Tesseract  3.02
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
tesseractclass.cpp
Go to the documentation of this file.
1 
2 // File: tesseractclass.cpp
3 // Description: An instance of Tesseract. For thread safety, *every*
4 // global variable goes in here, directly, or indirectly.
5 // Author: Ray Smith
6 // Created: Fri Mar 07 08:17:01 PST 2008
7 //
8 // (C) Copyright 2008, Google Inc.
9 // Licensed under the Apache License, Version 2.0 (the "License");
10 // you may not use this file except in compliance with the License.
11 // You may obtain a copy of the License at
12 // http://www.apache.org/licenses/LICENSE-2.0
13 // Unless required by applicable law or agreed to in writing, software
14 // distributed under the License is distributed on an "AS IS" BASIS,
15 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
16 // See the License for the specific language governing permissions and
17 // limitations under the License.
18 //
20 
21 #include "tesseractclass.h"
22 
23 #include "allheaders.h"
24 #include "cube_reco_context.h"
25 #include "edgblob.h"
26 #include "equationdetect.h"
27 #include "globals.h"
29 
30 // Include automatically generated configuration file if running autoconf.
31 #ifdef HAVE_CONFIG_H
32 #include "config_auto.h"
33 #endif
34 
35 namespace tesseract {
36 
38  : BOOL_MEMBER(tessedit_resegment_from_boxes, false,
39  "Take segmentation and labeling from box file",
40  this->params()),
41  BOOL_MEMBER(tessedit_resegment_from_line_boxes, false,
42  "Conversion of word/line box file to char box file",
43  this->params()),
44  BOOL_MEMBER(tessedit_train_from_boxes, false,
45  "Generate training data from boxed chars", this->params()),
46  BOOL_MEMBER(tessedit_make_boxes_from_boxes, false,
47  "Generate more boxes from boxed chars", this->params()),
48  BOOL_MEMBER(tessedit_dump_pageseg_images, false,
49  "Dump intermediate images made during page segmentation",
50  this->params()),
51  // The default for pageseg_mode is the old behaviour, so as not to
52  // upset anything that relies on that.
53  INT_MEMBER(tessedit_pageseg_mode, PSM_SINGLE_BLOCK,
54  "Page seg mode: 0=osd only, 1=auto+osd, 2=auto, 3=col, 4=block,"
55  " 5=line, 6=word, 7=char"
56  " (Values from PageSegMode enum in publictypes.h)",
57  this->params()),
58  INT_INIT_MEMBER(tessedit_ocr_engine_mode, tesseract::OEM_TESSERACT_ONLY,
59  "Which OCR engine(s) to run (Tesseract, Cube, both)."
60  " Defaults to loading and running only Tesseract"
61  " (no Cube,no combiner)."
62  " Values from OcrEngineMode enum in tesseractclass.h)",
63  this->params()),
64  STRING_MEMBER(tessedit_char_blacklist, "",
65  "Blacklist of chars not to recognize", this->params()),
66  STRING_MEMBER(tessedit_char_whitelist, "",
67  "Whitelist of chars to recognize", this->params()),
68  BOOL_MEMBER(tessedit_ambigs_training, false,
69  "Perform training for ambiguities", this->params()),
70  INT_MEMBER(pageseg_devanagari_split_strategy,
71  tesseract::ShiroRekhaSplitter::NO_SPLIT,
72  "Whether to use the top-line splitting process for Devanagari "
73  "documents while performing page-segmentation.", this->params()),
74  INT_MEMBER(ocr_devanagari_split_strategy,
75  tesseract::ShiroRekhaSplitter::NO_SPLIT,
76  "Whether to use the top-line splitting process for Devanagari "
77  "documents while performing ocr.", this->params()),
78  STRING_MEMBER(tessedit_write_params_to_file, "",
79  "Write all parameters to the given file.", this->params()),
