void Textord::cleanup_blocks(bool clean_noise, BLOCK_LIST* blocks) {
BLOCK_IT block_it = blocks; //iterator
ROW_IT row_it; //row iterator
int num_rows = 0;
int num_rows_all = 0;
int num_blocks = 0;
int num_blocks_all = 0;
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()) {
cleanup_nontext_block(block);
continue;
}
num_rows = 0;
num_rows_all = 0;
if (clean_noise) {
row_it.set_to_list(block->row_list());
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
ROW* row = row_it.data();
++num_rows_all;
clean_small_noise_from_words(row);
if ((textord_noise_rejrows && !row->word_list()->empty() &&
clean_noise_from_row(row)) ||
row->word_list()->empty()) {
delete row_it.extract(); // lose empty row.
} else {
if (textord_noise_rejwords)
clean_noise_from_words(row_it.data());
if (textord_blshift_maxshift >= 0)
tweak_row_baseline(row, textord_blshift_maxshift,
textord_blshift_xfraction);
++num_rows;
}
}
}
if (block->row_list()->empty()) {
delete block_it.extract(); // Lose empty text blocks.
} else {
++num_blocks;
}
++num_blocks_all;
if (textord_noise_debug)
tprintf("cleanup_blocks: # rows = %d / %d\n", num_rows, num_rows_all);
}
if (textord_noise_debug)
tprintf("cleanup_blocks: # blocks = %d / %d\n", num_blocks, num_blocks_all);
}
void PrintSegmentationStats(BLOCK_LIST* block_list) {
int num_blocks = 0;
int num_rows = 0;
int num_words = 0;
int num_blobs = 0;
BLOCK_IT block_it(block_list);
for (block_it.mark_cycle_pt(); !block_it.cycled_list(); block_it.forward()) {
BLOCK* block = block_it.data();
++num_blocks;
ROW_IT row_it(block->row_list());
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
++num_rows;
ROW* row = row_it.data();
// Iterate over all werds in the row.
WERD_IT werd_it(row->word_list());
for (werd_it.mark_cycle_pt(); !werd_it.cycled_list(); werd_it.forward()) {
WERD* werd = werd_it.data();
++num_words;
num_blobs += werd->cblob_list()->length();
}
}
}
tprintf("Block list stats:\nBlocks = %d\nRows = %d\nWords = %d\nBlobs = %d\n",
num_blocks, num_rows, num_words, num_blobs);
}
/// Builds a PAGE_RES from the block_list in the way required for ApplyBoxes:
/// All fuzzy spaces are removed, and all the words are maximally chopped.
PAGE_RES* Tesseract::SetupApplyBoxes(const GenericVector<TBOX>& boxes,
BLOCK_LIST *block_list) {
PreenXHeights(block_list);
// Strip all fuzzy space markers to simplify the PAGE_RES.
BLOCK_IT b_it(block_list);
for (b_it.mark_cycle_pt(); !b_it.cycled_list(); b_it.forward()) {
BLOCK* block = b_it.data();
ROW_IT r_it(block->row_list());
for (r_it.mark_cycle_pt(); !r_it.cycled_list(); r_it.forward ()) {
ROW* row = r_it.data();
WERD_IT w_it(row->word_list());
for (w_it.mark_cycle_pt(); !w_it.cycled_list(); w_it.forward()) {
WERD* word = w_it.data();
if (word->cblob_list()->empty()) {
delete w_it.extract();
} else {
word->set_flag(W_FUZZY_SP, false);
word->set_flag(W_FUZZY_NON, false);
}
}
}
}
PAGE_RES* page_res = new PAGE_RES(false, block_list, NULL);
PAGE_RES_IT pr_it(page_res);
WERD_RES* word_res;
while ((word_res = pr_it.word()) != NULL) {
MaximallyChopWord(boxes, pr_it.block()->block,
pr_it.row()->row, word_res);
pr_it.forward();
}
return page_res;
}
void RefreshWordBlobsFromNewBlobs(BLOCK_LIST* block_list,
C_BLOB_LIST* new_blobs,
C_BLOB_LIST* not_found_blobs) {
// Now iterate over all the blobs in the segmentation_block_list_, and just
// replace the corresponding c-blobs inside the werds.
