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 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
|
/*
Copyright (C) 1999-2001 Ricardo Ueda Karpischek
This is free software; you can redistribute it and/or modify
it under the terms of the version 2 of the GNU General Public
License as published by the Free Software Foundation.
This software is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this software; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
USA.
*/
/*
build.c: The function build.
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "common.h"
/*
Magic numbers
m_mwd .. Maximum horizontal distance accepted in order to
consider two words aligned, specified as a percentage of
x_height. Values greater than 100 are accepted as well.
m_msd .. Maximum horizontal distance accepted in order to
consider two symbols aligned, specified as a percentage of
x_height. Values greater than 100 are accepted as well.
m_mae .. Acceptable error when testing fitting of one symbol on the
expected limits for the ascent and descent, specified as a
percentage of the dot diameter. Values greater than 100 are
accepted as well.
m_ds .. Descent (relative to baseline) as a percentage of the dot
diameter. Usually greater than 100 (for instance 200).
m_as .. Ascent (relative to baseline) as a percentage of the dot
diameter. Usually greater than 100 (for instance 800).
m_xh .. x_height (relative to baseline) as a percentage of the dot
diameter. Usually greater than 100 (for instance 600).
m_fs .. number of steps to complete one unity on float steppers.
*/
#ifdef CF
/* Candido de Figueiredo */
int m_mwd = 1000;
int m_msd = 40;
int m_mae = 60;
int m_ds = 150;
int m_as = 450;
int m_xh = 300;
int m_fs = 20;
#else
int m_mwd = 1000;
int m_msd = 60;
int m_mae = 50;
int m_ds = 200;
int m_as = 800;
int m_xh = 600;
int m_fs = 20;
#endif
/*
Bold and italic distinguishers.
*/
int bw[256],bh[256],bp[256];
int iw[256],ih[256],ip[256];
/*
Validation of the recognition of the symbol i.
*/
void recog_validation(int i)
{
int k,b;
sdesc *m;
/* the symbol we're validating */
m = mc + (k=i);
/*
Ignore DOTS with invalid alignments. The rationale here
is that noise is easily classified as dot. So we must ignore all
dots without a typical dot context, like the top of "i" or
"j", period, etc. An easy approximation for this criteria is
considering the alignment. So we remove all SHAPE votes
from dots with unknown alignment.
*/
if ((m->tc == DOT) && (m->va < 0)) {
rmvotes(SHAPE,k,-1,0);
}
/*
Apply add-hoc filtering rules.
*/
{
unsigned char *t;
if ((m->tr != NULL) &&
(strlen(t = (unsigned char *)((m->tr)->t)) == 1) &&
(avoid(k,t[0]))) {
rmvotes(SHAPE,k,-1,0);
/*
BUG: We must reset lfa and bm only when the symbol
is not a pattern. By now we're using a simple but
broken test. Must change in the future.
*/
if (m->tr == NULL) {
m->lfa = -1;
m->bm = -1;
}
}
}
/* attach accents to their base letters */
if (m->tc == ACCENT) {
int *t;
/*
strangely enough, it seems that some untransliterated symbols
are being classified as accents.
*/
if (m->tr == NULL) {
db("oops.. untransliterated symbol classified as accent");
}
else if ((b = bsymb(k,0)) >= 0) {
if (box_dist(ps[i],b,NULL) < ((mc[b].b-mc[b].t)/2)) {
for (t = &(mc[b].sl); (*t >= 0); t = &(mc[*t].sl));
*t = i;
}
}
}
}
/*
Returns -1 if the line a precedes the line b,
or +1 if the line b precedes the line a, or
0 if cannot decide.