80  BOOL_MEMBER(tessedit_adapt_to_char_fragments, true,
81  "Adapt to words that contain "
82  " a character composed form fragments", this->params()),
83  BOOL_MEMBER(tessedit_adaption_debug, false, "Generate and print debug"
84  " information for adaption", this->params()),
85  INT_MEMBER(bidi_debug, 0, "Debug level for BiDi", this->params()),
86  INT_MEMBER(applybox_debug, 1, "Debug level", this->params()),
87  INT_MEMBER(applybox_page, 0,
88  "Page number to apply boxes from", this->params()),
89  STRING_MEMBER(applybox_exposure_pattern, ".exp", "Exposure value follows"
90  " this pattern in the image filename. The name of the image"
91  " files are expected to be in the form"
92  " [lang].[fontname].exp[num].tif", this->params()),
93  BOOL_MEMBER(applybox_learn_chars_and_char_frags_mode, false,
94  "Learn both character fragments (as is done in the"
95  " special low exposure mode) as well as unfragmented"
96  " characters.", this->params()),
97  BOOL_MEMBER(applybox_learn_ngrams_mode, false, "Each bounding box"
98  " is assumed to contain ngrams. Only learn the ngrams"
99  " whose outlines overlap horizontally.", this->params()),
100  BOOL_MEMBER(tessedit_display_outwords, false,
101  "Draw output words", this->params()),
102  BOOL_MEMBER(tessedit_training_tess, false,
103  "Call Tess to learn blobs", this->params()),
104  BOOL_MEMBER(tessedit_dump_choices, false,
105  "Dump char choices", this->params()),
106  BOOL_MEMBER(tessedit_fix_fuzzy_spaces, true,
107  "Try to improve fuzzy spaces", this->params()),
108  BOOL_MEMBER(tessedit_unrej_any_wd, false,
109  "Dont bother with word plausibility", this->params()),
110  BOOL_MEMBER(tessedit_fix_hyphens, true,
111  "Crunch double hyphens?", this->params()),
112  BOOL_MEMBER(tessedit_redo_xheight, true,
113  "Check/Correct x-height", this->params()),
114  BOOL_MEMBER(tessedit_enable_doc_dict, true,
115  "Add words to the document dictionary", this->params()),
116  BOOL_MEMBER(tessedit_debug_fonts, false,
117  "Output font info per char", this->params()),
118  BOOL_MEMBER(tessedit_debug_block_rejection, false,
119  "Block and Row stats", this->params()),
120  BOOL_MEMBER(tessedit_enable_bigram_correction, false,
121  "Enable correction based on the word bigram dictionary.",
122  this->params()),
123  INT_MEMBER(tessedit_bigram_debug, 0,
124  "Amount of debug output for bigram correction.",
125  this->params()),
126  INT_MEMBER(debug_x_ht_level, 0, "Reestimate debug", this->params()),
127  BOOL_MEMBER(debug_acceptable_wds, false,
128  "Dump word pass/fail chk", this->params()),
129  STRING_MEMBER(chs_leading_punct, "('`\"",
130  "Leading punctuation", this->params()),
131  STRING_MEMBER(chs_trailing_punct1, ").,;:?!",
132  "1st Trailing punctuation", this->params()),
133  STRING_MEMBER(chs_trailing_punct2, ")'`\"",
134  "2nd Trailing punctuation", this->params()),
135  double_MEMBER(quality_rej_pc, 0.08,
136  "good_quality_doc lte rejection limit", this->params()),
137  double_MEMBER(quality_blob_pc, 0.0,
138  "good_quality_doc gte good blobs limit", this->params()),
139  double_MEMBER(quality_outline_pc, 1.0,
140  "good_quality_doc lte outline error limit", this->params()),
141  double_MEMBER(quality_char_pc, 0.95,
142  "good_quality_doc gte good char limit", this->params()),
143  INT_MEMBER(quality_min_initial_alphas_reqd, 2,
144  "alphas in a good word", this->params()),
145  BOOL_MEMBER(tessedit_tess_adapt_to_rejmap, false,
