/*
** bmp8 must be grayscale
** (x1,y1) and (x2,y2) from top left of bitmap
*/
void jocr_single_word_from_bmp8(char *text,int maxlen,WILLUSBITMAP *bmp8,
int x1,int y1,int x2,int y2,int allow_spaces,
int std_proc)
{
job_t *job,_job;
int i,w,h,dw,dh,bw;
unsigned char *src,*dst;
// static char *funcname="jocr_single_word_from_bmp8";
char *buf;
int pgm2asc(job_t *job);
char *getTextLine(int);
if (x1>x2)
{
w=x1;
x1=x2;
x2=w;
}
w=x2-x1+1;
bw=w/40;
if (bw<6)
bw=6;
dw=w+bw*2;
if (y1>y2)
{
h=y1;
y1=y2;
y2=h;
}
h=y2-y1+1;
dh=h+bw*2;
job=&_job;
job_init(job);
job_init_image(job);
// willus_mem_alloc_warn((void **)&job->src.p.p,w*h,funcname,10);
/* Must use malloc since job_free_image counts on this. */
job->src.p.p=malloc(dw*dh);
job->src.p.x=dw;
job->src.p.y=dh;
job->src.p.bpp=1;
src=bmp_rowptr_from_top(bmp8,y1)+x1;
memset(job->src.p.p,255,dw*dh);
dst=(unsigned char *)job->src.p.p + dw*bw + bw;
for (i=y1;i<=y2;i++,dst+=dw,src+=bmp8->width)
memcpy(dst,src,w);
pgm2asc(job);
buf=getTextLine(0);
if (buf)
{
strncpy(text,buf,maxlen-1);
text[maxlen-1]='\0';
if (std_proc)
ocr_text_proc(text,allow_spaces);
}
else
text[0]='\0';
// willus_mem_free((double **)&job->src.p.p,funcname);
job_free_image(job);
}
/* -------------------------------------------------------------
// ------ MAIN - replace this by your own aplication!
// ------------------------------------------------------------- */
int main(int argn, char *argv[]) {
int multipnm=1;
job_t job1, *job; /* fixme, dont want global variables for lib */
job=OCR_JOB=&job1;
setvbuf(stdout, (char *) NULL, _IONBF, 0); /* not buffered */
job_init(job); /* init cfg and db */
process_arguments(job, argn, argv);
/* load character data base (JS1002: now outside pgm2asc) */
if ( job->cfg.mode & 2 ) /* check for db-option flag */
load_db(job);
/* load_db uses readpnm() and would conflict with multi images */
while (multipnm==1) { /* multi-image loop */
job_init_image(job); /* single image */
mark_start(job);
multipnm = read_picture(job);
/* separation of main and rest for using as lib
this will be changed later => introduction of set_option()
for better communication to the engine */
if (multipnm<0) break; /* read error */
/* call main loop */
pgm2asc(job);
mark_end(job);
print_output(job);
job_free_image(job);
}
return ((multipnm<0)?multipnm:0); /* -1=255 on error, 0 ok */
}
char* gocr_main(job_t* job) {
int multipnm=1;
char* output;
setvbuf(stdout, (char *) NULL, _IONBF, 0); /* not buffered */
/* load character data base (JS1002: now outside pgm2asc) */
if ( job->cfg.mode & 2 ) /* check for db-option flag */
load_db(job);
/* load_db uses readpnm() and would conflict with multi images */
while (multipnm==1) { /* multi-image loop */
job_init_image(job); /* single image */
mark_start(job);
multipnm = read_picture(job);
/* separation of main and rest for using as lib
this will be changed later => introduction of set_option()
for better communication to the engine */
if (multipnm<0) break; /* read error */
/* call main loop */
pgm2asc(job);
mark_end(job);
output = print_output(job);
job_free_image(job);
}
// return ((multipnm<0)?multipnm:0); /* -1=255 on error, 0 ok */
return output;
}
char get_atom_label(const Magick::Image &image, const Magick::ColorGray &bg, int x1, int y1, int x2, int y2,
double THRESHOLD, int dropx, int dropy, bool no_filtering, bool verbose, bool numbers)
{
char c = UNKNOWN_CHAR;
const int width = x2 - x1 + 1;
const int height = y2 - y1 + 1;
unsigned char *pixmap = (unsigned char *) malloc(width * height);
for (int i = y1; i <= y2; i++)
for (int j = x1; j <= x2; j++)
pixmap[(i - y1) * width + j - x1] = (unsigned char) (255 - 255 * get_pixel(image, bg, j, i, THRESHOLD));
// Here we drop down from the top of the box, middle of x coordinate and extract connected component
int t = 1;
int y = dropy - y1 + 1;
int x = dropx - x1;
while ((t != 0) && (y < height))
{
t = pixmap[y * width + x];
y++;
}
if (t != 0)
{
free(pixmap);
return 0;
}
#pragma omp critical
{
y--;
pixmap[y * width + x] = 2;
list<int> cx;
list<int> cy;
cx.push_back(x);
cy.push_back(y);
while (!cx.empty())
{
x = cx.front();
y = cy.front();
cx.pop_front();
cy.pop_front();
pixmap[y * width + x] = 1;
// this goes around 3x3 square touching the chosen pixel
for (int i = x - 1; i < x + 2; i++)
for (int j = y - 1; j < y + 2; j++)
if (i < width && j < height && i >= 0 && j >= 0 && pixmap[j * width + i] == 0)
{
cx.push_back(i);
cy.push_back(j);
pixmap[j * width + i] = 2;
}
}
// Flatten the bitmap. Note: the bitmap is inverted after this cycle (255 means "empty", 0 means "pixel").
