Leptonica 1.85.0
Image processing and image analysis suite
Loading...
Searching...
No Matches
recogident.c
Go to the documentation of this file.
1/*====================================================================*
2 - Copyright (C) 2001 Leptonica. All rights reserved.
3 -
4 - Redistribution and use in source and binary forms, with or without
5 - modification, are permitted provided that the following conditions
6 - are met:
7 - 1. Redistributions of source code must retain the above copyright
8 - notice, this list of conditions and the following disclaimer.
9 - 2. Redistributions in binary form must reproduce the above
10 - copyright notice, this list of conditions and the following
11 - disclaimer in the documentation and/or other materials
12 - provided with the distribution.
13 -
14 - THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
15 - ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
16 - LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
17 - A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL ANY
18 - CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
19 - EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
20 - PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
21 - PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
22 - OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
23 - NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
24 - SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 *====================================================================*/
26
92#ifdef HAVE_CONFIG_H
93#include <config_auto.h>
94#endif /* HAVE_CONFIG_H */
95
96#include <string.h>
97#include "allheaders.h"
98#include "array_internal.h"
99
100 /* There are two methods for splitting characters: DID and greedy.
101 * The default method is DID. */
102#define SPLIT_WITH_DID 1
103
104 /* Padding on pix1: added before correlations and removed from result */
105static const l_int32 LeftRightPadding = 32;
106
107 /* Parameters for filtering and sorting connected components in splitter */
108static const l_float32 MinFillFactor = 0.10f;
109static const l_int32 DefaultMinHeight = 15; /* min unscaled height */
110static const l_int32 MinOverlap1 = 6; /* in pass 1 of boxaSort2d() */
111static const l_int32 MinOverlap2 = 6; /* in pass 2 of boxaSort2d() */
112static const l_int32 MinHeightPass1 = 5; /* min height to start pass 1 */
113
114
115static l_int32 pixCorrelationBestShift(PIX *pix1, PIX *pix2, NUMA *nasum1,
116 NUMA *namoment1, l_int32 area2,
117 l_int32 ycent2, l_int32 maxyshift,
118 l_int32 *tab8, l_int32 *pdelx,
119 l_int32 *pdely, l_float32 *pscore,
120 l_int32 debugflag );
121static L_RCH *rchCreate(l_int32 index, l_float32 score, char *text,
122 l_int32 sample, l_int32 xloc, l_int32 yloc,
123 l_int32 width);
124static L_RCHA *rchaCreate();
125static l_int32 transferRchToRcha(L_RCH *rch, L_RCHA *rcha);
126static PIX *recogPreSplittingFilter(L_RECOG *recog, PIX *pixs, l_int32 minh,
127 l_float32 minaf, l_int32 debug);
128static l_int32 recogSplittingFilter(L_RECOG *recog, PIX *pixs, l_int32 min,
129 l_float32 minaf, l_int32 *premove,
130 l_int32 debug);
131static void l_showIndicatorSplitValues(NUMA *na1, NUMA *na2, NUMA *na3,
132 NUMA *na4, NUMA *na5, NUMA *na6);
133
134/*------------------------------------------------------------------------*
135 * Identification
136 *------------------------------------------------------------------------*/
162l_ok
163recogIdentifyMultiple(L_RECOG *recog,
164 PIX *pixs,
165 l_int32 minh,
166 l_int32 skipsplit,
167 BOXA **pboxa,
168 PIXA **ppixa,
169 PIX **ppixdb,
170 l_int32 debugsplit)
171{
172l_int32 n;
173BOXA *boxa;
174PIX *pixb;
175PIXA *pixa;
176
177 if (pboxa) *pboxa = NULL;
178 if (ppixa) *ppixa = NULL;
179 if (ppixdb) *ppixdb = NULL;
180 if (!recog)
181 return ERROR_INT("recog not defined", __func__, 2);
182 if (!recog->train_done)
183 return ERROR_INT("training not finished", __func__, 2);
184 if (!pixs)
185 return ERROR_INT("pixs not defined", __func__, 2);
186
187 /* Binarize if necessary */
188 if (pixGetDepth(pixs) > 1)
189 pixb = pixConvertTo1(pixs, recog->threshold);
190 else
191 pixb = pixClone(pixs);
192
193 /* Noise removal and splitting of touching characters */
194 recogSplitIntoCharacters(recog, pixb, minh, skipsplit, &boxa, &pixa,
195 debugsplit);
196 pixDestroy(&pixb);
197 if (!pixa || (n = pixaGetCount(pixa)) == 0) {
198 pixaDestroy(&pixa);
199 boxaDestroy(&boxa);
200 L_WARNING("nothing found\n", __func__);
201 return 1;
202 }
203
204 recogIdentifyPixa(recog, pixa, ppixdb);
205 if (pboxa)
206 *pboxa = boxa;
207 else
208 boxaDestroy(&boxa);
209 if (ppixa)
210 *ppixa = pixa;
211 else
212 pixaDestroy(&pixa);
213 return 0;
214}
215
216
217/*------------------------------------------------------------------------*
218 * Segmentation and noise removal *
219 *------------------------------------------------------------------------*/
248l_ok
249recogSplitIntoCharacters(L_RECOG *recog,
250 PIX *pixs,
251 l_int32 minh,
252 l_int32 skipsplit,
253 BOXA **pboxa,
254 PIXA **ppixa,
255 l_int32 debug)
256{
257static l_int32 ind = 0;
258char buf[32];
259l_int32 i, xoff, yoff, empty, maxw, bw, ncomp, scaling;
260BOX *box;
261BOXA *boxa1, *boxa2, *boxa3, *boxa4, *boxad;
262BOXAA *baa;
263PIX *pix, *pix1, *pix2, *pix3;
264PIXA *pixa;
265
266 lept_mkdir("lept/recog");
267
268 if (pboxa) *pboxa = NULL;
269 if (ppixa) *ppixa = NULL;
270 if (!pboxa || !ppixa)
271 return ERROR_INT("&boxa and &pixa not defined", __func__, 1);
272 if (!recog)
273 return ERROR_INT("recog not defined", __func__, 1);
274 if (!recog->train_done)
275 return ERROR_INT("training not finished", __func__, 1);
276 if (!pixs || pixGetDepth(pixs) != 1)
277 return ERROR_INT("pixs not defined or not 1 bpp", __func__, 1);
278 if (minh <= 0) minh = DefaultMinHeight;
279 pixZero(pixs, &empty);
280 if (empty) return 1;
281
282 /* Small vertical close for consolidation. Don't do a horizontal
283 * closing, because it might join separate characters. */
284 pix1 = pixMorphSequence(pixs, "c1.3", 0);
285
286 /* Carefully filter out noise */
287 pix2 = recogPreSplittingFilter(recog, pix1, minh, MinFillFactor, debug);
288 pixDestroy(&pix1);
289
290 /* Get the 8-connected components to be split/identified */
291 boxa1 = pixConnComp(pix2, NULL, 8);
292 pixDestroy(&pix2);
293 ncomp = boxaGetCount(boxa1);
294 if (ncomp == 0) {
295 boxaDestroy(&boxa1);
296 L_WARNING("all components removed\n", __func__);
297 return 1;
298 }
299
300 /* Save everything and split the large components */
301 boxa2 = boxaCreate(ncomp);
302 maxw = recog->maxwidth_u + 5;
303 scaling = (recog->scalew > 0 || recog->scaleh > 0) ? TRUE : FALSE;
304 pixa = (debug) ? pixaCreate(ncomp) : NULL;
305 for (i = 0; i < ncomp; i++) {
306 box = boxaGetBox(boxa1, i, L_CLONE);
307 boxGetGeometry(box, &xoff, &yoff, &bw, NULL);
308 /* Treat as one character if it is small, if the images
309 * have been scaled, or if splitting is not to be run. */
310 if (bw <= maxw || scaling || skipsplit) {
311 boxaAddBox(boxa2, box, L_INSERT);
312 } else {
313 pix = pixClipRectangle(pixs, box, NULL);
314#if SPLIT_WITH_DID
315 if (!debug) {
316 boxa3 = recogDecode(recog, pix, 2, NULL);
317 } else {
318 boxa3 = recogDecode(recog, pix, 2, &pix2);
319 pixaAddPix(pixa, pix2, L_INSERT);
320 }
321#else /* use greedy splitting */
322 recogCorrelationBestRow(recog, pix, &boxa3, NULL, NULL,
323 NULL, debug);
324 if (debug) {
325 pix2 = pixConvertTo32(pix);
326 pixRenderBoxaArb(pix2, boxa3, 2, 255, 0, 0);
327 pixaAddPix(pixa, pix2, L_INSERT);
328 }
329#endif /* SPLIT_WITH_DID */
330 pixDestroy(&pix);
331 boxDestroy(&box);
332 if (!boxa3) {
333 L_ERROR("boxa3 not found for component %d\n", __func__, i);
334 } else {
335 boxa4 = boxaTransform(boxa3, xoff, yoff, 1.0, 1.0);
336 boxaJoin(boxa2, boxa4, 0, -1);
337 boxaDestroy(&boxa3);
338 boxaDestroy(&boxa4);
339 }
340 }
341 }
342 boxaDestroy(&boxa1);
343 if (pixa) { /* debug */
344 pix3 = pixaDisplayTiledInColumns(pixa, 1, 1.0, 20, 2);
345 snprintf(buf, sizeof(buf), "/tmp/lept/recog/decode-%d.png", ind++);
346 pixWrite(buf, pix3, IFF_PNG);
347 pixaDestroy(&pixa);
348 pixDestroy(&pix3);
349 }
350
351 /* Do a 2D sort on the bounding boxes, and flatten the result to 1D.