146  "Use reject map to control Tesseract adaption", this->params()),
147  INT_MEMBER(tessedit_tess_adaption_mode, 0x27,
148  "Adaptation decision algorithm for tess", this->params()),
149  BOOL_MEMBER(tessedit_minimal_rej_pass1, false,
150  "Do minimal rejection on pass 1 output", this->params()),
151  BOOL_MEMBER(tessedit_test_adaption, false,
152  "Test adaption criteria", this->params()),
153  BOOL_MEMBER(tessedit_matcher_log, false,
154  "Log matcher activity", this->params()),
155  INT_MEMBER(tessedit_test_adaption_mode, 3,
156  "Adaptation decision algorithm for tess", this->params()),
157  BOOL_MEMBER(save_blob_choices, false,
158  "Save the results of the recognition step (blob_choices)"
159  " within the corresponding WERD_CHOICE", this->params()),
160  BOOL_MEMBER(test_pt, false, "Test for point", this->params()),
161  double_MEMBER(test_pt_x, 99999.99, "xcoord", this->params()),
162  double_MEMBER(test_pt_y, 99999.99, "ycoord", this->params()),
163  INT_MEMBER(paragraph_debug_level, 0, "Print paragraph debug info.",
164  this->params()),
165  INT_MEMBER(cube_debug_level, 0, "Print cube debug info.", this->params()),
166  STRING_MEMBER(outlines_odd, "%| ", "Non standard number of outlines",
167  this->params()),
168  STRING_MEMBER(outlines_2, "ij!?%\":;",
169  "Non standard number of outlines", this->params()),
170  BOOL_MEMBER(docqual_excuse_outline_errs, false,
171  "Allow outline errs in unrejection?", this->params()),
172  BOOL_MEMBER(tessedit_good_quality_unrej, true,
173  "Reduce rejection on good docs", this->params()),
174  BOOL_MEMBER(tessedit_use_reject_spaces, true,
175  "Reject spaces?", this->params()),
176  double_MEMBER(tessedit_reject_doc_percent, 65.00,
177  "%rej allowed before rej whole doc", this->params()),
178  double_MEMBER(tessedit_reject_block_percent, 45.00,
179  "%rej allowed before rej whole block", this->params()),
180  double_MEMBER(tessedit_reject_row_percent, 40.00,
181  "%rej allowed before rej whole row", this->params()),
182  double_MEMBER(tessedit_whole_wd_rej_row_percent, 70.00,
183  "Number of row rejects in whole word rejects"
184  "which prevents whole row rejection", this->params()),
185  BOOL_MEMBER(tessedit_preserve_blk_rej_perfect_wds, true,
186  "Only rej partially rejected words in block rejection",
187  this->params()),
188  BOOL_MEMBER(tessedit_preserve_row_rej_perfect_wds, true,
189  "Only rej partially rejected words in row rejection",
190  this->params()),
191  BOOL_MEMBER(tessedit_dont_blkrej_good_wds, false,
192  "Use word segmentation quality metric", this->params()),
193  BOOL_MEMBER(tessedit_dont_rowrej_good_wds, false,
194  "Use word segmentation quality metric", this->params()),
195  INT_MEMBER(tessedit_preserve_min_wd_len, 2,
196  "Only preserve wds longer than this", this->params()),
197  BOOL_MEMBER(tessedit_row_rej_good_docs, true,
198  "Apply row rejection to good docs", this->params()),
199  double_MEMBER(tessedit_good_doc_still_rowrej_wd, 1.1,
200  "rej good doc wd if more than this fraction rejected",
201  this->params()),
202  BOOL_MEMBER(tessedit_reject_bad_qual_wds, true,
203  "Reject all bad quality wds", this->params()),
204  BOOL_MEMBER(tessedit_debug_doc_rejection, false,
205  "Page stats", this->params()),
206  BOOL_MEMBER(tessedit_debug_quality_metrics, false,
207  "Output data to debug file", this->params()),
208  BOOL_MEMBER(bland_unrej, false,
209  "unrej potential with no chekcs", this->params()),