*/
int cmpln(int a,int b)
{
sdesc *p,*q;
p = mc + word[line[a].f].F;
q = mc + word[line[b].f].F;
if ((p->b+p->t)/2 + 15 < (q->b+q->t)/2)
return(-1);
else if ((q->b+q->t)/2 + 15 < (p->b+p->t)/2)
return(1);
else if ((p->l+p->r)/2 < (q->l+q->r)/2)
return(-1);
else if ((q->l+q->r)/2 < (p->l+p->r)/2)
return(1);
else
return(0);
}
/*
Account bold and italic distinguisher symbols. This simplistic
code assumes only one bold font size along the book. If more than
one such size exists, detection of bold words will become
compromised (not a big problem, though).
*/
void acc_f(int *fw,int *fh,int *fp,int FF)
{
int c,i,f[256],o[256];
int ow[256],oh[256],op[256];
pdesc *p;
/* prepare account */
for (i=0; i<256; ++i) {
fw[i] = fh[i] = fp[i] = 0;
ow[i] = oh[i] = op[i] = 0;
f[i] = o[i] = 0;
}
/* account featured symbols */
for (i=0; i<=topp; ++i) {
p = pattern + i;
if ((p->tr != NULL) &&
(strlen(p->tr) == 1)) {
c = ((unsigned char *)(p->tr))[0];
if (p->f & FF) {
fw[c] += p->bw;
fh[c] += p->bh;
fp[c] += p->bp;
++(f[c]);
}
else {
ow[c] += p->bw;
oh[c] += p->bh;
op[c] += p->bp;
++(o[c]);
}
}
}
/* search bold distinguishers */
for (c=0; c<256; ++c) {
if ((f[c] > 0) && (o[c] > 0)) {
fw[c] /= f[c];
fh[c] /= f[c];
fp[c] /= f[c];
ow[c] /= o[c];
oh[c] /= o[c];
op[c] /= o[c];
if (abs(fp[c]-op[c]) <= (fp[c]/10))
fp[c] = 0;
if (abs(fh[c]-oh[c]) <= (fh[c]/10))
fh[c] = 0;
if (abs(fw[c]-ow[c]) <= (fw[c]/10))
fw[c] = 0;
/*
if ((fp[c] > 0) || (fh[c] > 0) || (fw[c] > 0))
printf("%d distinguisher: %d %c\n",FF,c,c);
*/
}
}
}
/*
Compute word properties.
*/
void wprops(int i)
{
/* compute baseline (TO BE DONE) */
/* italic and bold flags, and the word type */
{
wdesc *w;
int j,it,bd;
w = word + i;
it = 0;
bd = 0;
w->a = UNDEF;
for (j=word[i].F; j>=0; j=mc[j].E) {
if (!uncertain(mc[j].tc)) {
unsigned char *t;
int c,w,h,p;
t = ((unsigned char *) (mc[j].tr->t));
if (strlen(t) == 1) {
c = t[0];
w = mc[j].r - mc[j].l + 1;
h = mc[j].b - mc[j].t + 1;
p = mc[j].nbp;
if (((mc[j].tr->f & F_ITALIC) != 0) &&
(((iw[c] > 0) && (abs(iw[c]-w) < iw[c]/20)) ||
((ih[c] > 0) && (abs(ih[c]-h) < ih[c]/20)) ||
((ip[c] > 0) && (abs(ip[c]-p) < ip[c]/20)))) {
it = 1;
}
if (((mc[j].tr->f & F_BOLD) != 0) &&
(((bw[c] > 0) && (abs(bw[c]-w) < bw[c]/20)) ||
((bh[c] > 0) && (abs(bh[c]-h) < bh[c]/20)) ||
((bp[c] > 0) && (abs(bp[c]-p) < bp[c]/20)))) {
bd = 1;
}
}
}
}
if (it)
C_SET(w->f,F_ITALIC);
else
C_UNSET(w->f,F_ITALIC);
if (bd)
C_SET(w->f,F_BOLD);
else
C_UNSET(w->f,F_BOLD);
}
/* compute bounding box */
{
int l,r,t,b;
sdesc *s;
s = mc + word[i].F;
l = s->l;
r = s->r;
t = s->t;
b = s->b;
while (s->E >= 0) {
s = mc + s->E;
if (s->l < l)
l = s->l;
if (s->r > r)
r = s->r;
if (s->t < t)