352 * For the 2D sort, to add a box to an existing boxa, we require
353 * specified minimum vertical overlaps for the first two passes
354 * of the 2D sort. In pass 1, only components with sufficient
355 * height can start a new boxa. */
356 baa = boxaSort2d(boxa2, NULL, MinOverlap1, MinOverlap2, MinHeightPass1);
357 boxa3 = boxaaFlattenToBoxa(baa, NULL, L_CLONE);
358 boxaaDestroy(&baa);
359 boxaDestroy(&boxa2);
360
361 /* Remove smaller components of overlapping pairs.
362 * We only remove the small component if the overlap is
363 * at least half its area and if its area is no more
364 * than 30% of the area of the large component. Because the
365 * components are in a flattened 2D sort, we don't need to
366 * look far ahead in the array to find all overlapping boxes;
367 * 10 boxes is plenty. */
368 boxad = boxaHandleOverlaps(boxa3, L_COMBINE, 10, 0.5f, 0.3f, NULL);
369 boxaDestroy(&boxa3);
370
371 /* Extract and save the image pieces from the input image. */
372 *ppixa = pixClipRectangles(pixs, boxad);
373 *pboxa = boxad;
374 return 0;
375}
376
377
378/*------------------------------------------------------------------------*
379 * Greedy character splitting *
380 *------------------------------------------------------------------------*/
401l_ok
402recogCorrelationBestRow(L_RECOG *recog,
403 PIX *pixs,
404 BOXA **pboxa,
405 NUMA **pnascore,
406 NUMA **pnaindex,
407 SARRAY **psachar,
408 l_int32 debug)
409{
410char *charstr;
411l_int32 index, remove, w, h, bx, bw, bxc, bwc, w1, w2, w3;
412l_float32 score;
413BOX *box, *boxc, *boxtrans, *boxl, *boxr, *boxlt, *boxrt;
414BOXA *boxat;
415NUMA *nascoret, *naindext, *nasort;
416PIX *pixb, *pixc, *pixl, *pixr, *pixdb, *pixd;
417PIXA *pixar, *pixadb;
418SARRAY *sachart;
419
420l_int32 iter;
421
422 if (pnascore) *pnascore = NULL;
423 if (pnaindex) *pnaindex = NULL;
424 if (psachar) *psachar = NULL;
425 if (!pboxa)
426 return ERROR_INT("&boxa not defined", __func__, 1);
427 *pboxa = NULL;
428 if (!recog)
429 return ERROR_INT("recog not defined", __func__, 1);
430 if (!pixs || pixGetDepth(pixs) != 1)
431 return ERROR_INT("pixs not defined or not 1 bpp", __func__, 1);
432 if (pixGetWidth(pixs) < recog->minwidth_u - 4)
433 return ERROR_INT("pixs too narrow", __func__, 1);
434 if (!recog->train_done)
435 return ERROR_INT("training not finished", __func__, 1);
436
437 /* Binarize and crop to foreground if necessary */
438 pixb = recogProcessToIdentify(recog, pixs, 0);
439
440 /* Initialize the arrays */
441 boxat = boxaCreate(4);
442 nascoret = numaCreate(4);
443 naindext = numaCreate(4);
444 sachart = sarrayCreate(4);
445 pixadb = (debug) ? pixaCreate(4) : NULL;
446
447 /* Initialize the images remaining to be processed with the input.
448 * These are stored in pixar, which is used here as a queue,
449 * on which we only put image fragments that are large enough to
450 * contain at least one character. */
451 pixar = pixaCreate(1);
452 pixGetDimensions(pixb, &w, &h, NULL);
453 box = boxCreate(0, 0, w, h);
454 pixaAddPix(pixar, pixb, L_INSERT);
455 pixaAddBox(pixar, box, L_INSERT);
456
457 /* Successively split on the best match until nothing is left.
458 * To be safe, we limit the search to 10 characters. */
459 for (iter = 0; iter < 11; iter++) {
460 if (pixaGetCount(pixar) == 0)
461 break;
462 if (iter == 10) {
463 L_WARNING("more than 10 chars; ending search\n", __func__);
464 break;
465 }
466
467 /* Pop one from the queue */
468 pixaRemovePixAndSave(pixar, 0, &pixc, &boxc);
469 boxGetGeometry(boxc, &bxc, NULL, &bwc, NULL);
470
471 /* This is a single component; if noise, remove it */
472 recogSplittingFilter(recog, pixc, 0, MinFillFactor, &remove, debug);
473 if (debug)
474 lept_stderr("iter = %d, removed = %d\n", iter, remove);
475 if (remove) {
476 pixDestroy(&pixc);
477 boxDestroy(&boxc);
478 continue;
479 }
480
481 /* Find the best character match */
482 if (debug) {
483 recogCorrelationBestChar(recog, pixc, &box, &score,
484 &index, &charstr, &pixdb);
485 pixaAddPix(pixadb, pixdb, L_INSERT);
486 } else {
487 recogCorrelationBestChar(recog, pixc, &box, &score,
488 &index, &charstr, NULL);
489 }
490
491 /* Find the box in original coordinates, and append
492 * the results to the arrays. */
493 boxtrans = boxTransform(box, bxc, 0, 1.0, 1.0);
494 boxaAddBox(boxat, boxtrans, L_INSERT);
495 numaAddNumber(nascoret, score);
496 numaAddNumber(naindext, index);
497 sarrayAddString(sachart, charstr, L_INSERT);
498
499 /* Split the current pixc into three regions and save
500 * each region if it is large enough. */
501 boxGetGeometry(box, &bx, NULL, &bw, NULL);
502 w1 = bx;
503 w2 = bw;
504 w3 = bwc - bx - bw;
505 if (debug)
506 lept_stderr(" w1 = %d, w2 = %d, w3 = %d\n", w1, w2, w3);
507 if (w1 < recog->minwidth_u - 4) {
508 if (debug) L_INFO("discarding width %d on left\n", __func__, w1);
509 } else { /* extract and save left region */
510 boxl = boxCreate(0, 0, bx + 1, h);
511 pixl = pixClipRectangle(pixc, boxl, NULL);
512 boxlt = boxTransform(boxl, bxc, 0, 1.0, 1.0);
513 pixaAddPix(pixar, pixl, L_INSERT);
514 pixaAddBox(pixar, boxlt, L_INSERT);
515 boxDestroy(&boxl);
516 }
517 if (w3 < recog->minwidth_u - 4) {
518 if (debug) L_INFO("discarding width %d on right\n", __func__, w3);
519 } else { /* extract and save left region */
520 boxr = boxCreate(bx + bw - 1, 0, w3 + 1, h);
521 pixr = pixClipRectangle(pixc, boxr, NULL);
522 boxrt = boxTransform(boxr, bxc, 0, 1.0, 1.0);
523 pixaAddPix(pixar, pixr, L_INSERT);
524 pixaAddBox(pixar, boxrt, L_INSERT);
525 boxDestroy(&boxr);
526 }
527 pixDestroy(&pixc);
528 boxDestroy(&box);
529 boxDestroy(&boxc);
530 }
531 pixaDestroy(&pixar);
532
533
534 /* Sort the output results by left-to-right in the boxa */
535 *pboxa = boxaSort(boxat, L_SORT_BY_X, L_SORT_INCREASING, &nasort);
536 if (pnascore)
537 *pnascore = numaSortByIndex(nascoret, nasort);
538 if (pnaindex)
539 *pnaindex = numaSortByIndex(naindext, nasort);
540 if (psachar)
541 *psachar = sarraySortByIndex(sachart, nasort);
542 numaDestroy(&nasort);
543 boxaDestroy(&boxat);
544 numaDestroy(&nascoret);
545 numaDestroy(&naindext);
546 sarrayDestroy(&sachart);
547
548 /* Final debug output */
549 if (debug) {