210  double_MEMBER(quality_rowrej_pc, 1.1,
211  "good_quality_doc gte good char limit", this->params()),
212  BOOL_MEMBER(unlv_tilde_crunching, true,
213  "Mark v.bad words for tilde crunch", this->params()),
214  BOOL_MEMBER(crunch_early_merge_tess_fails, true,
215  "Before word crunch?", this->params()),
216  BOOL_MEMBER(crunch_early_convert_bad_unlv_chs, false,
217  "Take out ~^ early?", this->params()),
218  double_MEMBER(crunch_terrible_rating, 80.0,
219  "crunch rating lt this", this->params()),
220  BOOL_MEMBER(crunch_terrible_garbage, true, "As it says", this->params()),
221  double_MEMBER(crunch_poor_garbage_cert, -9.0,
222  "crunch garbage cert lt this", this->params()),
223  double_MEMBER(crunch_poor_garbage_rate, 60,
224  "crunch garbage rating lt this", this->params()),
225  double_MEMBER(crunch_pot_poor_rate, 40,
226  "POTENTIAL crunch rating lt this", this->params()),
227  double_MEMBER(crunch_pot_poor_cert, -8.0,
228  "POTENTIAL crunch cert lt this", this->params()),
229  BOOL_MEMBER(crunch_pot_garbage, true,
230  "POTENTIAL crunch garbage", this->params()),
231  double_MEMBER(crunch_del_rating, 60,
232  "POTENTIAL crunch rating lt this", this->params()),
233  double_MEMBER(crunch_del_cert, -10.0,
234  "POTENTIAL crunch cert lt this", this->params()),
235  double_MEMBER(crunch_del_min_ht, 0.7,
236  "Del if word ht lt xht x this", this->params()),
237  double_MEMBER(crunch_del_max_ht, 3.0,
238  "Del if word ht gt xht x this", this->params()),
239  double_MEMBER(crunch_del_min_width, 3.0,
240  "Del if word width lt xht x this", this->params()),
241  double_MEMBER(crunch_del_high_word, 1.5,
242  "Del if word gt xht x this above bl", this->params()),
243  double_MEMBER(crunch_del_low_word, 0.5,
244  "Del if word gt xht x this below bl", this->params()),
245  double_MEMBER(crunch_small_outlines_size, 0.6,
246  "Small if lt xht x this", this->params()),
247  INT_MEMBER(crunch_rating_max, 10,
248  "For adj length in rating per ch", this->params()),
249  INT_MEMBER(crunch_pot_indicators, 1,
250  "How many potential indicators needed", this->params()),
251  BOOL_MEMBER(crunch_leave_ok_strings, true,
252  "Dont touch sensible strings", this->params()),
253  BOOL_MEMBER(crunch_accept_ok, true,
254  "Use acceptability in okstring", this->params()),
255  BOOL_MEMBER(crunch_leave_accept_strings, false,
256  "Dont pot crunch sensible strings", this->params()),
257  BOOL_MEMBER(crunch_include_numerals, false,
258  "Fiddle alpha figures", this->params()),
259  INT_MEMBER(crunch_leave_lc_strings, 4,
260  "Dont crunch words with long lower case strings",
261  this->params()),
262  INT_MEMBER(crunch_leave_uc_strings, 4,
263  "Dont crunch words with long lower case strings",
264  this->params()),
265  INT_MEMBER(crunch_long_repetitions, 3,
266  "Crunch words with long repetitions", this->params()),
267  INT_MEMBER(crunch_debug, 0, "As it says", this->params()),
268  INT_MEMBER(fixsp_non_noise_limit, 1,
269  "How many non-noise blbs either side?", this->params()),
270  double_MEMBER(fixsp_small_outlines_size, 0.28,
271  "Small if lt xht x this", this->params()),
272  BOOL_MEMBER(tessedit_prefer_joined_punct, false,
273  "Reward punctation joins", this->params()),
274  INT_MEMBER(fixsp_done_mode, 1,
275  "What constitues done for spacing", this->params()),
276  INT_MEMBER(debug_fix_space_level, 0,
277  "Contextual fixspace debug", this->params()),
278  STRING_MEMBER(numeric_punctuation, ".,",