BLOCK_IT block_it(block_list);
for (block_it.mark_cycle_pt(); !block_it.cycled_list(); block_it.forward()) {
BLOCK* block = block_it.data();
if (block->poly_block() != NULL && !block->poly_block()->IsText())
continue; // Don't touch non-text blocks.
// Iterate over all rows in the block.
ROW_IT row_it(block->row_list());
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
ROW* row = row_it.data();
// Iterate over all werds in the row.
WERD_IT werd_it(row->word_list());
WERD_LIST new_words;
WERD_IT new_words_it(&new_words);
for (werd_it.mark_cycle_pt(); !werd_it.cycled_list(); werd_it.forward()) {
WERD* werd = werd_it.extract();
WERD* new_werd = werd->ConstructWerdWithNewBlobs(new_blobs,
not_found_blobs);
if (new_werd) {
// Insert this new werd into the actual row's werd-list. Remove the
// existing one.
new_words_it.add_after_then_move(new_werd);
delete werd;
} else {
// Reinsert the older word back, for lack of better options.
// This is critical since dropping the words messes up segmentation:
// eg. 1st word in the row might otherwise have W_FUZZY_NON turned on.
new_words_it.add_after_then_move(werd);
}
}
// Get rid of the old word list & replace it with the new one.
row->word_list()->clear();
werd_it.move_to_first();
werd_it.add_list_after(&new_words);
}
}
}
void apply_box_training(BLOCK_LIST *block_list) {
BLOCK_IT block_it(block_list);
ROW_IT row_it;
ROW *row;
WERD_IT word_it;
WERD *word;
WERD *bln_word;
WERD copy_outword; // copy to denorm
PBLOB_IT blob_it;
DENORM denorm;
INT16 count = 0;
char ch[2];
ch[1] = '\0';
tprintf ("Generating training data\n");
for (block_it.mark_cycle_pt ();
!block_it.cycled_list (); block_it.forward ()) {
row_it.set_to_list (block_it.data ()->row_list ());
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
word_it.set_to_list (row->word_list ());
for (word_it.mark_cycle_pt ();
!word_it.cycled_list (); word_it.forward ()) {
word = word_it.data ();
if ((strlen (word->text ()) == 1) &&
(word->gblob_list ()->length () == 1)) {
/* Here is a word with a single char label and a single blob so train on it */
bln_word =
make_bln_copy (word, row, row->x_height (), &denorm);
blob_it.set_to_list (bln_word->blob_list ());
ch[0] = *word->text ();
tess_training_tester (blob_it.data (),
//single blob
&denorm, TRUE, //correct
ch, //correct ASCII char
1, //ASCII length
NULL);
copy_outword = *(bln_word);
copy_outword.baseline_denormalise (&denorm);
blob_it.set_to_list (copy_outword.blob_list ());
ch[0] = *word->text ();
delete bln_word;
count++;
}
}
}
}
tprintf ("Generated training data for %d blobs\n", count);
}
void ExtractBlobsFromSegmentation(BLOCK_LIST* blocks,
C_BLOB_LIST* output_blob_list) {
C_BLOB_IT return_list_it(output_blob_list);
BLOCK_IT block_it(blocks);
for (block_it.mark_cycle_pt(); !block_it.cycled_list(); block_it.forward()) {
BLOCK* block = block_it.data();
ROW_IT row_it(block->row_list());
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
ROW* row = row_it.data();
// Iterate over all werds in the row.
WERD_IT werd_it(row->word_list());
for (werd_it.mark_cycle_pt(); !werd_it.cycled_list(); werd_it.forward()) {
WERD* werd = werd_it.data();
return_list_it.move_to_last();
return_list_it.add_list_after(werd->cblob_list());
return_list_it.move_to_last();
return_list_it.add_list_after(werd->rej_cblob_list());
}
}
}
}
// Fixes the block so it obeys all the rules:
// Must have at least one ROW.