t = s->t;
if (s->b > b)
b = s->b;
}
word[i].l = l;
word[i].r = r;
word[i].t = t;
word[i].b = b;
}
}
/*
Word pairing
------------
The word pairing test is used to build the text lines. The function
w_pair tests word pairing, and returns the following diagnostics:
0 .. the words are paired
1 .. insufficient vertical intersection
2 .. incomplete data
3 .. maximum horizontal distance exceeded
*/
int w_pair(int wa,int wb)
{
int a,b;
int r,t,d,e;
int dd,as,bl,xh,ds;
int pa,pb;
a = word[wa].L;
b = word[wb].F;
/* consist */
if ((a < 0) || (tops < a) || (b < 0) || (tops < b)) {
db("inconsistent input at w_pair");
return(2);
}
/* compute horizontal distance */
if (mc[a].r < mc[b].l)
d = mc[b].l - mc[a].r;
else if (mc[a].l+mc[a].r < mc[b].l+mc[b].r)
d = 1;
else
d = 0;
/* vertical intersection */
t = intersize(mc[a].t,mc[a].b,mc[b].t,mc[b].b,NULL,NULL);
/* check if wa or wb are pseudo-words */
pa = ((mc[a].W < 0) && ((mc[a].tc == DOT) || (mc[a].tc == COMMA)));
pb = ((mc[b].E < 0) && ((mc[b].tc == DOT) || (mc[b].tc == COMMA)));
/* try to obtain x_height and baseline */
dd = complete_align(a,b,&as,&xh,&bl,&ds);
/* horizontal distance unacceptable (1st test) */
if ((d <= 0) || (((m_mwd*FS)/100) <= d))
r = 3;
/* pseudo-words special case */
else if (pa || pb) {
/* non-null intersection is the easy case */
if (t > 0)
return(0);
if (dd < 0) {
if (mc[a].tc == DOT)
dd = ((mc[a].r-mc[a].l)+(mc[a].b-mc[a].t)+2) / 2;
else if (mc[b].tc == DOT)
dd = ((mc[b].r-mc[b].l)+(mc[b].b-mc[b].t)+2) / 2;
else if (mc[a].tc == COMMA)
dd = (mc[a].r-mc[a].l)+2 / 2;
}
/* could not estimate the dot diameter */
if (dd < 0)
return(2);
/* vertical distance between a and b */
e = ldist(mc[a].b,mc[a].t,mc[b].b,mc[b].t);
/* small vertical distance */
if (e <= 2*dd)
r = 0;
/* large vertical distance */
else
r = 1;
}
/* insufficient data */
else if ((xh <= 0) || (bl <= 0))
r = 2;
/* horizontal distance unacceptable (2nd test) */
else if ((d <= 0) || (((m_mwd*(bl-xh))/100) <= d))
r = 3;
/*
BUG: hardcoded magic number
*/
else if (t > 5) {
r = 0;
}
/* no vertical intersection */
else
r = 1;
return(r);
}
/*
Diagnose word pairing.
*/
void diag_wpairing(int t)
{
int r,w1,w2;
if ((curr_mc < 0) ||
(t < 0) ||
((w1=mc[curr_mc].sw) < 0) ||
((w2=mc[t].sw) < 0)) {
show_hint(2,"invalid parameters");
return;
}
else if ((mc[curr_mc].l+mc[curr_mc].r) < (mc[t].l+mc[t].r))
r = w_pair(w1,w2);
else
r = w_pair(w2,w1);
if (r == 0)
show_hint(2,"pairing successful");
if (r == 1)
show_hint(2,"insuficcient vertical intersection");
if (r == 2)
show_hint(2,"incomplete data");
if (r == 3)
show_hint(2,"maximum horizontal distance exceeded");
}
/*
Diagnose pairing against current symbol.