550 pixd = pixaDisplayTiledInRows(pixadb, 32, 2000, 1.0, 0, 15, 2);
551 pixDisplay(pixd, 400, 400);
552 pixaAddPix(recog->pixadb_split, pixd, L_INSERT);
553 pixaDestroy(&pixadb);
554 }
555 return 0;
556}
557
558
581l_ok
582recogCorrelationBestChar(L_RECOG *recog,
583 PIX *pixs,
584 BOX **pbox,
585 l_float32 *pscore,
586 l_int32 *pindex,
587 char **pcharstr,
588 PIX **ppixdb)
589{
590l_int32 i, n, w1, h1, w2, area2, ycent2, delx, dely;
591l_int32 bestdelx, bestdely, bestindex;
592l_float32 score, bestscore;
593BOX *box;
594BOXA *boxa;
595NUMA *nasum, *namoment;
596PIX *pix1, *pix2;
597
598 if (pindex) *pindex = 0;
599 if (pcharstr) *pcharstr = NULL;
600 if (ppixdb) *ppixdb = NULL;
601 if (pbox) *pbox = NULL;
602 if (pscore) *pscore = 0.0;
603 if (!pbox || !pscore)
604 return ERROR_INT("&box and &score not both defined", __func__, 1);
605 if (!recog)
606 return ERROR_INT("recog not defined", __func__, 1);
607 if (!pixs || pixGetDepth(pixs) != 1)
608 return ERROR_INT("pixs not defined or not 1 bpp", __func__, 1);
609 if (!recog->train_done)
610 return ERROR_INT("training not finished", __func__, 1);
611
612 /* Binarize and crop to foreground if necessary. Add padding
613 * to both the left and right side; this is compensated for
614 * when reporting the bounding box of the best matched character. */
615 pix1 = recogProcessToIdentify(recog, pixs, LeftRightPadding);
616 pixGetDimensions(pix1, &w1, &h1, NULL);
617
618 /* Compute vertical sum and moment arrays */
619 nasum = pixCountPixelsByColumn(pix1);
620 namoment = pixGetMomentByColumn(pix1, 1);
621
622 /* Do shifted correlation against all averaged templates. */
623 n = recog->setsize;
624 boxa = boxaCreate(n); /* location of best fits for each character */
625 bestscore = 0.0;
626 bestindex = bestdelx = bestdely = 0;
627 for (i = 0; i < n; i++) {
628 pix2 = pixaGetPix(recog->pixa_u, i, L_CLONE);
629 w2 = pixGetWidth(pix2);
630 /* Note that the slightly expended w1 is typically larger
631 * than w2 (the template). */
632 if (w1 >= w2) {
633 numaGetIValue(recog->nasum_u, i, &area2);
634 ptaGetIPt(recog->pta_u, i, NULL, &ycent2);
635 pixCorrelationBestShift(pix1, pix2, nasum, namoment, area2, ycent2,
636 recog->maxyshift, recog->sumtab, &delx,
637 &dely, &score, 1);
638 if (ppixdb) {
639 lept_stderr(
640 "Best match template %d: (x,y) = (%d,%d), score = %5.3f\n",
641 i, delx, dely, score);
642 }
643 /* Compensate for padding */
644 box = boxCreate(delx - LeftRightPadding, 0, w2, h1);
645 if (score > bestscore) {
646 bestscore = score;
647 bestdelx = delx - LeftRightPadding;
648 bestdely = dely;
649 bestindex = i;
650 }
651 } else {
652 box = boxCreate(0, 0, 1, 1); /* placeholder */
653 if (ppixdb)
654 lept_stderr("Component too thin: w1 = %d, w2 = %d\n", w1, w2);
655 }
656 boxaAddBox(boxa, box, L_INSERT);
657 pixDestroy(&pix2);
658 }
659
660 *pscore = bestscore;
661 *pbox = boxaGetBox(boxa, bestindex, L_COPY);
662 if (pindex) *pindex = bestindex;
663 if (pcharstr)
664 recogGetClassString(recog, bestindex, pcharstr);
665
666 if (ppixdb) {
667 L_INFO("Best match: class %d; shifts (%d, %d)\n",
668 __func__, bestindex, bestdelx, bestdely);
669 pix2 = pixaGetPix(recog->pixa_u, bestindex, L_CLONE);
670 *ppixdb = recogShowMatch(recog, pix1, pix2, NULL, -1, 0.0);
671 pixDestroy(&pix2);
672 }
673
674 pixDestroy(&pix1);
675 boxaDestroy(&boxa);
676 numaDestroy(&nasum);
677 numaDestroy(&namoment);
678 return 0;
679}
680
681
721static l_int32
722pixCorrelationBestShift(PIX *pix1,
723 PIX *pix2,
724 NUMA *nasum1,
725 NUMA *namoment1,
726 l_int32 area2,
727 l_int32 ycent2,
728 l_int32 maxyshift,
729 l_int32 *tab8,
730 l_int32 *pdelx,
731 l_int32 *pdely,
732 l_float32 *pscore,
733 l_int32 debugflag)
734{
735l_int32 w1, w2, h1, h2, i, j, nx, shifty, delx, dely;
736l_int32 sum, moment, count;
737l_int32 *tab, *area1, *arraysum, *arraymoment;
738l_float32 maxscore, score;
739l_float32 *ycent1;
740FPIX *fpix = NULL;
741PIX *pixt, *pixt1, *pixt2;
742
743 if (pdelx) *pdelx = 0;
744 if (pdely) *pdely = 0;
745 if (pscore) *pscore = 0.0;
746 if (!pix1 || pixGetDepth(pix1) != 1)
747 return ERROR_INT("pix1 not defined or not 1 bpp", __func__, 1);
748 if (!pix2 || pixGetDepth(pix2) != 1)
749 return ERROR_INT("pix2 not defined or not 1 bpp", __func__, 1);
750 if (!nasum1 || !namoment1)
751 return ERROR_INT("nasum1 and namoment1 not both defined", __func__, 1);
752 if (area2 <= 0 || ycent2 <= 0)
753 return ERROR_INT("area2 and ycent2 must be > 0", __func__, 1);
754
755 /* If pix1 (the unknown image) is narrower than pix2,
756 * don't bother to try the match. pix1 is already padded with
757 * 2 pixels on each side. */
758 pixGetDimensions(pix1, &w1, &h1, NULL);
759 pixGetDimensions(pix2, &w2, &h2, NULL);
760 if (w1 < w2) {
761 if (debugflag > 0) {
762 L_INFO("skipping match with w1 = %d and w2 = %d\n",
763 __func__, w1, w2);
764 }
765 return 0;
766 }
767 nx = w1 - w2 + 1;
768
769 if (debugflag > 0)
770 fpix = fpixCreate(nx, 2 * maxyshift + 1);
771 if (!tab8)
772 tab = makePixelSumTab8();
773 else
774 tab = tab8;
775
776 /* Set up the arrays for area1 and ycent1. We have to do this
777 * for each template (pix2) because the window width is w2. */
778 area1 = (l_int32 *)LEPT_CALLOC(nx, sizeof(l_int32));
779 ycent1 = (l_float32 *)LEPT_CALLOC(nx, sizeof(l_int32));
780 arraysum = numaGetIArray(nasum1);
781 arraymoment = numaGetIArray(namoment1);
782 for (i = 0, sum = 0, moment = 0; i < w2; i++) {
783 sum += arraysum[i];
784 moment += arraymoment[i];
785 }
786 for (i = 0; i < nx - 1; i++) {
787 area1[i] = sum;
788 ycent1[i] = (sum == 0) ? ycent2 : (l_float32)moment / (l_float32)sum;
789 sum += arraysum[w2 + i] - arraysum[i];
790 moment += arraymoment[w2 + i] - arraymoment[i];
791 }
792 area1[nx - 1] = sum;
793 ycent1[nx - 1] = (sum == 0) ? ycent2 : (l_float32)moment / (l_float32)sum;
794
795 /* Find the best match location for pix2. At each location,
796 * to insure that pixels are ON only within the intersection of
797 * pix and the shifted pix2:
798 * (1) Start with pixt cleared and equal in size to pix1.
799 * (2) Blit the shifted pix2 onto pixt. Then all ON pixels
800 * are within the intersection of pix1 and the shifted pix2.
