forked from tesseract-ocr/tesseract
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcontrol.cpp
2072 lines (1958 loc) · 76.8 KB
/
control.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/******************************************************************
* File: control.cpp (Formerly control.c)
* Description: Module-independent matcher controller.
* Author: Ray Smith
* Created: Thu Apr 23 11:09:58 BST 1992
* ReHacked: Tue Sep 22 08:42:49 BST 1992 Phil Cheatle
*
* (C) Copyright 1992, Hewlett-Packard Ltd.
** Licensed under the Apache License, Version 2.0 (the "License");
** you may not use this file except in compliance with the License.
** You may obtain a copy of the License at
** http://www.apache.org/licenses/LICENSE-2.0
** Unless required by applicable law or agreed to in writing, software
** distributed under the License is distributed on an "AS IS" BASIS,
** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
** See the License for the specific language governing permissions and
** limitations under the License.
*
**********************************************************************/
// Include automatically generated configuration file if running autoconf.
#ifdef HAVE_CONFIG_H
#include "config_auto.h"
#endif
#include <string.h>
#include <math.h>
#ifdef __UNIX__
#include <assert.h>
#include <unistd.h>
#include <errno.h>
#endif
#include <ctype.h>
#include "callcpp.h"
#include "control.h"
#include "docqual.h"
#include "drawfx.h"
#include "fixspace.h"
#include "globals.h"
#include "lstmrecognizer.h"
#include "ocrclass.h"
#include "output.h"
#include "pgedit.h"
#include "reject.h"
#include "sorthelper.h"
#include "tessbox.h"
#include "tesseractclass.h"
#include "tessvars.h"
#include "werdit.h"
#define MIN_FONT_ROW_COUNT 8
#define MAX_XHEIGHT_DIFF 3
const char* const kBackUpConfigFile = "tempconfigdata.config";
// Min believable x-height for any text when refitting as a fraction of
// original x-height
const double kMinRefitXHeightFraction = 0.5;
/**
* Make a word from the selected blobs and run Tess on them.
*
* @param page_res recognise blobs
* @param selection_box within this box
*/
namespace tesseract {
void Tesseract::recog_pseudo_word(PAGE_RES* page_res,
TBOX &selection_box) {
PAGE_RES_IT* it = make_pseudo_word(page_res, selection_box);
if (it != NULL) {
recog_interactive(it);
it->DeleteCurrentWord();
delete it;
}
}
/**
* Recognize a single word in interactive mode.
*
* @param pr_it the page results iterator
*/
BOOL8 Tesseract::recog_interactive(PAGE_RES_IT* pr_it) {
inT16 char_qual;
inT16 good_char_qual;
WordData word_data(*pr_it);
SetupWordPassN(2, &word_data);
// LSTM doesn't run on pass2, but we want to run pass2 for tesseract.
if (lstm_recognizer_ == NULL) {
classify_word_and_language(2, pr_it, &word_data);
} else {
classify_word_and_language(1, pr_it, &word_data);
}
if (tessedit_debug_quality_metrics) {
WERD_RES* word_res = pr_it->word();
word_char_quality(word_res, pr_it->row()->row, &char_qual, &good_char_qual);
tprintf("\n%d chars; word_blob_quality: %d; outline_errs: %d; "
"char_quality: %d; good_char_quality: %d\n",
word_res->reject_map.length(),
word_blob_quality(word_res, pr_it->row()->row),
word_outline_errs(word_res), char_qual, good_char_qual);
}
return TRUE;
}
// Helper function to check for a target word and handle it appropriately.
// Inspired by Jetsoft's requirement to process only single words on pass2
// and beyond.
// If word_config is not null:
// If the word_box and target_word_box overlap, read the word_config file
// else reset to previous config data.
// return true.
// else
// If the word_box and target_word_box overlap or pass <= 1, return true.
// Note that this function uses a fixed temporary file for storing the previous
// configs, so it is neither thread-safe, nor process-safe, but the assumption
// is that it will only be used for one debug window at a time.
//
// Since this function is used for debugging (and not to change OCR results)
// set only debug params from the word config file.
bool Tesseract::ProcessTargetWord(const TBOX& word_box,
const TBOX& target_word_box,
const char* word_config,
int pass) {
if (word_config != NULL) {
if (word_box.major_overlap(target_word_box)) {
if (backup_config_file_ == NULL) {
backup_config_file_ = kBackUpConfigFile;
FILE* config_fp = fopen(backup_config_file_, "wb");
ParamUtils::PrintParams(config_fp, params());
fclose(config_fp);
ParamUtils::ReadParamsFile(word_config,
SET_PARAM_CONSTRAINT_DEBUG_ONLY,
params());
}
} else {
if (backup_config_file_ != NULL) {
ParamUtils::ReadParamsFile(backup_config_file_,
SET_PARAM_CONSTRAINT_DEBUG_ONLY,
params());
backup_config_file_ = NULL;
}
}
} else if (pass > 1 && !word_box.major_overlap(target_word_box)) {
return false;
}
return true;
}
/** If tesseract is to be run, sets the words up ready for it. */
void Tesseract::SetupAllWordsPassN(int pass_n,
const TBOX* target_word_box,
const char* word_config,
PAGE_RES* page_res,
GenericVector<WordData>* words) {
// Prepare all the words.