279  "Punct. chs expected WITHIN numbers", this->params()),
280  INT_MEMBER(x_ht_acceptance_tolerance, 8,
281  "Max allowed deviation of blob top outside of font data",
282  this->params()),
283  INT_MEMBER(x_ht_min_change, 8,
284  "Min change in xht before actually trying it", this->params()),
285  BOOL_MEMBER(tessedit_write_block_separators, false,
286  "Write block separators in output", this->params()),
287  BOOL_MEMBER(tessedit_write_rep_codes, false,
288  "Write repetition char code", this->params()),
289  BOOL_MEMBER(tessedit_write_unlv, false,
290  "Write .unlv output file", this->params()),
291  BOOL_MEMBER(tessedit_create_hocr, false,
292  "Write .html hOCR output file", this->params()),
293  STRING_MEMBER(unrecognised_char, "|",
294  "Output char for unidentified blobs", this->params()),
295  INT_MEMBER(suspect_level, 99, "Suspect marker level", this->params()),
296  INT_MEMBER(suspect_space_level, 100,
297  "Min suspect level for rejecting spaces", this->params()),
298  INT_MEMBER(suspect_short_words, 2,
299  "Dont Suspect dict wds longer than this", this->params()),
300  BOOL_MEMBER(suspect_constrain_1Il, false,
301  "UNLV keep 1Il chars rejected", this->params()),
302  double_MEMBER(suspect_rating_per_ch, 999.9,
303  "Dont touch bad rating limit", this->params()),
304  double_MEMBER(suspect_accept_rating, -999.9,
305  "Accept good rating limit", this->params()),
306  BOOL_MEMBER(tessedit_minimal_rejection, false,
307  "Only reject tess failures", this->params()),
308  BOOL_MEMBER(tessedit_zero_rejection, false,
309  "Dont reject ANYTHING", this->params()),
310  BOOL_MEMBER(tessedit_word_for_word, false,
311  "Make output have exactly one word per WERD", this->params()),
312  BOOL_MEMBER(tessedit_zero_kelvin_rejection, false,
313  "Dont reject ANYTHING AT ALL", this->params()),
314  BOOL_MEMBER(tessedit_consistent_reps, true,
315  "Force all rep chars the same", this->params()),
316  INT_MEMBER(tessedit_reject_mode, 0, "Rejection algorithm", this->params()),
317  INT_MEMBER(tessedit_ok_mode, 5,
318  "Acceptance decision algorithm", this->params()),
319  BOOL_MEMBER(tessedit_rejection_debug, false,
320  "Adaption debug", this->params()),
321  BOOL_MEMBER(tessedit_flip_0O, true,
322  "Contextual 0O O0 flips", this->params()),
323  double_MEMBER(tessedit_lower_flip_hyphen, 1.5,
324  "Aspect ratio dot/hyphen test", this->params()),
325  double_MEMBER(tessedit_upper_flip_hyphen, 1.8,
326  "Aspect ratio dot/hyphen test", this->params()),
327  BOOL_MEMBER(rej_trust_doc_dawg, false,
328  "Use DOC dawg in 11l conf. detector", this->params()),
329  BOOL_MEMBER(rej_1Il_use_dict_word, false,
330  "Use dictword test", this->params()),
331  BOOL_MEMBER(rej_1Il_trust_permuter_type, true,
332  "Dont double check", this->params()),
333  BOOL_MEMBER(rej_use_tess_accepted, true,
334  "Individual rejection control", this->params()),
335  BOOL_MEMBER(rej_use_tess_blanks, true,
336  "Individual rejection control", this->params()),
337  BOOL_MEMBER(rej_use_good_perm, true,
338  "Individual rejection control", this->params()),
339  BOOL_MEMBER(rej_use_sensible_wd, false,
340  "Extend permuter check", this->params()),
341  BOOL_MEMBER(rej_alphas_in_number_perm, false,
342  "Extend permuter check", this->params()),
343  double_MEMBER(rej_whole_of_mostly_reject_word_fract, 0.85,
344  "if >this fract", this->params()),
345  INT_MEMBER(tessedit_image_border, 2,
346  "Rej blbs near image edge limit", this->params()),