*/
void diag_pairing(int t)
{
int r;
if ((curr_mc < 0) || (t < 0)) {
show_hint(2,"invalid parameters");
return;
}
else if ((mc[curr_mc].l+mc[curr_mc].r) < (mc[t].l+mc[t].r))
r = s_pair(curr_mc,t,0,NULL);
else
r = s_pair(t,curr_mc,0,NULL);
if (r == 0)
show_hint(2,"pairing successful");
if (r == 1)
show_hint(2,"insuficcient vertical intersection");
if (r == 2)
show_hint(2,"one or both symbols above ascent");
if (r == 3)
show_hint(2,"one or both symbols below descent");
if (r == 4)
show_hint(2,"maximum horizontal distance exceeded");
if (r == 5)
show_hint(2,"incomplete data");
}
/*
Detect the word at the left side of the symbol t.
This routine was part of the "extend left word" feature. That
feature was removed from the OCR. This code is being mantained
here because it may be used again someday.
*/
int left_word(int t)
{
int i,w,a,d,l,x;
/*
To obtain the left word we use a
straighforward method. Just search all words and, from those
whose last symbol has vertical intersection with t, choose
the one horizontally nearest to t.
*/
w = -1;
l = (mc[t].l + mc[t].r) / 2;
for (i=0, d=-1; i<=topw; ++i) {
/* new best candidate */
if ((word[i].F >= 0) &&
((a=word[i].L) >= 0) &&
(intersize(mc[a].t,mc[a].b,mc[t].t,mc[t].b,NULL,NULL) > 0) &&
((x = (mc[a].r+mc[a].l) / 2) < l) &&
((w < 0) || (l-x < d))) {
w = i;
d = l-x;
}
}
/* return the left word (or -1) */
return(w);
}
/*
Create and return an unitary word for symbol k.
*/
int new_word(int k) {
int cw;
/* enlarge buffer */
if (++topw >= wsz) {
wsz = topw + 512;
word = c_realloc(word,sizeof(wdesc)*wsz,NULL);
}
word[topw--].F = -1;
/* find and use the next free entry */
for (cw=0; word[cw].F >= 0; ++cw);
word[cw].F = word[cw].L = k;
word[cw].bl = -1;
word[cw].E = -1;
word[cw].W = -1;
word[cw].f = 0;
if (cw > topw)
topw = cw;
mc[k].sw = cw;
return(cw);
}
/*
The build function
------------------
*/
int build(int reset)
{
static int m,k,st;
static int i;
int j;
/* reset state */
if (reset) {
st = 0;
topw = -1;
topln = -1;
show_hint(0,"preparing");
return(1);
}
/*
** STATE 0 **
Preparation
*/
if (st == 0) {
/* (devel)
The build step
--------------
The "build" OCR step, implemented by the "build"
function, distributes the symbols on words
(analysing the distance, heights and relative
position for each pair of symbols), and the words
on lines (analysing the distance, heights and
relative position for each pair of words). Various
important heuristics take effect here.
0. Preparation
The first step of build is to distribute the symbols
on words. This is achieved by:
a. Undefining the next-symbol ("E" field) and previous-symbol
("W" field) links for each symbol, the surrounding word ("sw"
field) of each symbol, and the next signal ("sl" field) for
each symbol.
Obs. The next-symbol and previous symbol links are used
to build the list of symbols of each word. For instance,
on the word "goal", "o" is the next for "g" and
the previous for "a", "g" has no previous and "l"
has no next).
*/
for (m=0; m<=tops; ++m) {
mc[m].E = mc[m].W = mc[m].sw = mc[m].sl = -1;
}
/* (devel)
b. Undefining the transliteration class of SCHARs and
the uncertain alignment information.