PAGE_RES_IT page_res_it(page_res);
for (page_res_it.restart_page(); page_res_it.word() != NULL;
page_res_it.forward()) {
if (target_word_box == NULL ||
ProcessTargetWord(page_res_it.word()->word->bounding_box(),
*target_word_box, word_config, 1)) {
words->push_back(WordData(page_res_it));
}
}
// Setup all the words for recognition with polygonal approximation.
for (int w = 0; w < words->size(); ++w) {
SetupWordPassN(pass_n, &(*words)[w]);
if (w > 0) (*words)[w].prev_word = &(*words)[w - 1];
}
}
// Sets up the single word ready for whichever engine is to be run.
void Tesseract::SetupWordPassN(int pass_n, WordData* word) {
if (pass_n == 1 || !word->word->done) {
if (pass_n == 1) {
word->word->SetupForRecognition(unicharset, this, BestPix(),
tessedit_ocr_engine_mode, NULL,
classify_bln_numeric_mode,
textord_use_cjk_fp_model,
poly_allow_detailed_fx,
word->row, word->block);
} else if (pass_n == 2) {
// TODO(rays) Should we do this on pass1 too?
word->word->caps_height = 0.0;
if (word->word->x_height == 0.0f)
word->word->x_height = word->row->x_height();
}
word->lang_words.truncate(0);
for (int s = 0; s <= sub_langs_.size(); ++s) {
// The sub_langs_.size() entry is for the master language.
Tesseract* lang_t = s < sub_langs_.size() ? sub_langs_[s] : this;
WERD_RES* word_res = new WERD_RES;
word_res->InitForRetryRecognition(*word->word);
word->lang_words.push_back(word_res);
// LSTM doesn't get setup for pass2.
if (pass_n == 1 || lang_t->tessedit_ocr_engine_mode != OEM_LSTM_ONLY) {
word_res->SetupForRecognition(
lang_t->unicharset, lang_t, BestPix(),
lang_t->tessedit_ocr_engine_mode, NULL,
lang_t->classify_bln_numeric_mode,
lang_t->textord_use_cjk_fp_model,
lang_t->poly_allow_detailed_fx, word->row, word->block);
}
}
}
}
// Runs word recognition on all the words.
bool Tesseract::RecogAllWordsPassN(int pass_n, ETEXT_DESC* monitor,
PAGE_RES_IT* pr_it,
GenericVector<WordData>* words) {
// TODO(rays) Before this loop can be parallelized (it would yield a massive
// speed-up) all remaining member globals need to be converted to local/heap
// (eg set_pass1 and set_pass2) and an intermediate adaption pass needs to be
// added. The results will be significantly different with adaption on, and
// deterioration will need investigation.
pr_it->restart_page();
for (int w = 0; w < words->size(); ++w) {
WordData* word = &(*words)[w];
if (w > 0) word->prev_word = &(*words)[w - 1];
if (monitor != NULL) {
monitor->ocr_alive = TRUE;
if (pass_n == 1) {
monitor->progress = 70 * w / words->size();
if (monitor->progress_callback != NULL) {
TBOX box = pr_it->word()->word->bounding_box();
(*monitor->progress_callback)(monitor->progress, box.left(),
box.right(), box.top(), box.bottom());
}
} else {
monitor->progress = 70 + 30 * w / words->size();
if (monitor->progress_callback != NULL) {
(*monitor->progress_callback)(monitor->progress, 0, 0, 0, 0);
}
}
if (monitor->deadline_exceeded() ||
(monitor->cancel != NULL && (*monitor->cancel)(monitor->cancel_this,
words->size()))) {
// Timeout. Fake out the rest of the words.
for (; w < words->size(); ++w) {
(*words)[w].word->SetupFake(unicharset);
}
return false;
}
}
if (word->word->tess_failed) {
int s;
for (s = 0; s < word->lang_words.size() &&
word->lang_words[s]->tess_failed; ++s) {}
// If all are failed, skip it. Image words are skipped by this test.
if (s > word->lang_words.size()) continue;
}
// Sync pr_it with the wth WordData.
while (pr_it->word() != NULL && pr_it->word() != word->word)
pr_it->forward();
ASSERT_HOST(pr_it->word() != NULL);
bool make_next_word_fuzzy = false;
if (!AnyLSTMLang() &&
ReassignDiacritics(pass_n, pr_it, &make_next_word_fuzzy)) {
// Needs to be setup again to see the new outlines in the chopped_word.