for (int i = 0; i < height; i++)
for (int j = 0; j < width; j++)
pixmap[i * width + j] = (pixmap[i * width + j] == 1 ? 0 : 255);
job_t gocr_job;
// The list of all characters, that can be recognised as atom label:
string char_filter = RECOGNIZED_CHARS;
if (numbers) char_filter = "1";
if (no_filtering) char_filter.clear();
job_init(&gocr_job);
job_init_image(&gocr_job);
//gocr_job.cfg.cs = 160;
//gocr_job.cfg.certainty = 80;
//gocr_job.cfg.dust_size = 1;
gocr_job.src.p.x = width;
gocr_job.src.p.y = height;
gocr_job.src.p.bpp = 1;
gocr_job.src.p.p = pixmap;
if (char_filter.empty())
gocr_job.cfg.cfilter = (char*)NULL;
else
gocr_job.cfg.cfilter = (char*) char_filter.c_str();
struct OCRAD_Pixmap *ocrad_pixmap = new OCRAD_Pixmap();
unsigned char *ocrad_bitmap = (unsigned char *) malloc(width * height);
memset(ocrad_bitmap, 0, width * height);
ocrad_pixmap->height = height;
ocrad_pixmap->width = width;
ocrad_pixmap->mode = OCRAD_bitmap;
ocrad_pixmap->data = ocrad_bitmap;
// Number of non-zero pixels on the bitmap, excluding the 1px border:
int pixmap_pixels_count = 0;
// Number of zero pixels on the bitmap, excluding the 1px border:
int pixmap_zeros_count = 0;
// The code below initialises "opix->data" buffer ("bitmap_data") for OCRAD from "tmp" buffer:
#ifdef HAVE_CUNEIFORM_LIB
Magick::Image cuneiform_img(Magick::Geometry(2 * width + 2, height), "white");
// From cuneiform_src/cli/cuneiform-cli.cpp::preprocess_image(Magick::Image&):168
cuneiform_img.monochrome();
cuneiform_img.type(Magick::BilevelType);
#endif
for (int y = 0; y < height; y++)
{
for (int x = 0; x < width; x++)
{
if (pixmap[y * width + x] == 0)
{
ocrad_bitmap[y * width + x] = 1;
#ifdef HAVE_CUNEIFORM_LIB
// Draw two identical samples that follow one another. We do so because Cuneiform has difficulties in recognizing single characters:
cuneiform_img.pixelColor(x, y, "black");
cuneiform_img.pixelColor(x + width + 2, y, "black");
#endif
if (x > 0 && x < width - 1 && y > 0 && y < height - 1)
pixmap_pixels_count++;
}
else if (x > 0 && x < width - 1 && y > 0 && y < height - 1)
pixmap_zeros_count++;
}
}
if (verbose)
{
cout << "Box to OCR: " << x1 << "x" << y1 << "-" << x2 << "x" << y2 << " w/h: " << width << "x" << height << endl;
for (int i = 0; i < height; i++)
{
for (int j = 0; j < width; j++)
cout << (gocr_job.src.p.p[i * width + j] / 255 ? '#' : '.');
cout << endl;
}
}
if (pixmap_pixels_count <= MIN_CHAR_POINTS || pixmap_zeros_count <= MIN_CHAR_POINTS)
goto FINALIZE;
c = osra_gocr_ocr(gocr_job);
if (verbose)
cout << "GOCR: c=" << c << endl;
//c = UNKNOWN_CHAR; // Switch off GOCR recognition
// Character recognition succeeded for GOCR:
if (c != UNKNOWN_CHAR)
goto FINALIZE;
// Character recognition failed for GOCR and we try OCRAD:
c = osra_ocrad_ocr(ocrad_pixmap, char_filter);
if (verbose)
cout << "OCRAD: c=" << c << endl;
//c = UNKNOWN_CHAR; // Switch off OCRAD recognition
// Character recognition succeeded for OCRAD:
if (c != UNKNOWN_CHAR)
goto FINALIZE;
#ifdef HAVE_TESSERACT_LIB
c = osra_tesseract_ocr(gocr_job.src.p.p, width, height, char_filter);
if (verbose)
cout << "Tesseract: c=" << c << endl;
//c = UNKNOWN_CHAR; // Switch off Tesseract recognition
// Character recognition succeeded for Tesseract:
if (c != UNKNOWN_CHAR)
goto FINALIZE;
#endif
#ifdef HAVE_CUNEIFORM_LIB
// TODO: Why box width should be more than 7 for Cuneiform?
if (width <= 7)
goto FINALIZE;
c = osra_cuneiform_ocr(cuneiform_img, char_filter);
if (verbose)
cout << "Cuneiform: c=" << c << endl;
//c = UNKNOWN_CHAR; // Switch off Cuneiform recognition
#endif
FINALIZE:
// "pixmap" is freed together with "gocr_job".
job_free_image(&gocr_job);
OCR_JOB = NULL;
JOB = NULL;
delete ocrad_pixmap; // delete OCRAD Pixmap
free(ocrad_bitmap);
// TODO: Why there are problems with "7" with a given box size? If the problem is engine-specific, it should be moved to appropriate section
if (c == '7' && (width <= 10 || height <= 20))
c = UNKNOWN_CHAR;
} // #pragma omp critical
return(c == UNKNOWN_CHAR ? 0 : c);
}