*/
for (m=0; m<=tops; ++m) {
if (mc[m].sw < 0) {
if (mc[m].tc == SCHAR)
mc[m].tc = UNDEF;
if (uncertain(mc[m].tc))
mc[m].va = -1;
}
}
/* prepare next state */
show_hint(0,"detecting words");
m = 0;
st = 1;
}
/*
** STATE 1 **
distributing symbols on words
*/
else if (st == 1) {
/* (devel)
2. Distributing symbols on words
The second step is, for each CHAR not in any word, define
a unitary word around it and extend it to right
and left applying the symbol pairing test. When
extending, merge words when necessary.
*/
if (m <= topps) {
k = ps[m];
if ((mc[k].tc == CHAR) && (mc[k].sw < 0)) {
int a,b,cw,ow,f,t;
/* create unitary word */
cw = new_word(k);
/* extend to right */
for (a=k; ((b=rsymb(a,0)) >= 0); a=b) {
/* classify as SCHAR */
if (mc[b].tc==UNDEF)
mc[b].tc = SCHAR;
/* extend word or... */
if ((ow=mc[b].sw) < 0) {
mc[a].E = b;
mc[b].W = a;
mc[b].sw = cw;
word[cw].L = b;
}
/*
... merge words
WARNING: this code creates unused entries on the
array word below topw. Some of them will be
reused by the word creation code, though.
*/
else {
/* medium vertical line */
t = mc[a].l + mc[a].r;
/* consist and merge */
if (((f=word[ow].F) == b) &&
(t < mc[f].l+mc[f].r)) {
for (f=word[ow].F; f>=0; f=mc[f].E)
mc[f].sw = cw;
mc[a].E = b;
mc[b].W = a;
mc[b].sw = cw;
word[cw].L = word[ow].L;
word[ow].F = -1;
}
}
}
/* extend to left */
for (a=k; (b=lsymb(a,0)) >= 0; a=b) {
/* classify as SCHAR */
if (mc[b].tc==UNDEF)
mc[b].tc = SCHAR;
/* extend word or... */
if ((ow=mc[b].sw) < 0) {
mc[b].E = a;
mc[a].W = b;
mc[b].sw = cw;
word[cw].F = b;
}
/*
... merge words
WARNING: this code creates unused entries on the
array word below topw. Some of them will be
reused by the word creation code, though.
*/
else {
/* medium vertical line */
t = mc[a].l + mc[a].r;
/* consist and merge */
if (((f=word[ow].L) == b) &&
(mc[f].l+mc[f].r < t)) {
for (f=word[ow].F; f>=0; f=mc[f].E)
mc[f].sw = cw;
mc[b].E = a;
mc[a].W = b;
mc[b].sw = cw;
word[cw].F = word[ow].F;
word[ow].F = -1;
}
}
}
}
/* prepare next symbol */
if ((++m % DPROG) == 0)
show_hint(0,"detecting words %d/%d",m,topps);
}
/* prepare next state */
else {
show_hint(0,"merging");
st = 2;
i = 0;
}
}
/*
** STATE 2 **
Currently empty
*/
else if (st == 2) {
/*
(Currently empty)
*/
/* prepare next state */
show_hint(0,"computing alignment");
i = 0;
st = 3;
}
/*
** STATE 3 **
Computing the alignment using the words
*/
else if (st == 3) {
int k,l,dd,as,xh,bl,ds;
sdesc *m,*t;
/* (devel)
3. Computing the alignment using the words
Some symbols do not have a well-defined alignment by
themselves. For instance, a dot may be baseline-aligned
(a final dot) or 0-aligned (the "i" dot). So when
computing their alignments, we need to analyse their
neighborhoods. This is performed in this step.
*/
/* compute the alignment of dots and accents */
if (i <= topps) {
m = mc + (k = ps[i]);
#ifdef MEMCHECK
checkidx(i,topps+1,"build state 3.1");
#endif
/*
Alignment for dots may be 0 ("i", "j"),
11 (the top one in ":") or 22 (period).
Alignment for commas may be 22 (comma)
or 0 (apostrophe).
We expect alignment 0 for latin accents.