// Must have at least one WERD.
// WERDs contain a fake blob.
void Textord::cleanup_nontext_block(BLOCK* block) {
// Non-text blocks must contain at least one row.
ROW_IT row_it(block->row_list());
if (row_it.empty()) {
const TBOX& box = block->pdblk.bounding_box();
float height = box.height();
int32_t xstarts[2] = {box.left(), box.right()};
double coeffs[3] = {0.0, 0.0, static_cast<double>(box.bottom())};
ROW* row = new ROW(1, xstarts, coeffs, height / 2.0f, height / 4.0f,
height / 4.0f, 0, 1);
row_it.add_after_then_move(row);
}
// Each row must contain at least one word.
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
ROW* row = row_it.data();
WERD_IT w_it(row->word_list());
if (w_it.empty()) {
// Make a fake blob to put in the word.
TBOX box = block->row_list()->singleton() ? block->pdblk.bounding_box()
: row->bounding_box();
C_BLOB* blob = C_BLOB::FakeBlob(box);
C_BLOB_LIST blobs;
C_BLOB_IT blob_it(&blobs);
blob_it.add_after_then_move(blob);
WERD* word = new WERD(&blobs, 0, nullptr);
w_it.add_after_then_move(word);
}
// Each word must contain a fake blob.
for (w_it.mark_cycle_pt(); !w_it.cycled_list(); w_it.forward()) {
WERD* word = w_it.data();
// Just assert that this is true, as it would be useful to find
// out why it isn't.
ASSERT_HOST(!word->cblob_list()->empty());
}
row->recalc_bounding_box();
}
}
void apply_box_testing(BLOCK_LIST *block_list) {
BLOCK_IT block_it(block_list);
ROW_IT row_it;
ROW *row;
INT16 row_count = 0;
WERD_IT word_it;
WERD *word;
WERD *bln_word;
INT16 word_count = 0;
PBLOB_IT blob_it;
DENORM denorm;
INT16 count = 0;
char ch[2];
WERD *outword; //bln best choice
//segmentation
WERD_CHOICE *best_choice; //tess output
WERD_CHOICE *raw_choice; //top choice permuter
//detailed results
BLOB_CHOICE_LIST_CLIST blob_choices;
INT16 char_count = 0;
INT16 correct_count = 0;
INT16 err_count = 0;
INT16 rej_count = 0;
#ifndef SECURE_NAMES
WERDSTATS wordstats; //As from newdiff
#endif
char tess_rej_str[3];
char tess_long_str[3];
ch[1] = '\0';
strcpy (tess_rej_str, "|A");
strcpy (tess_long_str, "|B");
for (block_it.mark_cycle_pt ();
!block_it.cycled_list (); block_it.forward ()) {
row_it.set_to_list (block_it.data ()->row_list ());
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
row_count++;
word_count = 0;
word_it.set_to_list (row->word_list ());
for (word_it.mark_cycle_pt ();
!word_it.cycled_list (); word_it.forward ()) {
word = word_it.data ();
word_count++;
if ((strlen (word->text ()) == 1) &&
!STRING (applybox_test_exclusions).contains (*word->text ())
&& (word->gblob_list ()->length () == 1)) {
/* Here is a word with a single char label and a single blob so test it */
bln_word =
make_bln_copy (word, row, row->x_height (), &denorm);
blob_it.set_to_list (bln_word->blob_list ());
ch[0] = *word->text ();
char_count++;
best_choice = tess_segment_pass1 (bln_word,
&denorm,
tess_default_matcher,
raw_choice,
&blob_choices, outword);
/*
Test for TESS screw up on word. Recog_word has already ensured that the
choice list, outword blob lists and best_choice string are the same
length. A TESS screw up is indicated by a blank filled or 0 length string.