801 * (3) AND pix1 with pixt. */
802 pixt = pixCreate(w2, h1, 1);
803 maxscore = 0;
804 delx = 0;
805 dely = 0; /* amount to shift pix2 relative to pix1 to get alignment */
806 for (i = 0; i < nx; i++) {
807 shifty = (l_int32)(ycent1[i] - ycent2 + 0.5);
808 for (j = -maxyshift; j <= maxyshift; j++) {
809 pixClearAll(pixt);
810 pixRasterop(pixt, 0, shifty + j, w2, h2, PIX_SRC, pix2, 0, 0);
811 pixRasterop(pixt, 0, 0, w2, h1, PIX_SRC & PIX_DST, pix1, i, 0);
812 pixCountPixels(pixt, &count, tab);
813 score = (l_float32)count * (l_float32)count /
814 ((l_float32)area1[i] * (l_float32)area2);
815 if (score > maxscore) {
816 maxscore = score;
817 delx = i;
818 dely = shifty + j;
819 }
820
821 if (debugflag > 0)
822 fpixSetPixel(fpix, i, maxyshift + j, 1000.0 * score);
823 }
824 }
825
826 if (debugflag > 0) {
827 char buf[128];
828 lept_mkdir("lept/recog");
829 pixt1 = fpixDisplayMaxDynamicRange(fpix);
830 pixt2 = pixExpandReplicate(pixt1, 5);
831 snprintf(buf, sizeof(buf), "/tmp/lept/recog/junkbs_%d.png", debugflag);
832 pixWrite(buf, pixt2, IFF_PNG);
833 pixDestroy(&pixt1);
834 pixDestroy(&pixt2);
835 fpixDestroy(&fpix);
836 }
837
838 if (pdelx) *pdelx = delx;
839 if (pdely) *pdely = dely;
840 if (pscore) *pscore = maxscore;
841 if (!tab8) LEPT_FREE(tab);
842 LEPT_FREE(area1);
843 LEPT_FREE(ycent1);
844 LEPT_FREE(arraysum);
845 LEPT_FREE(arraymoment);
846 pixDestroy(&pixt);
847 return 0;
848}
849
850
851/*------------------------------------------------------------------------*
852 * Low-level identification *
853 *------------------------------------------------------------------------*/
872l_ok
873recogIdentifyPixa(L_RECOG *recog,
874 PIXA *pixa,
875 PIX **ppixdb)
876{
877char *text;
878l_int32 i, n, fail, index, depth;
879l_float32 score;
880PIX *pix1, *pix2, *pix3;
881PIXA *pixa1;
882L_RCH *rch;
883
884 if (ppixdb) *ppixdb = NULL;
885 if (!recog)
886 return ERROR_INT("recog not defined", __func__, 1);
887 if (!pixa)
888 return ERROR_INT("pixa not defined", __func__, 1);
889
890 /* Run the recognizer on the set of images. This writes
891 * the text string into each pix in pixa. */
892 n = pixaGetCount(pixa);
893 rchaDestroy(&recog->rcha);
894 recog->rcha = rchaCreate();
895 pixa1 = (ppixdb) ? pixaCreate(n) : NULL;
896 depth = 1;
897 for (i = 0; i < n; i++) {
898 pix1 = pixaGetPix(pixa, i, L_CLONE);
899 pix2 = NULL;
900 fail = FALSE;
901 if (!ppixdb)
902 fail = recogIdentifyPix(recog, pix1, NULL);
903 else
904 fail = recogIdentifyPix(recog, pix1, &pix2);
905 if (fail)
906 recogSkipIdentify(recog);
907 if ((rch = recog->rch) == NULL) {
908 L_ERROR("rch not found for char %d\n", __func__, i);
909 pixDestroy(&pix1);
910 pixDestroy(&pix2);
911 continue;
912 }
913 rchExtract(rch, NULL, NULL, &text, NULL, NULL, NULL, NULL);
914 pixSetText(pix1, text);
915 LEPT_FREE(text);
916 if (ppixdb) {
917 rchExtract(rch, &index, &score, NULL, NULL, NULL, NULL, NULL);
918 pix3 = recogShowMatch(recog, pix2, NULL, NULL, index, score);
919 if (i == 0) depth = pixGetDepth(pix3);
920 pixaAddPix(pixa1, pix3, L_INSERT);
921 pixDestroy(&pix2);
922 }
923 transferRchToRcha(rch, recog->rcha);
924 pixDestroy(&pix1);
925 }
926
927 /* Package the images for debug */
928 if (ppixdb) {
929 *ppixdb = pixaDisplayTiledInRows(pixa1, depth, 2500, 1.0, 0, 20, 1);
930 pixaDestroy(&pixa1);
931 }
932
933 return 0;
934}
935
936
963l_ok
964recogIdentifyPix(L_RECOG *recog,
965 PIX *pixs,
966 PIX **ppixdb)
967{
968char *text;
969l_int32 i, j, n, bestindex, bestsample, area1, area2, ret;
970l_int32 shiftx, shifty, bestdelx, bestdely, bestwidth, maxyshift;
971l_float32 x1, y1, x2, y2, delx, dely, score, maxscore;
972NUMA *numa;
973PIX *pix0, *pix1, *pix2;
974PIXA *pixa;
975PTA *pta;
976
977 if (ppixdb) *ppixdb = NULL;
978 if (!recog)
979 return ERROR_INT("recog not defined", __func__, 1);
980 if (!pixs || pixGetDepth(pixs) != 1)
981 return ERROR_INT("pixs not defined or not 1 bpp", __func__, 1);
982
983 /* Do the averaging if required and not yet done. */
984 if (recog->templ_use == L_USE_AVERAGE_TEMPLATES && !recog->ave_done) {
985 ret = recogAverageSamples(recog, 0);
986 if (ret)
987 return ERROR_INT("averaging failed", __func__, 1);
988 }
989
990 /* Binarize and crop to foreground if necessary */
991 if ((pix0 = recogProcessToIdentify(recog, pixs, 0)) == NULL)
992 return ERROR_INT("no fg pixels in pix0", __func__, 1);
993
994 /* Optionally scale and/or convert to fixed stroke width */
995 pix1 = recogModifyTemplate(recog, pix0);
996 pixDestroy(&pix0);
997 if (!pix1)
998 return ERROR_INT("no fg pixels in pix1", __func__, 1);
999
1000 /* Do correlation at all positions within +-maxyshift of
1001 * the nominal centroid alignment. */
1002 pixCountPixels(pix1, &area1, recog->sumtab);
1003 pixCentroid(pix1, recog->centtab, recog->sumtab, &x1, &y1);
1004 bestindex = bestsample = bestdelx = bestdely = bestwidth = 0;
1005 maxscore = 0.0;
1006 maxyshift = recog->maxyshift;
1007 if (recog->templ_use == L_USE_AVERAGE_TEMPLATES) {
1008 for (i = 0; i < recog->setsize; i++) {
1009 numaGetIValue(recog->nasum, i, &area2);
1010 if (area2 == 0) continue; /* no template available */
1011 pix2 = pixaGetPix(recog->pixa, i, L_CLONE);
1012 ptaGetPt(recog->pta, i, &x2, &y2);
1013 delx = x1 - x2;
1014 dely = y1 - y2;
1015 for (shifty = -maxyshift; shifty <= maxyshift; shifty++) {
1016 for (shiftx = -maxyshift; shiftx <= maxyshift; shiftx++) {
1017 pixCorrelationScoreSimple(pix1, pix2, area1, area2,
1018 delx + shiftx, dely + shifty,
1019 5, 5, recog->sumtab, &score);
1020 if (score > maxscore) {
1021 bestindex = i;
1022 bestdelx = delx + shiftx;
1023 bestdely = dely + shifty;
1024 maxscore = score;
1025 }
1026 }
1027 }
1028 pixDestroy(&pix2);
1029 }
1030 } else { /* use all the samples */
1031 for (i = 0; i < recog->setsize; i++) {
1032 pixa = pixaaGetPixa(recog->pixaa, i, L_CLONE);
1033 n = pixaGetCount(pixa);
1034 if (n == 0) {
1035 pixaDestroy(&pixa);
1036 continue;
1037 }
1038 numa = numaaGetNuma(recog->naasum, i, L_CLONE);
1039 pta = ptaaGetPta(recog->ptaa, i, L_CLONE);
1040 for (j = 0; j < n; j++) {
1041 pix2 = pixaGetPix(pixa, j, L_CLONE);
1042 numaGetIValue(numa, j, &area2);
1043 ptaGetPt(pta, j, &x2, &y2);