*/
if ((m->tc == DOT) || (m->tc == COMMA) || (m->tc == ACCENT)) {
int x,y,w,h;
/* reset alignment */
m->va = -1;
/* compute neighbours */
x = (m->l+m->r)/2 - FS;
y = (m->t+m->b)/2 - FS;
w = h = 2*FS;
list_s(x,y,w,h);
#ifdef MEMCHECK
checkidx(list_s_sz,tops+1,"build state 3.2");
#endif
/*
Try to compute the alignment using
some known neighbour.
*/
for (l=0; (l<list_s_sz) && (m->va < 0); ++l) {
#ifdef MEMCHECK
checkidx(list_s_r[l],tops+1,"build state 3.3");
#endif
t = mc + list_s_r[l];
if ((intersize(m->t,m->b,t->t,t->b,NULL,NULL) > 0) &&
((t->tc == CHAR) || (t->tc == SCHAR))) {
/* obtain complete alignment data */
dd = complete_align(k,list_s_r[l],&as,&xh,&bl,&ds);
/* infer the alignment for k */
if (dd > 0)
m->va = geo_align(k,dd,as,xh,bl,ds);
}
}
}
/* prepare next symbol */
if ((++i % DPROG) == 0)
show_hint(0,"computing alignment %d/%d",i,topps);
}
/* prepare next state */
else {
show_hint(0,"validating");
st = 4;
}
}
/*
** STATE 4 **
Validating the recognition
*/
else if (st == 4) {
/* (devel)
4. Validating the recognition
Shape-based recognitions must be validated by special
heuristics. For instance, the left column of a broken
"u" may be recognized as the body of an "i" letter. A
validation heuristic may refuse this recognition for
instance because the dot was not found. These heuristics
are per-alphabet.
*/
for (i=0; i<=topps; ++i) {
recog_validation(ps[i]);
}
/* prepare next state */
show_hint(0,"handling punctuation signs");
st = 5;
}
/*
** STATE 5 **
Creating fake words for punctuation signs
*/
else if (st == 5) {
sdesc *m;
/* (devel)
5. Creating fake words for punctuation signs
To produce a clean output, symbols that do not belong to
any word are not included on the OCR output. So we need
to create fake words for punctuation signs like commas
of final dots.
*/
/*
Build unitary words for baseline-aligned dots and commas
*/
for (i=0; i<=topps; ++i) {
m = mc + (k = ps[i]);
if ((m->sw < 0) &&
((m->tc == DOT) || (m->tc == COMMA))) {
if ((m->va == 22) || (m->va == 23)) {
new_word(k);
}
}
}
/* prepare next state */
show_hint(0,"aligning words");
st = 6;
}
/*
** STATE 6 **
Aligning words
*/
else if (st == 6) {
int a,b,d1,d2,k,l;
/* (devel)
6. Aligning words
Words need to be aligned in order to detect the
page text lines. This is perfomed as follows:
*/
/* Mark the start of words */
for (i=0; i<=tops; ++i)
C_UNSET(mc[i].f,F_ISW);
for (i=0; i<=topw; ++i)
C_SET(mc[word[i].F].f,F_ISW);
/* (devel)
a. Undefine the next-word and previous-word
links for each word. These are links for the
previous and next word within lines. For instance,
on the line "our goal is", "goal" is the next
for "our" and the previous for "is", "our" has
no previous and "is" has no next.
*/
for (i=0; i<=topw; ++i) {
word[i].E = -1;
word[i].W = -1;
}
/* (devel)
b. Distribution of the words on lines. This is just
a matter of computing, for each word, its "next" word.
So for each pair of words, we test if they're "paired"
in the sense of the function w_pair. In affirmative
case, we make the left word point to the right word
as its "next" and the rigth point to the left as its
"previous".
The function w_pair does not test the existence of
intermediary words. So on the line "our goal is" that
function will report pairing between "our" and "is".
So after detecting pairing, our loop also checks if the
detected pairing is eventually "better" than those
already detected.