*/
if ((best_choice->string ().length () == 0) ||
(strspn (best_choice->string ().string (), " ") ==
best_choice->string ().length ())) {
rej_count++;
tprintf ("%d:%d: \"%s\" -> TESS FAILED\n",
row_count, word_count, ch);
#ifndef SECURE_NAMES
wordstats.word (tess_rej_str, 2, ch, 1);
#endif
}
else {
if ((best_choice->string ().length () !=
outword->blob_list ()->length ()) ||
(best_choice->string ().length () !=
blob_choices.length ())) {
tprintf
("ASSERT FAIL String:\"%s\"; Strlen=%d; #Blobs=%d; #Choices=%d\n",
best_choice->string ().string (),
best_choice->string ().length (),
outword->blob_list ()->length (),
blob_choices.length ());
}
ASSERT_HOST (best_choice->string ().length () ==
outword->blob_list ()->length ());
ASSERT_HOST (best_choice->string ().length () ==
blob_choices.length ());
fix_quotes ((char *) best_choice->string ().string (),
//turn to double
outword, &blob_choices);
if (strcmp (best_choice->string ().string (), ch) != 0) {
err_count++;
tprintf ("%d:%d: \"%s\" -> \"%s\"\n",
row_count, word_count, ch,
best_choice->string ().string ());
}
else
correct_count++;
#ifndef SECURE_NAMES
if (best_choice->string ().length () > 2)
wordstats.word (tess_long_str, 2, ch, 1);
else
wordstats.word ((char *) best_choice->string ().
string (),
best_choice->string ().length (), ch,
1);
#endif
}
delete bln_word;
delete outword;
delete best_choice;
delete raw_choice;
blob_choices.deep_clear ();
count++;
}
}
}
}
#ifndef SECURE_NAMES
wordstats.print (1, 100.0);
wordstats.conf_matrix ();
tprintf ("Tested %d chars: %d correct; %d rejected by tess; %d errs\n",
char_count, correct_count, rej_count, err_count);
#endif
}
/*************************************************************************
* tidy_up()
* - report >1 block
* - sort the words in each row.
* - report any rows with no labelled words.
* - report any remaining unlabelled words
* - report total labelled words
*
*************************************************************************/
void tidy_up( //
BLOCK_LIST *block_list, //real blocks
INT16 &ok_char_count,
INT16 &ok_row_count,
INT16 &unlabelled_words,
INT16 *tgt_char_counts,
INT16 &rebalance_count,
char &min_char,
INT16 &min_samples,
INT16 &final_labelled_blob_count) {
BLOCK_IT block_it(block_list);
ROW_IT row_it;
ROW *row;
WERD_IT word_it;
WERD *word;
WERD *duplicate_word;
INT16 block_idx = 0;
INT16 row_idx;
INT16 all_row_idx = 0;
BOOL8 row_ok;
BOOL8 rebalance_needed = FALSE;
//No. of unique labelled samples
INT16 labelled_char_counts[128];
INT16 i;
char ch;
char prev_ch = '\0';
BOOL8 at_dupe_of_prev_word;
ROW *prev_row = NULL;
INT16 left;
INT16 prev_left = -1;
for (i = 0; i < 128; i++)
labelled_char_counts[i] = 0;
ok_char_count = 0;
ok_row_count = 0;
unlabelled_words = 0;
if ((applybox_debug > 4) && (block_it.length () != 1))
tprintf ("APPLY_BOXES: More than one block??\n");
for (block_it.mark_cycle_pt ();
!block_it.cycled_list (); block_it.forward ()) {
block_idx++;
row_idx = 0;
row_ok = FALSE;
row_it.set_to_list (block_it.data ()->row_list ());
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row_idx++;
all_row_idx++;
row = row_it.data ();
word_it.set_to_list (row->word_list ());
word_it.sort (word_comparator);
for (word_it.mark_cycle_pt ();
!word_it.cycled_list (); word_it.forward ()) {
word = word_it.data ();
if (strlen (word->text ()) == 0) {
unlabelled_words++;
if (applybox_debug > 4) {
tprintf
("APPLY_BOXES: Unlabelled word blk:%d row:%d allrows:%d\n",