347  STRING_MEMBER(ok_repeated_ch_non_alphanum_wds, "-?*\075",
348  "Allow NN to unrej", this->params()),
349  STRING_MEMBER(conflict_set_I_l_1, "Il1[]",
350  "Il1 conflict set", this->params()),
351  INT_MEMBER(min_sane_x_ht_pixels, 8,
352  "Reject any x-ht lt or eq than this", this->params()),
353  BOOL_MEMBER(tessedit_create_boxfile, false,
354  "Output text with boxes", this->params()),
355  INT_MEMBER(tessedit_page_number, -1, "-1 -> All pages"
356  " , else specifc page to process", this->params()),
357  BOOL_MEMBER(tessedit_write_images, false,
358  "Capture the image from the IPE", this->params()),
359  BOOL_MEMBER(interactive_display_mode, false, "Run interactively?",
360  this->params()),
361  STRING_MEMBER(file_type, ".tif", "Filename extension", this->params()),
362  BOOL_MEMBER(tessedit_override_permuter, true,
363  "According to dict_word", this->params()),
364  INT_MEMBER(tessdata_manager_debug_level, 0, "Debug level for"
365  " TessdataManager functions.", this->params()),
366  STRING_MEMBER(tessedit_load_sublangs, "",
367  "List of languages to load with this one", this->params()),
368  double_MEMBER(min_orientation_margin, 7.0,
369  "Min acceptable orientation margin", this->params()),
370  BOOL_MEMBER(textord_tabfind_show_vlines, false, "Debug line finding",
371  this->params()),
372  BOOL_MEMBER(textord_use_cjk_fp_model, FALSE, "Use CJK fixed pitch model",
373  this->params()),
374  BOOL_INIT_MEMBER(tessedit_init_config_only, false,
375  "Only initialize with the config file. Useful if the "
376  "instance is not going to be used for OCR but say only "
377  "for layout analysis.", this->params()),
378  BOOL_MEMBER(textord_equation_detect, false, "Turn on equation detector",
379  this->params()),
380  backup_config_file_(NULL),
381  pix_binary_(NULL),
382  cube_binary_(NULL),
383  pix_grey_(NULL),
384  source_resolution_(0),
385  textord_(this),
386  right_to_left_(false),
387  scaled_color_(NULL),
388  scaled_factor_(-1),
389  deskew_(1.0f, 0.0f),
390  reskew_(1.0f, 0.0f),
391  most_recently_used_(this),
392  font_table_size_(0),
393  cube_cntxt_(NULL),
394  tess_cube_combiner_(NULL),
395  equ_detect_(NULL) {
396 }
397 
399  Clear();
400  end_tesseract();
401  sub_langs_.delete_data_pointers();
402  // Delete cube objects.
403  if (cube_cntxt_ != NULL) {
404  delete cube_cntxt_;
405  cube_cntxt_ = NULL;
406  }
407  if (tess_cube_combiner_ != NULL) {
408  delete tess_cube_combiner_;
409  tess_cube_combiner_ = NULL;
410  }
411 }
412 
414  pixDestroy(&pix_binary_);
415  pixDestroy(&cube_binary_);
416  pixDestroy(&pix_grey_);
417  pixDestroy(&scaled_color_);
418  deskew_ = FCOORD(1.0f, 0.0f);
419  reskew_ = FCOORD(1.0f, 0.0f);
420  splitter_.Clear();
421  scaled_factor_ = -1;
423  for (int i = 0; i < sub_langs_.size(); ++i)
424  sub_langs_[i]->Clear();
425 }
426 
428  equ_detect_ = detector;
429  equ_detect_->SetLangTesseract(this);
430 }
431 
432 // Clear all memory of adaption for this and all subclassifiers.
435  for (int i = 0; i < sub_langs_.size(); ++i) {
436  sub_langs_[i]->ResetAdaptiveClassifierInternal();
437  }
438 }
439 
440 // Clear the document dictionary for this and all subclassifiers.
443  for (int i = 0; i < sub_langs_.size(); ++i) {
444  sub_langs_[i]->getDict().ResetDocumentDictionary();
445  }
446 }
447 
449  // Set the white and blacklists (if any)
451  tessedit_char_whitelist.string());
452  // Black and white lists should apply to all loaded classifiers.