1044 delx = x1 - x2;
1045 dely = y1 - y2;
1046 for (shifty = -maxyshift; shifty <= maxyshift; shifty++) {
1047 for (shiftx = -maxyshift; shiftx <= maxyshift; shiftx++) {
1048 pixCorrelationScoreSimple(pix1, pix2, area1, area2,
1049 delx + shiftx, dely + shifty,
1050 5, 5, recog->sumtab, &score);
1051 if (score > maxscore) {
1052 bestindex = i;
1053 bestsample = j;
1054 bestdelx = delx + shiftx;
1055 bestdely = dely + shifty;
1056 maxscore = score;
1057 bestwidth = pixGetWidth(pix2);
1058 }
1059 }
1060 }
1061 pixDestroy(&pix2);
1062 }
1063 pixaDestroy(&pixa);
1064 numaDestroy(&numa);
1065 ptaDestroy(&pta);
1066 }
1067 }
1068
1069 /* Package up the results */
1070 recogGetClassString(recog, bestindex, &text);
1071 rchDestroy(&recog->rch);
1072 recog->rch = rchCreate(bestindex, maxscore, text, bestsample,
1073 bestdelx, bestdely, bestwidth);
1074
1075 if (ppixdb) {
1076 if (recog->templ_use == L_USE_AVERAGE_TEMPLATES) {
1077 L_INFO("Best match: str %s; class %d; sh (%d, %d); score %5.3f\n",
1078 __func__, text, bestindex, bestdelx, bestdely, maxscore);
1079 pix2 = pixaGetPix(recog->pixa, bestindex, L_CLONE);
1080 } else { /* L_USE_ALL_TEMPLATES */
1081 L_INFO("Best match: str %s; sample %d in class %d; score %5.3f\n",
1082 __func__, text, bestsample, bestindex, maxscore);
1083 if (maxyshift > 0 && (L_ABS(bestdelx) > 0 || L_ABS(bestdely) > 0)) {
1084 L_INFO(" Best shift: (%d, %d)\n",
1085 __func__, bestdelx, bestdely);
1086 }
1087 pix2 = pixaaGetPix(recog->pixaa, bestindex, bestsample, L_CLONE);
1088 }
1089 *ppixdb = recogShowMatch(recog, pix1, pix2, NULL, -1, 0.0);
1090 pixDestroy(&pix2);
1091 }
1092
1093 pixDestroy(&pix1);
1094 return 0;
1095}
1096
1097
1110l_ok
1111recogSkipIdentify(L_RECOG *recog)
1112{
1113 if (!recog)
1114 return ERROR_INT("recog not defined", __func__, 1);
1115
1116 /* Package up placeholder results */
1117 rchDestroy(&recog->rch);
1118 recog->rch = rchCreate(0, 0.0, stringNew(""), 0, 0, 0, 0);
1119 return 0;
1120}
1121
1122
1123/*------------------------------------------------------------------------*
1124 * Operations for handling identification results *
1125 *------------------------------------------------------------------------*/
1134static L_RCHA *
1135rchaCreate()
1136{
1137L_RCHA *rcha;
1138
1139 rcha = (L_RCHA *)LEPT_CALLOC(1, sizeof(L_RCHA));
1140 rcha->naindex = numaCreate(0);
1141 rcha->nascore = numaCreate(0);
1142 rcha->satext = sarrayCreate(0);
1143 rcha->nasample = numaCreate(0);
1144 rcha->naxloc = numaCreate(0);
1145 rcha->nayloc = numaCreate(0);
1146 rcha->nawidth = numaCreate(0);
1147 return rcha;
1148}
1149
1150
1156void
1157rchaDestroy(L_RCHA **prcha)
1158{
1159L_RCHA *rcha;
1160
1161 if (prcha == NULL) {
1162 L_WARNING("&rcha is null!\n", __func__);
1163 return;
1164 }
1165 if ((rcha = *prcha) == NULL)
1166 return;
1167
1168 numaDestroy(&rcha->naindex);
1169 numaDestroy(&rcha->nascore);
1170 sarrayDestroy(&rcha->satext);
1171 numaDestroy(&rcha->nasample);
1172 numaDestroy(&rcha->naxloc);
1173 numaDestroy(&rcha->nayloc);
1174 numaDestroy(&rcha->nawidth);
1175 LEPT_FREE(rcha);
1176 *prcha = NULL;
1177}
1178
1179
1199static L_RCH *
1200rchCreate(l_int32 index,
1201 l_float32 score,
1202 char *text,
1203 l_int32 sample,
1204 l_int32 xloc,
1205 l_int32 yloc,
1206 l_int32 width)
1207{
1208L_RCH *rch;
1209
1210 rch = (L_RCH *)LEPT_CALLOC(1, sizeof(L_RCH));
1211 rch->index = index;
1212 rch->score = score;
1213 rch->text = text;
1214 rch->sample = sample;
1215 rch->xloc = xloc;
1216 rch->yloc = yloc;
1217 rch->width = width;
1218 return rch;
1219}
1220
1221
1227void
1228rchDestroy(L_RCH **prch)
1229{
1230L_RCH *rch;
1231
1232 if (prch == NULL) {
1233 L_WARNING("&rch is null!\n", __func__);
1234 return;
1235 }
1236 if ((rch = *prch) == NULL)
1237 return;
1238 LEPT_FREE(rch->text);
1239 LEPT_FREE(rch);
1240 *prch = NULL;
1241}
1242
1243
1263l_ok
1264rchaExtract(L_RCHA *rcha,
1265 NUMA **pnaindex,
1266 NUMA **pnascore,
1267 SARRAY **psatext,
1268 NUMA **pnasample,
1269 NUMA **pnaxloc,
1270 NUMA **pnayloc,
1271 NUMA **pnawidth)
1272{
1273 if (pnaindex) *pnaindex = NULL;
1274 if (pnascore) *pnascore = NULL;
1275 if (psatext) *psatext = NULL;
1276 if (pnasample) *pnasample = NULL;
1277 if (pnaxloc) *pnaxloc = NULL;
1278 if (pnayloc) *pnayloc = NULL;
1279 if (pnawidth) *pnawidth = NULL;
1280 if (!rcha)
1281 return ERROR_INT("rcha not defined", __func__, 1);
1282
1283 if (pnaindex) *pnaindex = numaClone(rcha->naindex);
1284 if (pnascore) *pnascore = numaClone(rcha->nascore);
1285 if (psatext) *psatext = sarrayClone(rcha->satext);
1286 if (pnasample) *pnasample = numaClone(rcha->nasample);
1287 if (pnaxloc) *pnaxloc = numaClone(rcha->naxloc);
1288 if (pnayloc) *pnayloc = numaClone(rcha->nayloc);
1289 if (pnawidth) *pnawidth = numaClone(rcha->nawidth);
1290 return 0;
1291}
1292
1293
1307l_ok
1308rchExtract(L_RCH *rch,
1309 l_int32 *pindex,
1310 l_float32 *pscore,
1311 char **ptext,
1312 l_int32 *psample,
1313 l_int32 *pxloc,
1314 l_int32 *pyloc,
1315 l_int32 *pwidth)
1316{
1317 if (pindex) *pindex = 0;
1318 if (pscore) *pscore = 0.0;
1319 if (ptext) *ptext = NULL;
1320 if (psample) *psample = 0;
1321 if (pxloc) *pxloc = 0;
1322 if (pyloc) *pyloc = 0;
1323 if (pwidth) *pwidth = 0;
1324 if (!rch)
1325 return ERROR_INT("rch not defined", __func__, 1);
1326
1327 if (pindex) *pindex = rch->index;
1328 if (pscore) *pscore = rch->score;
1329 if (ptext) *ptext = stringNew(rch->text); /* new string: owned by caller */
1330 if (psample) *psample = rch->sample;
1331 if (pxloc) *pxloc = rch->xloc;
1332 if (pyloc) *pyloc = rch->yloc;
1333 if (pwidth) *pwidth = rch->width;
1334 return 0;
1335}
1336
1337
1351static l_int32
1352transferRchToRcha(L_RCH *rch,
1353 L_RCHA *rcha)
1354{
1355
1356 if (!rch)
1357 return ERROR_INT("rch not defined", __func__, 1);
1358 if (!rcha)
1359 return ERROR_INT("rcha not defined", __func__, 1);
1360
1361 numaAddNumber(rcha->naindex, rch->index);
1362 numaAddNumber(rcha->nascore, rch->score);
1363 sarrayAddString(rcha->satext, rch->text, L_COPY);
1364 numaAddNumber(rcha->nasample, rch->sample);
1365 numaAddNumber(rcha->naxloc, rch->xloc);
1366 numaAddNumber(rcha->nayloc, rch->yloc);