block_idx, row_idx, all_row_idx);
}
}
else {
if (word->gblob_list ()->length () != 1)
tprintf
("APPLY_BOXES: FATALITY - MULTIBLOB Labelled word blk:%d row:%d allrows:%d\n",
block_idx, row_idx, all_row_idx);
ok_char_count++;
labelled_char_counts[*word->text ()]++;
row_ok = TRUE;
}
}
if ((applybox_debug > 4) && (!row_ok)) {
tprintf
("APPLY_BOXES: Row with no labelled words blk:%d row:%d allrows:%d\n",
block_idx, row_idx, all_row_idx);
}
else
ok_row_count++;
}
}
min_samples = 9999;
for (i = 0; i < 128; i++) {
if (tgt_char_counts[i] > labelled_char_counts[i]) {
if (labelled_char_counts[i] <= 1) {
tprintf
("APPLY_BOXES: FATALITY - %d labelled samples of \"%c\" - target is %d\n",
labelled_char_counts[i], (char) i, tgt_char_counts[i]);
}
else {
rebalance_needed = TRUE;
if (applybox_debug > 0)
tprintf
("APPLY_BOXES: REBALANCE REQD \"%c\" - target of %d from %d labelled samples\n",
(char) i, tgt_char_counts[i], labelled_char_counts[i]);
}
}
if ((min_samples > labelled_char_counts[i]) && (tgt_char_counts[i] > 0)) {
min_samples = labelled_char_counts[i];
min_char = (char) i;
}
}
while (applybox_rebalance && rebalance_needed) {
block_it.set_to_list (block_list);
for (block_it.mark_cycle_pt ();
!block_it.cycled_list (); block_it.forward ()) {
row_it.set_to_list (block_it.data ()->row_list ());
for (row_it.mark_cycle_pt ();
!row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
word_it.set_to_list (row->word_list ());
for (word_it.mark_cycle_pt ();
!word_it.cycled_list (); word_it.forward ()) {
word = word_it.data ();
left = word->bounding_box ().left ();
ch = *word->text ();
at_dupe_of_prev_word = ((row == prev_row) &&
(left = prev_left) &&
(ch == prev_ch));
if ((ch != '\0') &&
(labelled_char_counts[ch] > 1) &&
(tgt_char_counts[ch] > labelled_char_counts[ch]) &&
(!at_dupe_of_prev_word)) {
/* Duplicate the word to rebalance the labelled samples */
if (applybox_debug > 9) {
tprintf ("Duping \"%c\" from ", ch);
word->bounding_box ().print ();
}
duplicate_word = new WERD;
*duplicate_word = *word;
word_it.add_after_then_move (duplicate_word);
rebalance_count++;
labelled_char_counts[ch]++;
}
prev_row = row;
prev_left = left;
prev_ch = ch;
}
}
}
rebalance_needed = FALSE;
for (i = 0; i < 128; i++) {
if ((tgt_char_counts[i] > labelled_char_counts[i]) &&
(labelled_char_counts[i] > 1)) {
rebalance_needed = TRUE;
break;
}
}
}
/* Now final check - count labelled blobs */
final_labelled_blob_count = 0;
block_it.set_to_list (block_list);
for (block_it.mark_cycle_pt ();
!block_it.cycled_list (); block_it.forward ()) {
row_it.set_to_list (block_it.data ()->row_list ());
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
word_it.set_to_list (row->word_list ());
word_it.sort (word_comparator);
for (word_it.mark_cycle_pt ();
!word_it.cycled_list (); word_it.forward ()) {
word = word_it.data ();
if ((strlen (word->text ()) == 1) &&
(word->gblob_list ()->length () == 1))
final_labelled_blob_count++;
}
}
}
}
ROW *find_row_of_box( //
BLOCK_LIST *block_list, //real blocks
BOX box, //from boxfile
INT16 &block_id,
INT16 &row_id_to_process) {
BLOCK_IT block_it(block_list);
BLOCK *block;
ROW_IT row_it;
ROW *row;
ROW *row_to_process = NULL;
INT16 row_id;
WERD_IT word_it;
WERD *word;
BOOL8 polyg;
PBLOB_IT blob_it;
PBLOB *blob;
OUTLINE_IT outline_it;
OUTLINE *outline;
/*
Find row to process - error if box REALLY overlaps more than one row. (I.e
it overlaps blobs in the row - not just overlaps the bounding box of the
whole row.)
