void find_components(
BLOCK_LIST *blocks,
TO_BLOCK_LIST *land_blocks,
TO_BLOCK_LIST *port_blocks,
TBOX *page_box) {
BLOCK *block; //current block
PDBLK_CLIST pd_blocks; //copy of list
BLOCK_IT block_it = blocks; //iterator
PDBLK_C_IT pd_it = &pd_blocks; //iterator
IMAGE thresh_image; //thresholded
int width = page_image.get_xsize();
int height = page_image.get_ysize();
if (width > MAX_INT16 || height > MAX_INT16) {
tprintf("Input image too large! (%d, %d)\n", width, height);
return; // Can't handle it.
}
ICOORD page_tr(width, height);
block_it.set_to_list (blocks);
if (global_monitor != NULL)
global_monitor->ocr_alive = TRUE;
set_global_loc_code(LOC_EDGE_PROG);
if (!page_image.white_high ())
invert_image(&page_image);
#ifndef EMBEDDED
previous_cpu = clock ();
#endif
for (block_it.mark_cycle_pt(); !block_it.cycled_list();
block_it.forward()) {
block = block_it.data();
if (block->poly_block() == NULL ||
block->poly_block()->IsText()) {
#ifndef GRAPHICS_DISABLED
extract_edges(NULL, &page_image, &page_image, page_tr, block);
#else
extract_edges(&page_image, &page_image, page_tr, block);
#endif
*page_box += block->bounding_box ();
}
}
if (global_monitor != NULL) {
global_monitor->ocr_alive = TRUE;
global_monitor->progress = 10;
}
assign_blobs_to_blocks2(blocks, land_blocks, port_blocks);
if (global_monitor != NULL)
global_monitor->ocr_alive = TRUE;
filter_blobs (page_box->topright (), land_blocks, textord_test_landscape);
#ifndef EMBEDDED
previous_cpu = clock ();
#endif
filter_blobs (page_box->topright (), port_blocks, !textord_test_landscape);
if (global_monitor != NULL)
global_monitor->ocr_alive = TRUE;
}
/// Return a TBLOB * from the whole page_image.
/// To be freed later with free_blob().
TBLOB *make_tesseract_blob(float baseline, float xheight, float descender, float ascender) {
BLOCK *block = new BLOCK ("a character",
TRUE,
0, 0,
0, 0,
page_image.get_xsize(),
page_image.get_ysize());
// Create C_BLOBs from the page
extract_edges(
#ifndef GRAPHICS_DISABLED
NULL,
#endif
&page_image, &page_image,
ICOORD(page_image.get_xsize(), page_image.get_ysize()),
block);
// Create one PBLOB from all C_BLOBs
C_BLOB_LIST *list = block->blob_list();
C_BLOB_IT c_blob_it(list);
PBLOB *pblob = new PBLOB; // will be (hopefully) deleted by the pblob_list
for (c_blob_it.mark_cycle_pt();
!c_blob_it.cycled_list();
c_blob_it.forward()) {
C_BLOB *c_blob = c_blob_it.data();
PBLOB c_as_p(c_blob, baseline + xheight);
merge_blobs(pblob, &c_as_p);
}
PBLOB_LIST *pblob_list = new PBLOB_LIST; // will be deleted by the word
PBLOB_IT pblob_it(pblob_list);
pblob_it.add_after_then_move(pblob);
// Normalize PBLOB
WERD word(pblob_list, 0, " ");
ROW *row = make_tess_ocrrow(baseline, xheight, descender, ascender);
word.baseline_normalise(row);
delete row;
// Create a TBLOB from PBLOB
return make_tess_blob(pblob, /* flatten: */ TRUE);
}
// Perform steps to prepare underlying binary image/other data structures for
// OCR. The current segmentation is required by this method.
// Note that this method resets pix_binary_ to the original binarized image,
// which may be different from the image actually used for OCR depending on the
// value of devanagari_ocr_split_strategy.
void Tesseract::PrepareForTessOCR(BLOCK_LIST* block_list,
Tesseract* osd_tess, OSResults* osr) {
// Find the max splitter strategy over all langs.
ShiroRekhaSplitter::SplitStrategy max_ocr_strategy =
static_cast<ShiroRekhaSplitter::SplitStrategy>(
static_cast<inT32>(ocr_devanagari_split_strategy));
for (int i = 0; i < sub_langs_.size(); ++i) {
ShiroRekhaSplitter::SplitStrategy ocr_strategy =
static_cast<ShiroRekhaSplitter::SplitStrategy>(
static_cast<inT32>(sub_langs_[i]->ocr_devanagari_split_strategy));
if (ocr_strategy > max_ocr_strategy)
max_ocr_strategy = ocr_strategy;
}
// Utilize the segmentation information available.
splitter_.set_segmentation_block_list(block_list);
splitter_.set_ocr_split_strategy(max_ocr_strategy);
// Run the splitter for OCR
bool split_for_ocr = splitter_.Split(false);
// Restore pix_binary to the binarized original pix for future reference.