1367 numaAddNumber(rcha->nawidth, rch->width);
1368 return 0;
1369}
1370
1371
1372/*------------------------------------------------------------------------*
1373 * Preprocessing and filtering *
1374 *------------------------------------------------------------------------*/
1391PIX *
1392recogProcessToIdentify(L_RECOG *recog,
1393 PIX *pixs,
1394 l_int32 pad)
1395{
1396l_int32 canclip;
1397PIX *pix1, *pix2, *pixd;
1398
1399 if (!recog)
1400 return (PIX *)ERROR_PTR("recog not defined", __func__, NULL);
1401 if (!pixs)
1402 return (PIX *)ERROR_PTR("pixs not defined", __func__, NULL);
1403
1404 if (pixGetDepth(pixs) != 1)
1405 pix1 = pixThresholdToBinary(pixs, recog->threshold);
1406 else
1407 pix1 = pixClone(pixs);
1408 pixTestClipToForeground(pix1, &canclip);
1409 if (canclip)
1410 pixClipToForeground(pix1, &pix2, NULL);
1411 else
1412 pix2 = pixClone(pix1);
1413 pixDestroy(&pix1);
1414 if (!pix2)
1415 return (PIX *)ERROR_PTR("no foreground pixels", __func__, NULL);
1416
1417 pixd = pixAddBorderGeneral(pix2, pad, pad, 0, 0, 0);
1418 pixDestroy(&pix2);
1419 return pixd;
1420}
1421
1422
1433static PIX *
1434recogPreSplittingFilter(L_RECOG *recog,
1435 PIX *pixs,
1436 l_int32 minh,
1437 l_float32 minaf,
1438 l_int32 debug)
1439{
1440l_int32 scaling, minsplitw, maxsplith, maxasp;
1441BOXA *boxas;
1442NUMA *naw, *nah, *na1, *na1c, *na2, *na3, *na4, *na5, *na6, *na7;
1443PIX *pixd;
1444PIXA *pixas;
1445
1446 if (!recog)
1447 return (PIX *)ERROR_PTR("recog not defined", __func__, NULL);
1448 if (!pixs)
1449 return (PIX *)ERROR_PTR("pixs not defined", __func__, NULL);
1450
1451 /* If there is scaling, do not remove components based on the
1452 * values of min_splitw and max_splith. */
1453 scaling = (recog->scalew > 0 || recog->scaleh > 0) ? TRUE : FALSE;
1454 minsplitw = (scaling) ? 1 : recog->min_splitw - 3;
1455 maxsplith = (scaling) ? 150 : recog->max_splith;
1456 maxasp = recog->max_wh_ratio;
1457
1458 /* Generate an indicator array of connected components to remove:
1459 * short stuff
1460 * tall stuff
1461 * components with large width/height ratio
1462 * components with small area fill fraction */
1463 boxas = pixConnComp(pixs, &pixas, 8);
1464 pixaFindDimensions(pixas, &naw, &nah);
1465 na1 = numaMakeThresholdIndicator(naw, minsplitw, L_SELECT_IF_LT);
1466 na1c = numaCopy(na1);
1467 na2 = numaMakeThresholdIndicator(nah, minh, L_SELECT_IF_LT);
1468 na3 = numaMakeThresholdIndicator(nah, maxsplith, L_SELECT_IF_GT);
1469 na4 = pixaFindWidthHeightRatio(pixas);
1470 na5 = numaMakeThresholdIndicator(na4, maxasp, L_SELECT_IF_GT);
1471 na6 = pixaFindAreaFraction(pixas);
1472 na7 = numaMakeThresholdIndicator(na6, minaf, L_SELECT_IF_LT);
1473 numaLogicalOp(na1, na1, na2, L_UNION);
1474 numaLogicalOp(na1, na1, na3, L_UNION);
1475 numaLogicalOp(na1, na1, na5, L_UNION);
1476 numaLogicalOp(na1, na1, na7, L_UNION);
1477 pixd = pixCopy(NULL, pixs);
1478 pixRemoveWithIndicator(pixd, pixas, na1);
1479 if (debug)
1480 l_showIndicatorSplitValues(na1c, na2, na3, na5, na7, na1);
1481 numaDestroy(&naw);
1482 numaDestroy(&nah);
1483 numaDestroy(&na1);
1484 numaDestroy(&na1c);
1485 numaDestroy(&na2);
1486 numaDestroy(&na3);
1487 numaDestroy(&na4);
1488 numaDestroy(&na5);
1489 numaDestroy(&na6);
1490 numaDestroy(&na7);
1491 boxaDestroy(&boxas);
1492 pixaDestroy(&pixas);
1493 return pixd;
1494}
1495
1496
1508static l_int32
1509recogSplittingFilter(L_RECOG *recog,
1510 PIX *pixs,
1511 l_int32 minh,
1512 l_float32 minaf,
1513 l_int32 *premove,
1514 l_int32 debug)
1515{
1516l_int32 w, h;
1517l_float32 aspratio, fract;
1518
1519 if (!premove)
1520 return ERROR_INT("&remove not defined", __func__, 1);
1521 *premove = 0;
1522 if (!recog)
1523 return ERROR_INT("recog not defined", __func__, 1);
1524 if (!pixs)
1525 return ERROR_INT("pixs not defined", __func__, 1);
1526 if (minh <= 0) minh = DefaultMinHeight;
1527
1528 /* Remove from further consideration:
1529 * small stuff
1530 * components with large width/height ratio
1531 * components with small area fill fraction */
1532 pixGetDimensions(pixs, &w, &h, NULL);
1533 if (w < recog->min_splitw) {
1534 if (debug) L_INFO("w = %d < %d\n", __func__, w, recog->min_splitw);
1535 *premove = 1;
1536 return 0;
1537 }
1538 if (h < minh) {
1539 if (debug) L_INFO("h = %d < %d\n", __func__, h, minh);
1540 *premove = 1;
1541 return 0;
1542 }
1543 aspratio = (l_float32)w / (l_float32)h;
1544 if (aspratio > recog->max_wh_ratio) {
1545 if (debug) L_INFO("w/h = %5.3f too large\n", __func__, aspratio);
1546 *premove = 1;
1547 return 0;
1548 }
1549 pixFindAreaFraction(pixs, recog->sumtab, &fract);
1550 if (fract < minaf) {
1551 if (debug) L_INFO("area fill fract %5.3f < %5.3f\n",
1552 __func__, fract, minaf);
1553 *premove = 1;
1554 return 0;
1555 }
1556
1557 return 0;
1558}
1559
1560
1561/*------------------------------------------------------------------------*
1562 * Postprocessing *
1563 *------------------------------------------------------------------------*/
1598SARRAY *
1599recogExtractNumbers(L_RECOG *recog,
1600 BOXA *boxas,
1601 l_float32 scorethresh,
1602 l_int32 spacethresh,
1603 BOXAA **pbaa,
1604 NUMAA **pnaa)
1605{
1606char *str, *text;
1607l_int32 i, n, x1, x2, h_ovl, v_ovl, h_sep, v_sep;
1608l_float32 score;
1609BOX *box, *prebox;
1610BOXA *ba = NULL;
1611BOXAA *baa;
1612NUMA *nascore, *na = NULL;
1613NUMAA *naa;
1614SARRAY *satext, *sa = NULL, *saout;
1615
1616 if (pbaa) *pbaa = NULL;
1617 if (pnaa) *pnaa = NULL;
1618 if (!recog || !recog->rcha)
1619 return (SARRAY *)ERROR_PTR("recog and rcha not both defined",
1620 __func__, NULL);
1621 if (!boxas)
1622 return (SARRAY *)ERROR_PTR("boxas not defined", __func__, NULL);
1623
1624 if (spacethresh < 0)
1625 spacethresh = L_MAX(recog->maxheight_u, 20);
1626 rchaExtract(recog->rcha, NULL, &nascore, &satext, NULL, NULL, NULL, NULL);
1627 if (!nascore || !satext) {
1628 numaDestroy(&nascore);
1629 sarrayDestroy(&satext);
1630 return (SARRAY *)ERROR_PTR("nascore and satext not both returned",
1631 __func__, NULL);
1632 }
1633
1634 saout = sarrayCreate(0);
1635 naa = numaaCreate(0);
1636 baa = boxaaCreate(0);
1637 prebox = NULL;
1638 n = numaGetCount(nascore);
1639 for (i = 0; i < n; i++) {