*/
block_id = 0;
for (block_it.mark_cycle_pt ();
!block_it.cycled_list (); block_it.forward ()) {
block_id++;
row_id = 0;
block = block_it.data ();
if (block->bounding_box ().overlap (box)) {
row_it.set_to_list (block->row_list ());
for (row_it.mark_cycle_pt ();
!row_it.cycled_list (); row_it.forward ()) {
row_id++;
row = row_it.data ();
if (row->bounding_box ().overlap (box)) {
word_it.set_to_list (row->word_list ());
for (word_it.mark_cycle_pt ();
!word_it.cycled_list (); word_it.forward ()) {
word = word_it.data ();
polyg = word->flag (W_POLYGON);
if (word->bounding_box ().overlap (box)) {
blob_it.set_to_list (word->gblob_list ());
for (blob_it.mark_cycle_pt ();
!blob_it.cycled_list (); blob_it.forward ()) {
blob = blob_it.data ();
if (gblob_bounding_box (blob, polyg).
overlap (box)) {
outline_it.
set_to_list (gblob_out_list
(blob, polyg));
for (outline_it.mark_cycle_pt ();
!outline_it.cycled_list ();
outline_it.forward ()) {
outline = outline_it.data ();
if (goutline_bounding_box
(outline, polyg).major_overlap (box)) {
if ((row_to_process == NULL) ||
(row_to_process == row)) {
row_to_process = row;
row_id_to_process = row_id;
}
else
/* RETURN ERROR Box overlaps blobs in more than one row */
return NULL;
}
}
}
}
}
}
}
}
}
}
return row_to_process;
}
// Groups blocks by rotation, then, for each group, makes a WordGrid and calls
// TransferDiacriticsToWords to copy the diacritic blobs to the most
// appropriate words in the group of blocks. Source blobs are not touched.
void Textord::TransferDiacriticsToBlockGroups(BLOBNBOX_LIST* diacritic_blobs,
BLOCK_LIST* blocks) {
// Angle difference larger than this is too much to consider equal.
// They should only be in multiples of M_PI/2 anyway.
const double kMaxAngleDiff = 0.01; // About 0.6 degrees.
PointerVector<BlockGroup> groups;
BLOCK_IT bk_it(blocks);
for (bk_it.mark_cycle_pt(); !bk_it.cycled_list(); bk_it.forward()) {
BLOCK* block = bk_it.data();
if (block->pdblk.poly_block() != nullptr && !block->pdblk.poly_block()->IsText()) {
continue;
}
// Linear search of the groups to find a matching rotation.
float block_angle = block->re_rotation().angle();
int best_g = 0;
float best_angle_diff = MAX_FLOAT32;
for (int g = 0; g < groups.size(); ++g) {
double angle_diff = fabs(block_angle - groups[g]->angle);
if (angle_diff > M_PI) angle_diff = fabs(angle_diff - 2.0 * M_PI);
if (angle_diff < best_angle_diff) {
best_angle_diff = angle_diff;
best_g = g;
}
}
if (best_angle_diff > kMaxAngleDiff) {
groups.push_back(new BlockGroup(block));
} else {
groups[best_g]->blocks.push_back(block);
groups[best_g]->bounding_box += block->pdblk.bounding_box();
float x_height = block->x_height();
if (x_height < groups[best_g]->min_xheight)
groups[best_g]->min_xheight = x_height;
}
}
// Now process each group of blocks.