ASSERT_HOST(splitter_.orig_pix());
pixDestroy(&pix_binary_);
pix_binary_ = pixClone(splitter_.orig_pix());
// If the pageseg and ocr strategies are different, refresh the block list
// (from the last SegmentImage call) with blobs from the real image to be used
// for OCR.
if (splitter_.HasDifferentSplitStrategies()) {
BLOCK block("", TRUE, 0, 0, 0, 0, pixGetWidth(pix_binary_),
pixGetHeight(pix_binary_));
Pix* pix_for_ocr = split_for_ocr ? splitter_.splitted_image() :
splitter_.orig_pix();
extract_edges(pix_for_ocr, &block);
splitter_.RefreshSegmentationWithNewBlobs(block.blob_list());
}
// The splitter isn't needed any more after this, so save memory by clearing.
splitter_.Clear();
}
void Textord::find_components(Pix* pix, BLOCK_LIST *blocks,
TO_BLOCK_LIST *to_blocks) {
int width = pixGetWidth(pix);
int height = pixGetHeight(pix);
if (width > INT16_MAX || height > INT16_MAX) {
tprintf("Input image too large! (%d, %d)\n", width, height);
return; // Can't handle it.
}
set_global_loc_code(LOC_EDGE_PROG);
BLOCK_IT block_it(blocks); // iterator
for (block_it.mark_cycle_pt(); !block_it.cycled_list();
block_it.forward()) {
BLOCK* block = block_it.data();
if (block->pdblk.poly_block() == nullptr || block->pdblk.poly_block()->IsText()) {
extract_edges(pix, block);
}
}
assign_blobs_to_blocks2(pix, blocks, to_blocks);
ICOORD page_tr(width, height);
filter_blobs(page_tr, to_blocks, !textord_test_landscape);
}
void edges_and_textord( //read .pb file
const char *filename, //.pb file
BLOCK_LIST *blocks) {
BLOCK *block; //current block
char *lastdot; //of name
STRING name = filename; //truncated name
ICOORD page_tr;
BOX page_box; //bounding_box
PDBLK_CLIST pd_blocks; //copy of list
BLOCK_IT block_it = blocks; //iterator
PDBLK_C_IT pd_it = &pd_blocks; //iterator
//different orientations
TO_BLOCK_LIST land_blocks, port_blocks;
IMAGE thresh_image; //thresholded
lastdot = strrchr (name.string (), '.');
if (lastdot != NULL)
*lastdot = '\0';
if (page_image.get_bpp () == 0) {
name += tessedit_image_ext;
if (page_image.read_header (name.string ()))
CANTOPENFILE.error ("edges_and_textord", EXIT, name.string ());
if (page_image.read (0))
READFAILED.error ("edges_and_textord", EXIT, name.string ());
name = filename;
lastdot = strrchr (name.string (), '.');
if (lastdot != NULL)
*lastdot = '\0';
}
page_tr = ICOORD (page_image.get_xsize (), page_image.get_ysize ());
read_pd_file (name, page_image.get_xsize (), page_image.get_ysize (),
blocks);
block_it.set_to_list (blocks);
if (global_monitor != NULL)
global_monitor->ocr_alive = TRUE;
if (page_image.get_bpp () > 1) {
set_global_loc_code(LOC_ADAPTIVE);
for (block_it.mark_cycle_pt (); !block_it.cycled_list ();
block_it.forward ()) {
block = block_it.data ();
pd_it.add_after_then_move (block);
}
// adaptive_threshold(&page_image,&pd_blocks,&thresh_image);
set_global_loc_code(LOC_EDGE_PROG);
#ifndef EMBEDDED
previous_cpu = clock ();
#endif
for (block_it.mark_cycle_pt (); !block_it.cycled_list ();
block_it.forward ()) {
block = block_it.data ();
if (!polygon_tess_approximation)
invert_image(&page_image);
#ifndef GRAPHICS_DISABLED
extract_edges(NO_WINDOW, &page_image, &thresh_image, page_tr, block);
#else
extract_edges(&page_image, &thresh_image, page_tr, block);
#endif
page_box += block->bounding_box ();
}
page_image = thresh_image; //everyone else gets it
}
else {
set_global_loc_code(LOC_EDGE_PROG);
if (!page_image.white_high ())
invert_image(&page_image);
#ifndef EMBEDDED
previous_cpu = clock ();
#endif
for (block_it.mark_cycle_pt (); !block_it.cycled_list ();
block_it.forward ()) {
block = block_it.data ();
#ifndef GRAPHICS_DISABLED
extract_edges(NO_WINDOW, &page_image, &page_image, page_tr, block);
#else
extract_edges(&page_image, &page_image, page_tr, block);
#endif
page_box += block->bounding_box ();
}
}
if (global_monitor != NULL) {
global_monitor->ocr_alive = TRUE;
global_monitor->progress = 10;
}
assign_blobs_to_blocks2(blocks, &land_blocks, &port_blocks);
if (global_monitor != NULL)
global_monitor->ocr_alive = TRUE;
filter_blobs (page_box.topright (), &land_blocks, textord_test_landscape);
#ifndef EMBEDDED
previous_cpu = clock ();
#endif
filter_blobs (page_box.topright (), &port_blocks, !textord_test_landscape);
if (global_monitor != NULL)
global_monitor->ocr_alive = TRUE;
textord_page (page_box.topright (), blocks, &land_blocks, &port_blocks);
}