1640 numaGetFValue(nascore, i, &score);
1641 text = sarrayGetString(satext, i, L_NOCOPY);
1642 if (prebox == NULL) { /* no current run */
1643 if (score < scorethresh) {
1644 continue;
1645 } else { /* start a number run */
1646 sa = sarrayCreate(0);
1647 ba = boxaCreate(0);
1648 na = numaCreate(0);
1649 sarrayAddString(sa, text, L_COPY);
1650 prebox = boxaGetBox(boxas, i, L_CLONE);
1651 boxaAddBox(ba, prebox, L_COPY);
1652 numaAddNumber(na, score);
1653 }
1654 } else { /* in a current number run */
1655 box = boxaGetBox(boxas, i, L_CLONE);
1656 boxGetGeometry(prebox, &x1, NULL, NULL, NULL);
1657 boxGetGeometry(box, &x2, NULL, NULL, NULL);
1658 boxOverlapDistance(box, prebox, &h_ovl, &v_ovl);
1659 h_sep = -h_ovl;
1660 v_sep = -v_ovl;
1661 boxDestroy(&prebox);
1662 if (x1 < x2 && h_sep <= spacethresh &&
1663 v_sep < 0 && score >= scorethresh) { /* add to number */
1664 sarrayAddString(sa, text, L_COPY);
1665 boxaAddBox(ba, box, L_COPY);
1666 numaAddNumber(na, score);
1667 prebox = box;
1668 } else { /* save the completed number */
1669 str = sarrayToString(sa, 0);
1670 sarrayAddString(saout, str, L_INSERT);
1671 sarrayDestroy(&sa);
1672 boxaaAddBoxa(baa, ba, L_INSERT);
1673 numaaAddNuma(naa, na, L_INSERT);
1674 boxDestroy(&box);
1675 if (score >= scorethresh) { /* start a new number */
1676 i--;
1677 continue;
1678 }
1679 }
1680 }
1681 }
1682
1683 if (prebox) { /* save the last number */
1684 str = sarrayToString(sa, 0);
1685 sarrayAddString(saout, str, L_INSERT);
1686 boxaaAddBoxa(baa, ba, L_INSERT);
1687 numaaAddNuma(naa, na, L_INSERT);
1688 sarrayDestroy(&sa);
1689 boxDestroy(&prebox);
1690 }
1691
1692 numaDestroy(&nascore);
1693 sarrayDestroy(&satext);
1694 if (sarrayGetCount(saout) == 0) {
1695 sarrayDestroy(&saout);
1696 boxaaDestroy(&baa);
1697 numaaDestroy(&naa);
1698 L_INFO("saout has no identified text\n", __func__);
1699 return NULL;
1700 }
1701
1702 if (pbaa)
1703 *pbaa = baa;
1704 else
1705 boxaaDestroy(&baa);
1706 if (pnaa)
1707 *pnaa = naa;
1708 else
1709 numaaDestroy(&naa);
1710 return saout;
1711}
1712
1731PIXA *
1732showExtractNumbers(PIX *pixs,
1733 SARRAY *sa,
1734 BOXAA *baa,
1735 NUMAA *naa,
1736 PIX **ppixdb)
1737{
1738char buf[128];
1739char *textstr, *scorestr;
1740l_int32 i, j, n, nchar, len;
1741l_float32 score;
1742L_BMF *bmf;
1743BOX *box1, *box2;
1744BOXA *ba;
1745NUMA *na;
1746PIX *pix1, *pix2, *pix3, *pix4;
1747PIXA *pixa;
1748
1749 if (ppixdb) *ppixdb = NULL;
1750 if (!pixs)
1751 return (PIXA *)ERROR_PTR("pixs not defined", __func__, NULL);
1752 if (!sa)
1753 return (PIXA *)ERROR_PTR("sa not defined", __func__, NULL);
1754 if (!baa)
1755 return (PIXA *)ERROR_PTR("baa not defined", __func__, NULL);
1756 if (!naa)
1757 return (PIXA *)ERROR_PTR("naa not defined", __func__, NULL);
1758
1759 n = sarrayGetCount(sa);
1760 pixa = pixaCreate(n);
1761 bmf = bmfCreate(NULL, 6);
1762 if (ppixdb) *ppixdb = pixConvertTo8(pixs, 1);
1763 for (i = 0; i < n; i++) {
1764 textstr = sarrayGetString(sa, i, L_NOCOPY);
1765 ba = boxaaGetBoxa(baa, i, L_CLONE);
1766 na = numaaGetNuma(naa, i, L_CLONE);
1767 boxaGetExtent(ba, NULL, NULL, &box1);
1768 box2 = boxAdjustSides(NULL, box1, -5, 5, -5, 5);
1769 if (ppixdb) pixRenderBoxArb(*ppixdb, box2, 3, 255, 0, 0);
1770 pix1 = pixClipRectangle(pixs, box1, NULL);
1771 len = strlen(textstr) + 1;
1772 pix2 = pixAddBlackOrWhiteBorder(pix1, 14 * len, 14 * len,
1773 5, 3, L_SET_WHITE);
1774 pix3 = pixConvertTo8(pix2, 1);
1775 nchar = numaGetCount(na);
1776 scorestr = NULL;
1777 for (j = 0; j < nchar; j++) {
1778 numaGetFValue(na, j, &score);
1779 snprintf(buf, sizeof(buf), "%d", (l_int32)(100 * score));
1780 stringJoinIP(&scorestr, buf);
1781 if (j < nchar - 1) stringJoinIP(&scorestr, ",");
1782 }
1783 snprintf(buf, sizeof(buf), "%s: %s\n", textstr, scorestr);
1784 pix4 = pixAddTextlines(pix3, bmf, buf, 0xff000000, L_ADD_BELOW);
1785 pixaAddPix(pixa, pix4, L_INSERT);
1786 boxDestroy(&box1);
1787 boxDestroy(&box2);
1788 pixDestroy(&pix1);
1789 pixDestroy(&pix2);
1790 pixDestroy(&pix3);
1791 boxaDestroy(&ba);
1792 numaDestroy(&na);
1793 LEPT_FREE(scorestr);
1794 }
1795
1796 bmfDestroy(&bmf);
1797 return pixa;
1798}
1799
1800
1801/*------------------------------------------------------------------------*
1802 * Static debug helper *
1803 *------------------------------------------------------------------------*/
1816static void
1817l_showIndicatorSplitValues(NUMA *na1,
1818 NUMA *na2,
1819 NUMA *na3,
1820 NUMA *na4,
1821 NUMA *na5,
1822 NUMA *na6)
1823{
1824l_int32 i, n;
1825
1826 n = numaGetCount(na1);
1827 lept_stderr("================================================\n");
1828 lept_stderr("lt minw: ");
1829 for (i = 0; i < n; i++)
1830 lept_stderr("%4d ", (l_int32)na1->array[i]);
1831 lept_stderr("\nlt minh: ");
1832 for (i = 0; i < n; i++)
1833 lept_stderr("%4d ", (l_int32)na2->array[i]);
1834 lept_stderr("\ngt maxh: ");
1835 for (i = 0; i < n; i++)
1836 lept_stderr("%4d ", (l_int32)na3->array[i]);
1837 lept_stderr("\ngt maxasp: ");
1838 for (i = 0; i < n; i++)
1839 lept_stderr("%4d ", (l_int32)na4->array[i]);
1840 lept_stderr("\nlt minaf: ");
1841 for (i = 0; i < n; i++)
1842 lept_stderr("%4d ", (l_int32)na5->array[i]);
1843 lept_stderr("\n------------------------------------------------");
1844 lept_stderr("\nresult: ");
1845 for (i = 0; i < n; i++)
1846 lept_stderr("%4d ", (l_int32)na6->array[i]);
1847 lept_stderr("\n================================================\n");
1848}
array_internal.h
PIX_DST
#define PIX_DST
Definition pix.h:445
L_SELECT_IF_LT
@ L_SELECT_IF_LT
Definition pix.h:575
L_SELECT_IF_GT
@ L_SELECT_IF_GT
Definition pix.h:576
L_SORT_BY_X
@ L_SORT_BY_X
Definition pix.h:528
L_ADD_BELOW
@ L_ADD_BELOW
Definition pix.h:1003
L_COPY
@ L_COPY
Definition pix.h:505
L_CLONE
@ L_CLONE
Definition pix.h:506
L_NOCOPY
@ L_NOCOPY
Definition pix.h:503
L_INSERT
@ L_INSERT
Definition pix.h:504
L_SET_WHITE
@ L_SET_WHITE
Definition pix.h:699
L_COMBINE
@ L_COMBINE
Definition pix.h:880
PIX_SRC
#define PIX_SRC
Definition pix.h:444
L_SORT_INCREASING
@ L_SORT_INCREASING
Definition pix.h:522
L_USE_AVERAGE_TEMPLATES
@ L_USE_AVERAGE_TEMPLATES
Definition recog.h:261
recogProcessToIdentify