SetupWordPassN(pass_n, word);
}
classify_word_and_language(pass_n, pr_it, word);
if (tessedit_dump_choices || debug_noise_removal) {
tprintf("Pass%d: %s [%s]\n", pass_n,
word->word->best_choice->unichar_string().string(),
word->word->best_choice->debug_string().string());
}
pr_it->forward();
if (make_next_word_fuzzy && pr_it->word() != NULL) {
pr_it->MakeCurrentWordFuzzy();
}
}
return true;
}
/**
* recog_all_words()
*
* Walk the page_res, recognizing all the words.
* If monitor is not null, it is used as a progress monitor/timeout/cancel.
* If dopasses is 0, all recognition passes are run,
* 1 just pass 1, 2 passes2 and higher.
* If target_word_box is not null, special things are done to words that
* overlap the target_word_box:
* if word_config is not null, the word config file is read for just the
* target word(s), otherwise, on pass 2 and beyond ONLY the target words
* are processed (Jetsoft modification.)
* Returns false if we cancelled prematurely.
*
* @param page_res page structure
* @param monitor progress monitor
* @param word_config word_config file
* @param target_word_box specifies just to extract a rectangle
* @param dopasses 0 - all, 1 just pass 1, 2 passes 2 and higher
*/
bool Tesseract::recog_all_words(PAGE_RES* page_res,
ETEXT_DESC* monitor,
const TBOX* target_word_box,
const char* word_config,
int dopasses) {
PAGE_RES_IT page_res_it(page_res);
if (tessedit_minimal_rej_pass1) {
tessedit_test_adaption.set_value (TRUE);
tessedit_minimal_rejection.set_value (TRUE);
}
if (dopasses==0 || dopasses==1) {
page_res_it.restart_page();
// ****************** Pass 1 *******************
// If the adaptive classifier is full switch to one we prepared earlier,
// ie on the previous page. If the current adaptive classifier is non-empty,
// prepare a backup starting at this page, in case it fills up. Do all this
// independently for each language.
if (AdaptiveClassifierIsFull()) {
SwitchAdaptiveClassifier();
} else if (!AdaptiveClassifierIsEmpty()) {
StartBackupAdaptiveClassifier();
}
// Now check the sub-langs as well.
for (int i = 0; i < sub_langs_.size(); ++i) {
if (sub_langs_[i]->AdaptiveClassifierIsFull()) {
sub_langs_[i]->SwitchAdaptiveClassifier();
} else if (!sub_langs_[i]->AdaptiveClassifierIsEmpty()) {
sub_langs_[i]->StartBackupAdaptiveClassifier();
}
}
// Set up all words ready for recognition, so that if parallelism is on
// all the input and output classes are ready to run the classifier.
GenericVector<WordData> words;
SetupAllWordsPassN(1, target_word_box, word_config, page_res, &words);
if (tessedit_parallelize) {
PrerecAllWordsPar(words);
}
stats_.word_count = words.size();
stats_.dict_words = 0;
stats_.doc_blob_quality = 0;
stats_.doc_outline_errs = 0;
stats_.doc_char_quality = 0;
stats_.good_char_count = 0;
stats_.doc_good_char_quality = 0;
most_recently_used_ = this;
// Run pass 1 word recognition.
if (!RecogAllWordsPassN(1, monitor, &page_res_it, &words)) return false;
// Pass 1 post-processing.
for (page_res_it.restart_page(); page_res_it.word() != NULL;
page_res_it.forward()) {
if (page_res_it.word()->word->flag(W_REP_CHAR)) {
fix_rep_char(&page_res_it);
continue;
}
// Count dict words.
if (page_res_it.word()->best_choice->permuter() == USER_DAWG_PERM)
++(stats_.dict_words);
// Update misadaption log (we only need to do it on pass 1, since
// adaption only happens on this pass).
if (page_res_it.word()->blamer_bundle != NULL &&
page_res_it.word()->blamer_bundle->misadaption_debug().length() > 0) {
page_res->misadaption_log.push_back(
page_res_it.word()->blamer_bundle->misadaption_debug());
}
}
}
if (dopasses == 1) return true;
// ****************** Pass 2 *******************
if (tessedit_tess_adaption_mode != 0x0 && !tessedit_test_adaption &&
AnyTessLang()) {
page_res_it.restart_page();
GenericVector<WordData> words;
SetupAllWordsPassN(2, target_word_box, word_config, page_res, &words);
if (tessedit_parallelize) {
PrerecAllWordsPar(words);
}
most_recently_used_ = this;
// Run pass 2 word recognition.
if (!RecogAllWordsPassN(2, monitor, &page_res_it, &words)) return false;
}
// The next passes are only required for Tess-only.
if (AnyTessLang() && !AnyLSTMLang()) {
// ****************** Pass 3 *******************
// Fix fuzzy spaces.