PointerVector<WordWithBox> word_ptrs;
for (int g = 0; g < groups.size(); ++g) {
const BlockGroup* group = groups[g];
if (group->bounding_box.null_box()) continue;
WordGrid word_grid(group->min_xheight, group->bounding_box.botleft(),
group->bounding_box.topright());
for (int b = 0; b < group->blocks.size(); ++b) {
ROW_IT row_it(group->blocks[b]->row_list());
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
ROW* row = row_it.data();
// Put the words of the row into the grid.
WERD_IT w_it(row->word_list());
for (w_it.mark_cycle_pt(); !w_it.cycled_list(); w_it.forward()) {
WERD* word = w_it.data();
WordWithBox* box_word = new WordWithBox(word);
word_grid.InsertBBox(true, true, box_word);
// Save the pointer where it will be auto-deleted.
word_ptrs.push_back(box_word);
}
}
}
FCOORD rotation = group->rotation;
// Make it a forward rotation that will transform blob coords to block.
rotation.set_y(-rotation.y());
TransferDiacriticsToWords(diacritic_blobs, rotation, &word_grid);
}
}
/// Consume all source blobs that strongly overlap the given box,
/// putting them into a new word, with the correct_text label.
/// Fights over which box owns which blobs are settled by
/// applying the blobs to box or next_box with the least non-overlap.
/// @return false if the box was in error, which can only be caused by
/// failing to find an overlapping blob for a box.
bool Tesseract::ResegmentWordBox(BLOCK_LIST *block_list,
const TBOX& box, const TBOX& next_box,
const char* correct_text) {
if (applybox_debug > 1) {
tprintf("\nAPPLY_BOX: in ResegmentWordBox() for %s\n", correct_text);
}
WERD* new_word = NULL;
BLOCK_IT b_it(block_list);
for (b_it.mark_cycle_pt(); !b_it.cycled_list(); b_it.forward()) {
BLOCK* block = b_it.data();
if (!box.major_overlap(block->bounding_box()))
continue;
ROW_IT r_it(block->row_list());
for (r_it.mark_cycle_pt(); !r_it.cycled_list(); r_it.forward()) {
ROW* row = r_it.data();
if (!box.major_overlap(row->bounding_box()))
continue;
WERD_IT w_it(row->word_list());
for (w_it.mark_cycle_pt(); !w_it.cycled_list(); w_it.forward()) {
WERD* word = w_it.data();
if (applybox_debug > 2) {
tprintf("Checking word:");
word->bounding_box().print();
}
if (word->text() != NULL && word->text()[0] != '\0')
continue; // Ignore words that are already done.
if (!box.major_overlap(word->bounding_box()))
continue;
C_BLOB_IT blob_it(word->cblob_list());
for (blob_it.mark_cycle_pt(); !blob_it.cycled_list();
blob_it.forward()) {
C_BLOB* blob = blob_it.data();
TBOX blob_box = blob->bounding_box();
if (!blob_box.major_overlap(box))
continue;
double current_box_miss_metric = BoxMissMetric(blob_box, box);
double next_box_miss_metric = BoxMissMetric(blob_box, next_box);
if (applybox_debug > 2) {
tprintf("Checking blob:");
blob_box.print();
tprintf("Current miss metric = %g, next = %g\n",
current_box_miss_metric, next_box_miss_metric);
}
if (current_box_miss_metric > next_box_miss_metric)
continue; // Blob is a better match for next box.
if (applybox_debug > 2) {
tprintf("Blob match: blob:");
blob_box.print();
tprintf("Matches box:");
box.print();
tprintf("With next box:");
next_box.print();
}
if (new_word == NULL) {
// Make a new word with a single blob.
new_word = word->shallow_copy();
new_word->set_text(correct_text);
w_it.add_to_end(new_word);
}
C_BLOB_IT new_blob_it(new_word->cblob_list());
new_blob_it.add_to_end(blob_it.extract());
}
}
}
}
if (new_word == NULL && applybox_debug > 0) tprintf("FAIL!\n");
return new_word != NULL;
}
