColumn()) {
        if (captBox->x + captBox->w <= center and
            proposal->x + proposal->w > center) {
          boxRelocateOneSide(proposal, proposal, center - 1, L_FROM_RIGHT);
        } else if (captBox->x >= center and proposal->x < center) {
          boxRelocateOneSide(proposal, proposal, center + 1, L_FROM_LEFT);
        }
      }

      BOX *clippedProposal;
      pixClipBoxToForeground(original, proposal, NULL, &clippedProposal);
      if (clippedProposal != NULL and
          scoreBox(clippedProposal, pageRegions.captions.at(i).type, bodytext,
                   graphics, claimedImages, original) > 0) {
        boxaAddBox(proposals, clippedProposal, L_CLONE);
      }
    }

    if (proposals->n > 0) {
      boxaaAddBoxa(allProposals, proposals, L_CLONE);
    } else {
      // Give up on this caption
      int on_caption = i - (total_captions - unassigned_captions.size());
      errors.push_back(Figure(unassigned_captions.at(on_caption), NULL));
      unassigned_captions.erase(unassigned_captions.begin() + on_caption);
    }
  }
  std::vector<Figure> figures = std::vector<Figure>();
  if (unassigned_captions.size() == 0) {
    return figures;
  }

  // Now go through every possible assignment of captions
  // to proposals pick the highest scorign one
  int numConfigurations = 1;
  for (int i = 0; i < allProposals->n; ++i) {
    numConfigurations *= allProposals->boxa[i]->n;
  }

  if (verbose)
    printf("Found %d possible configurations\n", numConfigurations);

  BOXA *bestProposals = NULL;
  std::vector<bool> bestKeep;
  int bestFound = -1;
  double bestScore = -1;
  for (int onConfig = 0; onConfig < numConfigurations; ++onConfig) {

    // Gather the proposed regions based on the configuration number
    int configNum = onConfig;
    BOXA *proposals = boxaCreate(allProposals->n);
    std::vector<bool> keep;
    for (int i = 0; i < allProposals->n; ++i) {
      int numProposals = allProposals->boxa[i]->n;
      int selected = configNum % numProposals;
      configNum = configNum / numProposals;
      boxaAddBox(proposals, allProposals->boxa[i]->box[selected], L_COPY);
    }

    // Attempt to split any overlapping regions
    for (int i = 0; i < proposals->n; ++i) {
      for (int j = i; j < proposals->n; ++j) {
        BOX *p1 = proposals->box[i];
        BOX *p2 = proposals->box[j];
        int eq;
        boxEqual(p1, p2, &eq);
        if (not eq)
          continue;
        int vertical, horizontal;
        boxAlignment(unassigned_captions.at(i).boundingBox,
                     unassigned_captions.at(j).boundingBox, 2, &horizontal,
                     &vertical);
        if (vertical == 0 or horizontal != 0)
          continue;

        double split = splitBoxVertical(original, p1);
        if (split > 0) {
          BOX *topClipped;
          BOX *botClipped;
          BOX *top = boxRelocateOneSide(NULL, p1, split - 1, L_FROM_BOT);
          pixClipBoxToForeground(original, top, NULL, &topClipped);
          BOX *bot = boxRelocateOneSide(NULL, p1, split + 1, L_FROM_TOP);
          pixClipBoxToForeground(original, bot, NULL, &botClipped);
          if (vertical == -1) {
            proposals->box[i] = topClipped;
            proposals->box[j] = botClipped;
          } else {
            proposals->box[i] = botClipped;
            proposals->box[j] = topClipped;
          }
          if (verbose)
            printf("Split a region vertically\n");
        }
      }
    }

    if (showSteps) {
      pixaAddPix(steps, pixDrawBoxa(original, proposals, 4, 0xff000000),
                 L_CLONE);
    }

    // Score the proposals
    int numFound = 0;
    double totalScore = 0;
    for (int i = 0; i < proposals->n; ++i) {
      double score =
          scoreBox(proposals->box[i], pageRegions.captions.at(i).type, bodytext,
                   graphics, proposals, original);
      totalScore += score;
      if (score > 0) {
        numFound += 1;
        keep.push_back(true);
      } else {
        keep.push_back(false);
      }
    }

    // Switch in for the current best needed
    if (numFound > bestFound or
        (numFound == bestFound and totalScore > bestScore)) {
      bestFound = numFound;
      bestScore = totalScore;
      bestProposals = proposals;
      bestKeep = keep;
    }
  }

  if (showSteps) {
    BOX *clip;
    PIXA *show = pixaCreate(4);
    pixClipBoxToForeground(original, NULL, NULL, &clip);
    int pad = 10;
    clip->x -= 10;
    clip->y -= 10;
    clip->w += pad * 2;
    clip->h += pad * 2;
    for (int i = 0; i < steps->n; ++i) {
      pixaAddPix(show, pixClipRectangle(steps->pix[i], clip, NULL), L_CLONE);
    }
    pixDisplay(pixaDisplayTiled(pixaConvertTo32(show), 4000, 1, 30), 0, 0);
  }