set_global_loc_code(LOC_FUZZY_SPACE);
if (!tessedit_test_adaption && tessedit_fix_fuzzy_spaces
&& !tessedit_word_for_word && !right_to_left())
fix_fuzzy_spaces(monitor, stats_.word_count, page_res);
// ****************** Pass 4 *******************
if (tessedit_enable_dict_correction) dictionary_correction_pass(page_res);
if (tessedit_enable_bigram_correction) bigram_correction_pass(page_res);
// ****************** Pass 5,6 *******************
rejection_passes(page_res, monitor, target_word_box, word_config);
// ****************** Pass 8 *******************
font_recognition_pass(page_res);
// ****************** Pass 9 *******************
// Check the correctness of the final results.
blamer_pass(page_res);
script_pos_pass(page_res);
}
// Write results pass.
set_global_loc_code(LOC_WRITE_RESULTS);
// This is now redundant, but retained commented so show how to obtain
// bounding boxes and style information.
// changed by jetsoft
// needed for dll to output memory structure
if ((dopasses == 0 || dopasses == 2) && (monitor || tessedit_write_unlv))
output_pass(page_res_it, target_word_box);
// end jetsoft
PageSegMode pageseg_mode = static_cast<PageSegMode>(
static_cast<int>(tessedit_pageseg_mode));
textord_.CleanupSingleRowResult(pageseg_mode, page_res);
// Remove empty words, as these mess up the result iterators.
for (page_res_it.restart_page(); page_res_it.word() != NULL;
page_res_it.forward()) {
WERD_RES* word = page_res_it.word();
POLY_BLOCK* pb = page_res_it.block()->block != NULL
? page_res_it.block()->block->poly_block()
: NULL;
if (word->best_choice == NULL || word->best_choice->length() == 0 ||
(word->best_choice->IsAllSpaces() && (pb == NULL || pb->IsText()))) {
page_res_it.DeleteCurrentWord();
}
}
if (monitor != NULL) {
monitor->progress = 100;
}
return true;
}
void Tesseract::bigram_correction_pass(PAGE_RES *page_res) {
PAGE_RES_IT word_it(page_res);
WERD_RES *w_prev = NULL;
WERD_RES *w = word_it.word();
while (1) {
w_prev = w;
while (word_it.forward() != NULL &&
(!word_it.word() || word_it.word()->part_of_combo)) {
// advance word_it, skipping over parts of combos
}
if (!word_it.word()) break;
w = word_it.word();
if (!w || !w_prev || w->uch_set != w_prev->uch_set) {
continue;
}
if (w_prev->word->flag(W_REP_CHAR) || w->word->flag(W_REP_CHAR)) {
if (tessedit_bigram_debug) {
tprintf("Skipping because one of the words is W_REP_CHAR\n");
}
continue;
}
// Two words sharing the same language model, excellent!
GenericVector<WERD_CHOICE *> overrides_word1;
GenericVector<WERD_CHOICE *> overrides_word2;
STRING orig_w1_str = w_prev->best_choice->unichar_string();
STRING orig_w2_str = w->best_choice->unichar_string();
WERD_CHOICE prev_best(w->uch_set);
{
int w1start, w1end;
w_prev->best_choice->GetNonSuperscriptSpan(&w1start, &w1end);
prev_best = w_prev->best_choice->shallow_copy(w1start, w1end);
}
WERD_CHOICE this_best(w->uch_set);
{
int w2start, w2end;
w->best_choice->GetNonSuperscriptSpan(&w2start, &w2end);
this_best = w->best_choice->shallow_copy(w2start, w2end);
}
if (w->tesseract->getDict().valid_bigram(prev_best, this_best)) {
if (tessedit_bigram_debug) {
tprintf("Top choice \"%s %s\" verified by bigram model.\n",
orig_w1_str.string(), orig_w2_str.string());
}
continue;
}
if (tessedit_bigram_debug > 2) {
tprintf("Examining alt choices for \"%s %s\".\n",
orig_w1_str.string(), orig_w2_str.string());
}
if (tessedit_bigram_debug > 1) {
if (!w_prev->best_choices.singleton()) {
w_prev->PrintBestChoices();
}
if (!w->best_choices.singleton()) {
w->PrintBestChoices();
}
}
float best_rating = 0.0;
int best_idx = 0;
WERD_CHOICE_IT prev_it(&w_prev->best_choices);
for (prev_it.mark_cycle_pt(); !prev_it.cycled_list(); prev_it.forward()) {
WERD_CHOICE *p1 = prev_it.data();
WERD_CHOICE strip1(w->uch_set);
{
int p1start, p1end;
p1->GetNonSuperscriptSpan(&p1start, &p1end);
strip1 = p1->shallow_copy(p1start, p1end);
}
WERD_CHOICE_IT w_it(&w->best_choices);
for (w_it.mark_cycle_pt(); !w_it.cycled_list(); w_it.forward()) {
WERD_CHOICE *p2 = w_it.data();
WERD_CHOICE strip2(w->uch_set);
{
int p2start, p2end;
p2->GetNonSuperscriptSpan(&p2start, &p2end);
strip2 = p2->shallow_copy(p2start, p2end);
}
if (w->tesseract->getDict().valid_bigram(strip1, strip2)) {
overrides_word1.push_back(p1);
overrides_word2.push_back(p2);
if (overrides_word1.size() == 1 ||
p1->rating() + p2->rating() < best_rating) {
best_rating = p1->rating() + p2->rating();
best_idx = overrides_word1.size() - 1;
}
}
}
}
if (!overrides_word1.empty()) {
// Excellent, we have some bigram matches.