*/
for (i=0; i<=topw; ++i) {
for (j=i+1; j<=topw; ++j) {
/* ignore empty words */
if ((word[i].F < 0) || (word[j].F < 0))
a = b = -1;
/* i-j or j-i pairing detected */
else if (w_pair(i,j) == 0) {
a = i;
b = j;
}
else if (w_pair(j,i) == 0) {
a = j;
b = i;
}
else
a = b = -1;
/*
Compare the pairings a-b and a-k where k=word[a].E,
that is, a-k is a perhaps non-optimal previously
detected pairing.
*/
if ((a >= 0) && (word[a].E >= 0)) {
k = word[a].E;
d1 = mc[word[b].F].l - mc[word[a].L].r;
d2 = mc[word[k].F].l - mc[word[a].L].r;
/* a-k is preferred */
if (d1 >= d2) {
a = -1;
}
}
else
k = -1;
/*
Compare the pairings a-b and l-b where l=word[b].W,
that is, l-b is a perhaps non-optimal previously
detected pairing.
*/
if ((a >= 0) && (word[b].W >= 0)) {
l = word[b].W;
d1 = mc[word[b].F].l - mc[word[a].L].r;
d2 = mc[word[b].F].l - mc[word[l].L].r;
/* l-b is preferred */
if (d1 >= d2) {
a = -1;
}
}
else
l = -1;
/* pairing a-b is preferred */
if (a >= 0) {
/* break link a-k */
if (k >= 0) {
word[k].W = -1;
}
/* break link l-b */
if (l >= 0) {
word[l].E = -1;
}
/* create link a-b */
word[a].E = b;
word[b].W = a;
}
}
}
/* count lines */
for (topln=-1, k=0; k <= topw; ++k) {
if ((word[k].W < 0) && (word[k].F >=0))
++topln;
}
/* enlarge memory area for lines */
if (lnsz <= topln) {
lnsz = topln + 512;
line = c_realloc(line,sizeof(lndesc)*lnsz,NULL);
}
/* line heads */
for (topln=-1, k=0; k <= topw; ++k) {
if ((word[k].W < 0) && (word[k].F >= 0)) {
line[++topln].f = k;
line[topln].hw = 0;
}
}
/* (devel)
c. Sort the lines. The lines are sorted based on the
comparison performed by the function "cmpln".
*/
{
int *a;
lndesc *l;
a = alloca(sizeof(int)*(topln+1));
l = alloca(sizeof(lndesc)*(topln+1));
for (i=0; i<=topln; ++i)
a[i] = i;
qsf(a,0,topln,0,cmpln);
for (i=0; i<=topln; ++i) {
memcpy(l+i,line+a[i],sizeof(lndesc));
}
memcpy(line,l,(topln+1)*sizeof(lndesc));
}
/* word[i].tl must be the line where the word i was placed */
for (i=0; i<=topw; ++i)
word[i].tl = -1;
for (i=0; i<=topln; ++i)
for (j=line[i].f; j>=0; j=word[j].E)
word[j].tl = i;
/* prepare next state */
show_hint(0,"computing word properties");
st = 7;
}
/*
** STATE 7 **
Computing word properties
*/
else if (st == 7) {
int i;
/* account featured symbols */
acc_f(bw,bh,bp,F_BOLD);
acc_f(iw,ih,ip,F_ITALIC);
/* (devel)
7. Computing word properties
Finally, word properties can be computed once we
have detected the words. Some of these properties are
applied to untransliterated symbols. The properties are:
1. The baseline left and right ordinates.
2. The italic and bold flags.
3. The alphabet.
4. The word bounding box.
All these properties are computed by the
function wprops.
*/
for (i=0; i<=topw; ++i) {
if (word[i].F >= 0)
wprops(i);
}
/* finished */
st = 0;
return(0);
}
/* invalid state */
else {
st = 0;
return(0);
}
/* did not complete */
return(1);
}
|