  for (int i = 0; i < bestProposals->n; ++i) {
    if (bestKeep.at(i)) {
      BOX *imageBox = bestProposals->box[i];
      int pad = 2;
      imageBox->x -= pad;
      imageBox->y -= pad;
      imageBox->w += pad * 2;
      imageBox->h += pad * 2;
      figures.push_back(Figure(unassigned_captions.at(i), imageBox));
    } else {
      errors.push_back(Figure(unassigned_captions.at(i), NULL));
    }
  }
  return figures;
}
Exemple #30
0
void k2pdfopt_reflow_bmp(KOPTContext *kctx) {
    K2PDFOPT_SETTINGS _k2settings, *k2settings;
    MASTERINFO _masterinfo, *masterinfo;
    WILLUSBITMAP _srcgrey, *srcgrey;
    WILLUSBITMAP *src, *dst;
    BMPREGION region;
    int i, bw, marbot, marleft;

    src = &kctx->src;
    srcgrey = &_srcgrey;
    bmp_init(srcgrey);

    k2settings = &_k2settings;
    masterinfo = &_masterinfo;
    /* Initialize settings */
    k2pdfopt_settings_init_from_koptcontext(k2settings, kctx);
    k2pdfopt_settings_quick_sanity_check(k2settings);
    /* Init for new source doc */
    k2pdfopt_settings_new_source_document_init(k2settings);
    /* Init master output structure */
    masterinfo_init(masterinfo, k2settings);
    wrapbmp_init(&masterinfo->wrapbmp, k2settings->dst_color);
    /* Init new source bitmap */
    bmpregion_init(&region);
    masterinfo_new_source_page_init(masterinfo, k2settings, src, srcgrey, NULL,
            &region, k2settings->src_rot, NULL, NULL, 1, -1, NULL );
    /* Set output size */
    k2pdfopt_settings_set_margins_and_devsize(k2settings,&region,masterinfo,-1.,0);
    /* Process single source page */
    bmpregion_source_page_add(&region, k2settings, masterinfo, 1, 0);
    wrapbmp_flush(masterinfo, k2settings, 0);

    if (fabs(k2settings->dst_gamma - 1.0) > .001)
        bmp_gamma_correct(&masterinfo->bmp, &masterinfo->bmp,
                k2settings->dst_gamma);

    /* copy master bitmap to context dst bitmap */
    dst = &kctx->dst;
    marbot = (int) (k2settings->dst_dpi * k2settings->dstmargins.box[1] + .5);
    marleft = (int) (k2settings->dst_dpi * k2settings->dstmargins.box[0] + .5);
    dst->bpp = masterinfo->bmp.bpp;
    dst->width = masterinfo->bmp.width;
    dst->height = masterinfo->rows > kctx->page_height ? masterinfo->rows + marbot : kctx->page_height;
    bmp_alloc(dst);
    bmp_fill(dst, 255, 255, 255);
    bw = bmp_bytewidth(&masterinfo->bmp);
    for (i = 0; i < masterinfo->rows; i++)
        memcpy(bmp_rowptr_from_top(dst, i),
                bmp_rowptr_from_top(&masterinfo->bmp, i), bw);

    kctx->page_width = kctx->dst.width;
    kctx->page_height = kctx->dst.height;
    kctx->precache = 0;

    int j;
    BOXA *rboxa = boxaCreate(masterinfo->rectmaps.n);
    BOXA *nboxa = boxaCreate(masterinfo->rectmaps.n);
    for (j = 0; j < masterinfo->rectmaps.n; j++) {
        WRECTMAP * rectmap = &masterinfo->rectmaps.wrectmap[j];
        rectmap->coords[1].x += marleft;
        BOX* rlbox = boxCreate(rectmap->coords[1].x,
                              rectmap->coords[1].y,
                              rectmap->coords[2].x,
                              rectmap->coords[2].y);
        BOX* nlbox = boxCreate(rectmap->coords[0].x*k2settings->src_dpi/rectmap->srcdpiw/kctx->zoom + kctx->bbox.x0,
                              rectmap->coords[0].y*k2settings->src_dpi/rectmap->srcdpih/kctx->zoom + kctx->bbox.y0,
                              rectmap->coords[2].x*k2settings->src_dpi/rectmap->srcdpiw/kctx->zoom,
                              rectmap->coords[2].y*k2settings->src_dpi/rectmap->srcdpih/kctx->zoom);
        boxaAddBox(rboxa, rlbox, L_INSERT);
        boxaAddBox(nboxa, nlbox, L_INSERT);
        wrectmaps_add_wrectmap(&kctx->rectmaps, rectmap);

        /*printf("rectmap:coords:\t%.1f %.1f\t%.1f %.1f\t%.1f %.1f\t%.1f %.1f\n",
                rectmap->coords[0].x, rectmap->coords[0].y,
                rectmap->coords[1].x, rectmap->coords[1].y,
                rectmap->coords[2].x, rectmap->coords[2].y,
                rectmap->srcdpiw,     rectmap->srcdpih);*/
    }
    /* 2D sort the bounding boxes of these words. */
    BOXAA *rbaa = boxaSort2d(rboxa, NULL, 3, -5, 5);
    BOXAA *nbaa = boxaSort2d(nboxa, NULL, 3, -5, 5);

    /* Flatten the boxaa, saving the boxa index for each box */
    kctx->rboxa = boxaaFlattenToBoxa(rbaa, &kctx->rnai, L_CLONE);
    kctx->nboxa = boxaaFlattenToBoxa(nbaa, &kctx->nnai, L_CLONE);

    boxaDestroy(&rboxa);
    boxaaDestroy(&rbaa);
    boxaDestroy(&nboxa);
    boxaaDestroy(&nbaa);

    bmp_free(src);
    bmp_free(srcgrey);
    bmpregion_free(&region);
    masterinfo_free(masterinfo, k2settings);
}