if (EqualIgnoringCaseAndTerminalPunct(*w_prev->best_choice,
*overrides_word1[best_idx]) &&
EqualIgnoringCaseAndTerminalPunct(*w->best_choice,
*overrides_word2[best_idx])) {
if (tessedit_bigram_debug > 1) {
tprintf("Top choice \"%s %s\" verified (sans case) by bigram "
"model.\n", orig_w1_str.string(), orig_w2_str.string());
}
continue;
}
STRING new_w1_str = overrides_word1[best_idx]->unichar_string();
STRING new_w2_str = overrides_word2[best_idx]->unichar_string();
if (new_w1_str != orig_w1_str) {
w_prev->ReplaceBestChoice(overrides_word1[best_idx]);
}
if (new_w2_str != orig_w2_str) {
w->ReplaceBestChoice(overrides_word2[best_idx]);
}
if (tessedit_bigram_debug > 0) {
STRING choices_description;
int num_bigram_choices
= overrides_word1.size() * overrides_word2.size();
if (num_bigram_choices == 1) {
choices_description = "This was the unique bigram choice.";
} else {
if (tessedit_bigram_debug > 1) {
STRING bigrams_list;
const int kMaxChoicesToPrint = 20;
for (int i = 0; i < overrides_word1.size() &&
i < kMaxChoicesToPrint; i++) {
if (i > 0) { bigrams_list += ", "; }
WERD_CHOICE *p1 = overrides_word1[i];
WERD_CHOICE *p2 = overrides_word2[i];
bigrams_list += p1->unichar_string() + " " + p2->unichar_string();
if (i == kMaxChoicesToPrint) {
bigrams_list += " ...";
}
}
choices_description = "There were many choices: {";
choices_description += bigrams_list;
choices_description += "}";
} else {
choices_description.add_str_int("There were ", num_bigram_choices);
choices_description += " compatible bigrams.";
}
}
tprintf("Replaced \"%s %s\" with \"%s %s\" with bigram model. %s\n",
orig_w1_str.string(), orig_w2_str.string(),
new_w1_str.string(), new_w2_str.string(),
choices_description.string());
}
}
}
}
void Tesseract::rejection_passes(PAGE_RES* page_res,
ETEXT_DESC* monitor,
const TBOX* target_word_box,
const char* word_config) {
PAGE_RES_IT page_res_it(page_res);
// ****************** Pass 5 *******************
// Gather statistics on rejects.
int word_index = 0;
while (!tessedit_test_adaption && page_res_it.word() != NULL) {
set_global_loc_code(LOC_MM_ADAPT);
WERD_RES* word = page_res_it.word();
word_index++;
if (monitor != NULL) {
monitor->ocr_alive = TRUE;
monitor->progress = 95 + 5 * word_index / stats_.word_count;
}
if (word->rebuild_word == NULL) {
// Word was not processed by tesseract.
page_res_it.forward();
continue;
}
check_debug_pt(word, 70);
// changed by jetsoft
// specific to its needs to extract one word when need
if (target_word_box &&
!ProcessTargetWord(word->word->bounding_box(),
*target_word_box, word_config, 4)) {
page_res_it.forward();
continue;
}
// end jetsoft
page_res_it.rej_stat_word();
int chars_in_word = word->reject_map.length();
int rejects_in_word = word->reject_map.reject_count();
int blob_quality = word_blob_quality(word, page_res_it.row()->row);
stats_.doc_blob_quality += blob_quality;
int outline_errs = word_outline_errs(word);
stats_.doc_outline_errs += outline_errs;
inT16 all_char_quality;
inT16 accepted_all_char_quality;
word_char_quality(word, page_res_it.row()->row,
&all_char_quality, &accepted_all_char_quality);
stats_.doc_char_quality += all_char_quality;
uint8_t permuter_type = word->best_choice->permuter();
if ((permuter_type == SYSTEM_DAWG_PERM) ||
(permuter_type == FREQ_DAWG_PERM) ||
(permuter_type == USER_DAWG_PERM)) {
stats_.good_char_count += chars_in_word - rejects_in_word;
stats_.doc_good_char_quality += accepted_all_char_quality;
}
check_debug_pt(word, 80);
if (tessedit_reject_bad_qual_wds &&
(blob_quality == 0) && (outline_errs >= chars_in_word))
word->reject_map.rej_word_bad_quality();
check_debug_pt(word, 90);
page_res_it.forward();
}
if (tessedit_debug_quality_metrics) {
tprintf
("QUALITY: num_chs= %d num_rejs= %d %5.3f blob_qual= %d %5.3f"
" outline_errs= %d %5.3f char_qual= %d %5.3f good_ch_qual= %d %5.3f\n",
page_res->char_count, page_res->rej_count,
page_res->rej_count / static_cast<float>(page_res->char_count),
stats_.doc_blob_quality,
stats_.doc_blob_quality / static_cast<float>(page_res->char_count),
stats_.doc_outline_errs,
stats_.doc_outline_errs / static_cast<float>(page_res->char_count),
stats_.doc_char_quality,
stats_.doc_char_quality / static_cast<float>(page_res->char_count),
stats_.doc_good_char_quality,
(stats_.good_char_count > 0) ?