PIX * recogProcessToIdentify(L_RECOG *recog, PIX *pixs, l_int32 pad)
recogProcessToIdentify()
Definition recogident.c:1392
recogCorrelationBestChar
l_ok recogCorrelationBestChar(L_RECOG *recog, PIX *pixs, BOX **pbox, l_float32 *pscore, l_int32 *pindex, char **pcharstr, PIX **ppixdb)
recogCorrelationBestChar()
Definition recogident.c:582
rchaDestroy
void rchaDestroy(L_RCHA **prcha)
rchaDestroy()
Definition recogident.c:1157
rchExtract
l_ok rchExtract(L_RCH *rch, l_int32 *pindex, l_float32 *pscore, char **ptext, l_int32 *psample, l_int32 *pxloc, l_int32 *pyloc, l_int32 *pwidth)
rchExtract()
Definition recogident.c:1308
recogSplittingFilter
static l_int32 recogSplittingFilter(L_RECOG *recog, PIX *pixs, l_int32 min, l_float32 minaf, l_int32 *premove, l_int32 debug)
recogSplittingFilter()
Definition recogident.c:1509
l_showIndicatorSplitValues
static void l_showIndicatorSplitValues(NUMA *na1, NUMA *na2, NUMA *na3, NUMA *na4, NUMA *na5, NUMA *na6)
l_showIndicatorSplitValues()
Definition recogident.c:1817
transferRchToRcha
static l_int32 transferRchToRcha(L_RCH *rch, L_RCHA *rcha)
transferRchToRcha()
Definition recogident.c:1352
pixCorrelationBestShift
static l_int32 pixCorrelationBestShift(PIX *pix1, PIX *pix2, NUMA *nasum1, NUMA *namoment1, l_int32 area2, l_int32 ycent2, l_int32 maxyshift, l_int32 *tab8, l_int32 *pdelx, l_int32 *pdely, l_float32 *pscore, l_int32 debugflag)
pixCorrelationBestShift()
Definition recogident.c:722
rchDestroy
void rchDestroy(L_RCH **prch)
rchDestroy()
Definition recogident.c:1228
recogIdentifyPixa
l_ok recogIdentifyPixa(L_RECOG *recog, PIXA *pixa, PIX **ppixdb)
recogIdentifyPixa()
Definition recogident.c:873
recogIdentifyPix
l_ok recogIdentifyPix(L_RECOG *recog, PIX *pixs, PIX **ppixdb)
recogIdentifyPix()
Definition recogident.c:964
recogExtractNumbers
SARRAY * recogExtractNumbers(L_RECOG *recog, BOXA *boxas, l_float32 scorethresh, l_int32 spacethresh, BOXAA **pbaa, NUMAA **pnaa)
recogExtractNumbers()
Definition recogident.c:1599
rchaCreate
static L_RCHA * rchaCreate()
rchaCreate()
Definition recogident.c:1135
recogCorrelationBestRow
l_ok recogCorrelationBestRow(L_RECOG *recog, PIX *pixs, BOXA **pboxa, NUMA **pnascore, NUMA **pnaindex, SARRAY **psachar, l_int32 debug)
recogCorrelationBestRow()
Definition recogident.c:402
recogSkipIdentify
l_ok recogSkipIdentify(L_RECOG *recog)
recogSkipIdentify()
Definition recogident.c:1111
recogIdentifyMultiple
l_ok recogIdentifyMultiple(L_RECOG *recog, PIX *pixs, l_int32 minh, l_int32 skipsplit, BOXA **pboxa, PIXA **ppixa, PIX **ppixdb, l_int32 debugsplit)
recogIdentifyMultiple()
Definition recogident.c:163
recogPreSplittingFilter
static PIX * recogPreSplittingFilter(L_RECOG *recog, PIX *pixs, l_int32 minh, l_float32 minaf, l_int32 debug)
recogPreSplittingFilter()
Definition recogident.c:1434
recogSplitIntoCharacters
l_ok recogSplitIntoCharacters(L_RECOG *recog, PIX *pixs, l_int32 minh, l_int32 skipsplit, BOXA **pboxa, PIXA **ppixa, l_int32 debug)
recogSplitIntoCharacters()
Definition recogident.c:249
rchaExtract
l_ok rchaExtract(L_RCHA *rcha, NUMA **pnaindex, NUMA **pnascore, SARRAY **psatext, NUMA **pnasample, NUMA **pnaxloc, NUMA **pnayloc, NUMA **pnawidth)
rchaExtract()
Definition recogident.c:1264
showExtractNumbers
PIXA * showExtractNumbers(PIX *pixs, SARRAY *sa, BOXAA *baa, NUMAA *naa, PIX **ppixdb)
showExtractNumbers()
Definition recogident.c:1732
rchCreate
static L_RCH * rchCreate(l_int32 index, l_float32 score, char *text, l_int32 sample, l_int32 xloc, l_int32 yloc, l_int32 width)
rchCreate()
Definition recogident.c:1200
L_Bmf
Definition bmf.h:47
L_Rch
Definition recog.h:182
L_Rch::sample
l_int32 sample
Definition recog.h:186
L_Rch::yloc
l_int32 yloc
Definition recog.h:189
L_Rch::score
l_float32 score
Definition recog.h:184
L_Rch::text
char * text
Definition recog.h:185
L_Rch::index
l_int32 index
Definition recog.h:183
L_Rch::xloc
l_int32 xloc
Definition recog.h:188
L_Rch::width
l_int32 width
Definition recog.h:190
L_Rcha
Definition recog.h:197
L_Rcha::nasample
struct Numa * nasample
Definition recog.h:201
L_Rcha::satext
struct Sarray * satext
Definition recog.h:200
L_Rcha::nascore
struct Numa * nascore
Definition recog.h:199
L_Rcha::nawidth
struct Numa * nawidth
Definition recog.h:204
L_Rcha::naxloc
struct Numa * naxloc
Definition recog.h:202
L_Rcha::naindex
struct Numa * naindex
Definition recog.h:198
L_Rcha::nayloc
struct Numa * nayloc
Definition recog.h:203
L_Recog
Definition recog.h:116
L_Recog::nasum
struct Numa * nasum
Definition recog.h:163
L_Recog::ave_done
l_int32 ave_done
Definition recog.h:141
L_Recog::templ_use
l_int32 templ_use
Definition recog.h:123
L_Recog::centtab
l_int32 * centtab
Definition recog.h:150
L_Recog::rch
struct L_Rch * rch
Definition recog.h:174
L_Recog::pixa_u
struct Pixa * pixa_u
Definition recog.h:158
L_Recog::train_done
l_int32 train_done
Definition recog.h:142
L_Recog::maxwidth_u
l_int32 maxwidth_u
Definition recog.h:136
L_Recog::pta_u
struct Pta * pta_u
Definition recog.h:159
L_Recog::maxyshift
l_int32 maxyshift
Definition recog.h:129
L_Recog::scalew
l_int32 scalew
Definition recog.h:117
L_Recog::ptaa
struct Ptaa * ptaa
Definition recog.h:156
L_Recog::maxheight_u
l_int32 maxheight_u
Definition recog.h:138
L_Recog::min_splitw
l_int32 min_splitw
Definition recog.h:146
L_Recog::threshold
l_int32 threshold
Definition recog.h:128
L_Recog::nasum_u
struct Numa * nasum_u
Definition recog.h:160
L_Recog::rcha
struct L_Rcha * rcha
Definition recog.h:175
L_Recog::scaleh
l_int32 scaleh
Definition recog.h:119
L_Recog::max_splith
l_int32 max_splith
Definition recog.h:147
L_Recog::sumtab
l_int32 * sumtab
Definition recog.h:151
L_Recog::pixadb_split
struct Pixa * pixadb_split
Definition recog.h:170
L_Recog::pta
struct Pta * pta
Definition recog.h:162
L_Recog::minwidth_u
l_int32 minwidth_u
Definition recog.h:135
L_Recog::naasum
struct Numaa * naasum
Definition recog.h:157
L_Recog::max_wh_ratio
l_float32 max_wh_ratio
Definition recog.h:144
L_Recog::setsize
l_int32 setsize
Definition recog.h:127
L_Recog::pixaa
struct Pixaa * pixaa
Definition recog.h:155
L_Recog::pixa
struct Pixa * pixa
Definition recog.h:161
Numa::array
l_float32 * array
Definition array_internal.h:74