453  for (int i = 0; i < sub_langs_.size(); ++i) {
454  sub_langs_[i]->unicharset.set_black_and_whitelist(
456  }
457 }
458 
459 // Perform steps to prepare underlying binary image/other data structures for
460 // page segmentation.
463  pixDestroy(&cube_binary_);
464  cube_binary_ = pixClone(pix_binary());
465  // Find the max splitter strategy over all langs.
466  ShiroRekhaSplitter::SplitStrategy max_pageseg_strategy =
469  for (int i = 0; i < sub_langs_.size(); ++i) {
470  ShiroRekhaSplitter::SplitStrategy pageseg_strategy =
472  static_cast<inT32>(sub_langs_[i]->pageseg_devanagari_split_strategy));
473  if (pageseg_strategy > max_pageseg_strategy)
474  max_pageseg_strategy = pageseg_strategy;
475  // Clone the cube image to all the sub langs too.
476  pixDestroy(&sub_langs_[i]->cube_binary_);
477  sub_langs_[i]->cube_binary_ = pixClone(pix_binary());
478  pixDestroy(&sub_langs_[i]->pix_binary_);
479  sub_langs_[i]->pix_binary_ = pixClone(pix_binary());
480  }
481  // Perform shiro-rekha (top-line) splitting and replace the current image by
482  // the newly splitted image.
483  splitter_.set_orig_pix(pix_binary());
484  splitter_.set_pageseg_split_strategy(max_pageseg_strategy);
485  if (splitter_.Split(true)) {
486  ASSERT_HOST(splitter_.splitted_image());
487  pixDestroy(&pix_binary_);
488  pix_binary_ = pixClone(splitter_.splitted_image());
489  }
490 }
491 
492 // Perform steps to prepare underlying binary image/other data structures for
493 // OCR. The current segmentation is required by this method.
494 // Note that this method resets pix_binary_ to the original binarized image,
495 // which may be different from the image actually used for OCR depending on the
496 // value of devanagari_ocr_split_strategy.
497 void Tesseract::PrepareForTessOCR(BLOCK_LIST* block_list,
498  Tesseract* osd_tess, OSResults* osr) {
499  // Find the max splitter strategy over all langs.
500  ShiroRekhaSplitter::SplitStrategy max_ocr_strategy =
502  static_cast<inT32>(ocr_devanagari_split_strategy));
503  for (int i = 0; i < sub_langs_.size(); ++i) {
504  ShiroRekhaSplitter::SplitStrategy ocr_strategy =
506  static_cast<inT32>(sub_langs_[i]->ocr_devanagari_split_strategy));
507  if (ocr_strategy > max_ocr_strategy)
508  max_ocr_strategy = ocr_strategy;
509  }
510  // Utilize the segmentation information available.
511  splitter_.set_segmentation_block_list(block_list);
512  splitter_.set_ocr_split_strategy(max_ocr_strategy);
513  // Run the splitter for OCR
514  bool split_for_ocr = splitter_.Split(false);
515  // Restore pix_binary to the binarized original pix for future reference.
516  ASSERT_HOST(splitter_.orig_pix());
517  pixDestroy(&pix_binary_);
518  pix_binary_ = pixClone(splitter_.orig_pix());
519  // If the pageseg and ocr strategies are different, refresh the block list
520  // (from the last SegmentImage call) with blobs from the real image to be used
521  // for OCR.
522  if (splitter_.HasDifferentSplitStrategies()) {
523  BLOCK block("", TRUE, 0, 0, 0, 0, pixGetWidth(pix_binary_),
524  pixGetHeight(pix_binary_));
525  Pix* pix_for_ocr = split_for_ocr ? splitter_.splitted_image() :
526  splitter_.orig_pix();
527  extract_edges(pix_for_ocr, &block);
528  splitter_.RefreshSegmentationWithNewBlobs(block.blob_list());
529  }
530  // The splitter isn't needed any more after this, so save memory by clearing.
531  splitter_.Clear();
532 }
533 
534 } // namespace tesseract