(stats_.doc_good_char_quality /
static_cast<float>(stats_.good_char_count)) : 0.0);
}
BOOL8 good_quality_doc =
((page_res->rej_count / static_cast<float>(page_res->char_count)) <=
quality_rej_pc) &&
(stats_.doc_blob_quality / static_cast<float>(page_res->char_count) >=
quality_blob_pc) &&
(stats_.doc_outline_errs / static_cast<float>(page_res->char_count) <=
quality_outline_pc) &&
(stats_.doc_char_quality / static_cast<float>(page_res->char_count) >=
quality_char_pc);
// ****************** Pass 6 *******************
// Do whole document or whole block rejection pass
if (!tessedit_test_adaption) {
set_global_loc_code(LOC_DOC_BLK_REJ);
quality_based_rejection(page_res_it, good_quality_doc);
}
}
void Tesseract::blamer_pass(PAGE_RES* page_res) {
if (!wordrec_run_blamer) return;
PAGE_RES_IT page_res_it(page_res);
for (page_res_it.restart_page(); page_res_it.word() != NULL;
page_res_it.forward()) {
WERD_RES *word = page_res_it.word();
BlamerBundle::LastChanceBlame(wordrec_debug_blamer, word);
page_res->blame_reasons[word->blamer_bundle->incorrect_result_reason()]++;
}
tprintf("Blame reasons:\n");
for (int bl = 0; bl < IRR_NUM_REASONS; ++bl) {
tprintf("%s %d\n", BlamerBundle::IncorrectReasonName(
static_cast<IncorrectResultReason>(bl)),
page_res->blame_reasons[bl]);
}
if (page_res->misadaption_log.length() > 0) {
tprintf("Misadaption log:\n");
for (int i = 0; i < page_res->misadaption_log.length(); ++i) {
tprintf("%s\n", page_res->misadaption_log[i].string());
}
}
}
// Sets script positions and detects smallcaps on all output words.
void Tesseract::script_pos_pass(PAGE_RES* page_res) {
PAGE_RES_IT page_res_it(page_res);
for (page_res_it.restart_page(); page_res_it.word() != NULL;
page_res_it.forward()) {
WERD_RES* word = page_res_it.word();
if (word->word->flag(W_REP_CHAR)) {
page_res_it.forward();
continue;
}
float x_height = page_res_it.block()->block->x_height();
float word_x_height = word->x_height;
if (word_x_height < word->best_choice->min_x_height() ||
word_x_height > word->best_choice->max_x_height()) {
word_x_height = (word->best_choice->min_x_height() +
word->best_choice->max_x_height()) / 2.0f;
}
// Test for small caps. Word capheight must be close to block xheight,
// and word must contain no lower case letters, and at least one upper case.
double small_cap_xheight = x_height * kXHeightCapRatio;
double small_cap_delta = (x_height - small_cap_xheight) / 2.0;
if (word->uch_set->script_has_xheight() &&
small_cap_xheight - small_cap_delta <= word_x_height &&
word_x_height <= small_cap_xheight + small_cap_delta) {
// Scan for upper/lower.
int num_upper = 0;
int num_lower = 0;
for (int i = 0; i < word->best_choice->length(); ++i) {
if (word->uch_set->get_isupper(word->best_choice->unichar_id(i)))
++num_upper;
else if (word->uch_set->get_islower(word->best_choice->unichar_id(i)))
++num_lower;
}
if (num_upper > 0 && num_lower == 0)
word->small_caps = true;
}
word->SetScriptPositions();
}
}
// Helper finds the gap between the index word and the next.
static void WordGap(const PointerVector<WERD_RES>& words, int index, int* right,
int* next_left) {
*right = -MAX_INT32;
*next_left = MAX_INT32;
if (index < words.size()) {
*right = words[index]->word->bounding_box().right();
if (index + 1 < words.size())
*next_left = words[index + 1]->word->bounding_box().left();
}
}
// Factored helper computes the rating, certainty, badness and validity of
// the permuter of the words in [first_index, end_index).
static void EvaluateWordSpan(const PointerVector<WERD_RES>& words,
int first_index, int end_index, float* rating,
float* certainty, bool* bad,
bool* valid_permuter) {
if (end_index <= first_index) {
*bad = true;
*valid_permuter = false;
}
for (int index = first_index; index < end_index && index < words.size();
++index) {
WERD_CHOICE* choice = words[index]->best_choice;
if (choice == nullptr) {
*bad = true;
} else {
*rating += choice->rating();
*certainty = MIN(*certainty, choice->certainty());
if (!Dict::valid_word_permuter(choice->permuter(), false))
*valid_permuter = false;
}
}
}
// Helper chooses the best combination of words, transferring good ones from
// new_words to best_words. To win, a new word must have (better rating and
// certainty) or (better permuter status and rating within rating ratio and
// certainty within certainty margin) than current best.
// All the new_words are consumed (moved to best_words or deleted.)
// The return value is the number of new_words used minus the number of
// best_words that remain in the output.
static int SelectBestWords(double rating_ratio,
double certainty_margin,
bool debug,
PointerVector<WERD_RES>* new_words,
PointerVector<WERD_RES>* best_words) {
// Process the smallest groups of words that have an overlapping word
// boundary at the end.
GenericVector<WERD_RES*> out_words;
// Index into each word vector (best, new).
int b = 0, n = 0;
int num_best = 0, num_new = 0;
while (b < best_words->size() || n < new_words->size()) {
// Start of the current run in each.
int start_b = b, start_n = n;
while (b < best_words->size() || n < new_words->size()) {
int b_right = -MAX_INT32;
int next_b_left = MAX_INT32;
WordGap(*best_words, b, &b_right, &next_b_left);
int n_right = -MAX_INT32;
int next_n_left = MAX_INT32;
WordGap(*new_words, n, &n_right, &next_n_left);
if (MAX(b_right, n_right) < MIN(next_b_left, next_n_left)) {
// The word breaks overlap. [start_b,b] and [start_n, n] match.
break;
}
// Keep searching for the matching word break.
if ((b_right < n_right && b < best_words->size()) ||
n == new_words->size())
++b;
else
++n;
}
// Rating of the current run in each.
float b_rating = 0.0f, n_rating = 0.0f;
// Certainty of the current run in each.
float b_certainty = 0.0f, n_certainty = 0.0f;
// True if any word is missing its best choice.
bool b_bad = false, n_bad = false;
// True if all words have a valid permuter.
bool b_valid_permuter = true, n_valid_permuter = true;
int end_b = b < best_words->size() ? b + 1 : b;
int end_n = n < new_words->size() ? n + 1 : n;
EvaluateWordSpan(*best_words, start_b, end_b, &b_rating, &b_certainty,
&b_bad, &b_valid_permuter);
EvaluateWordSpan(*new_words, start_n, end_n, &n_rating, &n_certainty,
&n_bad, &n_valid_permuter);
bool new_better = false;
if (!n_bad && (b_bad || (n_certainty > b_certainty &&
n_rating < b_rating) ||
(!b_valid_permuter && n_valid_permuter &&
n_rating < b_rating * rating_ratio &&
n_certainty > b_certainty - certainty_margin))) {
// New is better.
for (int i = start_n; i < end_n; ++i) {
out_words.push_back((*new_words)[i]);
(*new_words)[i] = NULL;
++num_new;
}
new_better = true;
} else if (!b_bad) {
// Current best is better.
for (int i = start_b; i < end_b; ++i) {
out_words.push_back((*best_words)[i]);
(*best_words)[i] = NULL;
++num_best;
}
}
if (debug) {
tprintf("%d new words %s than %d old words: r: %g v %g c: %g v %g"
" valid dict: %d v %d\n",
end_n - start_n, new_better ? "better" : "worse",
end_b - start_b, n_rating, b_rating,
n_certainty, b_certainty, n_valid_permuter, b_valid_permuter);
}
// Move on to the next group.
b = end_b;
n = end_n;
}
// Transfer from out_words to best_words.
best_words->clear();
for (int i = 0; i < out_words.size(); ++i)
best_words->push_back(out_words[i]);
return num_new - num_best;
}
// Helper to recognize the word using the given (language-specific) tesseract.
// Returns positive if this recognizer found more new best words than the
// number kept from best_words.
int Tesseract::RetryWithLanguage(const WordData& word_data,
WordRecognizer recognizer, bool debug,
WERD_RES** in_word,
PointerVector<WERD_RES>* best_words) {
if (debug) {
tprintf("Trying word using lang %s, oem %d\n",
lang.string(), static_cast<int>(tessedit_ocr_engine_mode));
}
// Run the recognizer on the word.
PointerVector<WERD_RES> new_words;
(this->*recognizer)(word_data, in_word, &new_words);
if (new_words.empty()) {
// Transfer input word to new_words, as the classifier must have put
// the result back in the input.
new_words.push_back(*in_word);
*in_word = NULL;
}
if (debug) {
for (int i = 0; i < new_words.size(); ++i)
new_words[i]->DebugTopChoice("Lang result");
}
// Initial version is a bit of a hack based on better certainty and rating
// or a dictionary vs non-dictionary word.
return SelectBestWords(classify_max_rating_ratio,
classify_max_certainty_margin,
debug, &new_words, best_words);
}
// Helper returns true if all the words are acceptable.
static bool WordsAcceptable(const PointerVector<WERD_RES>& words) {
for (int w = 0; w < words.size(); ++w) {
if (words[w]->tess_failed || !words[w]->tess_accepted) return false;
}
return true;
}
// Moves good-looking "noise"/diacritics from the reject list to the main
// blob list on the current word. Returns true if anything was done, and
// sets make_next_word_fuzzy if blob(s) were added to the end of the word.
bool Tesseract::ReassignDiacritics(int pass, PAGE_RES_IT* pr_it,
bool* make_next_word_fuzzy) {
*make_next_word_fuzzy = false;
WERD* real_word = pr_it->word()->word;
if (real_word->rej_cblob_list()->empty() ||
real_word->cblob_list()->empty() ||
real_word->rej_cblob_list()->length() > noise_maxperword)
return false;
real_word->rej_cblob_list()->sort(&C_BLOB::SortByXMiddle);
// Get the noise outlines into a vector with matching bool map.
GenericVector<C_OUTLINE*> outlines;
real_word->GetNoiseOutlines(&outlines);
GenericVector<bool> word_wanted;
GenericVector<bool> overlapped_any_blob;
GenericVector<C_BLOB*> target_blobs;
AssignDiacriticsToOverlappingBlobs(outlines, pass, real_word, pr_it,
&word_wanted, &overlapped_any_blob,
&target_blobs);
// Filter the outlines that overlapped any blob and put them into the word
// now. This simplifies the remaining task and also makes it more accurate
// as it has more completed blobs to work on.
GenericVector<bool> wanted;
GenericVector<C_BLOB*> wanted_blobs;
GenericVector<C_OUTLINE*> wanted_outlines;
int num_overlapped = 0;
int num_overlapped_used = 0;
for (int i = 0; i < overlapped_any_blob.size(); ++i) {
if (overlapped_any_blob[i]) {
++num_overlapped;
if (word_wanted[i]) ++num_overlapped_used;
wanted.push_back(word_wanted[i]);
wanted_blobs.push_back(target_blobs[i]);
wanted_outlines.push_back(outlines[i]);
outlines[i] = NULL;
}
}
real_word->AddSelectedOutlines(wanted, wanted_blobs, wanted_outlines, NULL);
AssignDiacriticsToNewBlobs(outlines, pass, real_word, pr_it, &word_wanted,
&target_blobs);
int non_overlapped = 0;
int non_overlapped_used = 0;
for (int i = 0; i < word_wanted.size(); ++i) {
if (word_wanted[i]) ++non_overlapped_used;
if (outlines[i] != NULL) ++non_overlapped_used;
}
if (debug_noise_removal) {
tprintf("Used %d/%d overlapped %d/%d non-overlaped diacritics on word:",
num_overlapped_used, num_overlapped, non_overlapped_used,
non_overlapped);
real_word->bounding_box().print();
}
// Now we have decided which outlines we want, put them into the real_word.
if (real_word->AddSelectedOutlines(word_wanted, target_blobs, outlines,
make_next_word_fuzzy)) {
pr_it->MakeCurrentWordFuzzy();
}
// TODO(rays) Parts of combos have a deep copy of the real word, and need
// to have their noise outlines moved/assigned in the same way!!
return num_overlapped_used != 0 || non_overlapped_used != 0;
}
// Attempts to put noise/diacritic outlines into the blobs that they overlap.
// Input: a set of noisy outlines that probably belong to the real_word.
// Output: word_wanted indicates which outlines are to be assigned to a blob,
// target_blobs indicates which to assign to, and overlapped_any_blob is
// true for all outlines that overlapped a blob.
void Tesseract::AssignDiacriticsToOverlappingBlobs(
const GenericVector<C_OUTLINE*>& outlines, int pass, WERD* real_word,
PAGE_RES_IT* pr_it, GenericVector<bool>* word_wanted,
GenericVector<bool>* overlapped_any_blob,
GenericVector<C_BLOB*>* target_blobs) {
GenericVector<bool> blob_wanted;
word_wanted->init_to_size(outlines.size(), false);
overlapped_any_blob->init_to_size(outlines.size(), false);