X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?p=ypp-sc-tools.db-test.git;a=blobdiff_plain;f=pctb%2Focr.c;h=fe264e8110ce8ffce6cbb6be0282c004151dec31;hp=bc8057725291a482fb7a31a9cf0428112e6442ad;hb=cf8e9b2a56d12305cd61f8d4a9284d4319265215;hpb=83b1dc4a5d774abe3c0c4a89c67f78b7c5974c51 diff --git a/pctb/ocr.c b/pctb/ocr.c index bc80577..fe264e8 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -40,6 +40,18 @@ typedef struct DatabaseNode { DatabaseLink *links; } DatabaseNode; +typedef struct { + OcrReader *rd; + int w; + Pixcol *cols; + int x; + unsigned ctxmap; +} FindCharArgs; +typedef struct { + DatabaseNode *match; + int rx; +} FindCharResults; + static const char *context_names[]= { "Lower", "Upper", @@ -275,17 +287,8 @@ static void add_result(OcrReader *rd, const char *s, int l, int r, } -const char *ocr_celltype_name(OcrCellType ct) { return ct->name; } - -typedef struct { - OcrReader *rd; - int w; - Pixcol *cols; - int x; -} FindCharArgs; - -static DatabaseNode *findchar(const FindCharArgs *fca, - DatabaseNode *start, int *matchx_r) { +static DatabaseNode *findchar_1ctx(const FindCharArgs *fca, + DatabaseNode *start, int *matchx_r) { DatabaseNode *current= start; DatabaseNode *bestmatch= 0; int i; @@ -319,15 +322,51 @@ static DatabaseNode *findchar(const FindCharArgs *fca, return bestmatch; } +static DatabaseNode *findchar(const FindCharArgs *fca, int *match_rx) { + FindCharResults results[NCONTEXTS]; + int ctxi, match=-1, nmatches=0; + + debugf("OCR lx=%d ct_state=%x ", fca->x, fca->ctxmap); + for (ctxi=0; ctxictxmap & (1u << ctxi))) continue; + debugf(" || %s",context_names[ctxi]); + + results[ctxi].match= findchar_1ctx(fca, &fca->rd->contexts[ctxi], + &results[ctxi].rx); + if (!results[ctxi].match) continue; + + match= ctxi; + nmatches++; + } + if (nmatches==1) { + debugf( " unambiguous"); + } else { + match=-1; + } + if (match<0) + return 0; + + *match_rx= results[ctxi].rx; + return results[ctxi].match; +} + +const char *ocr_celltype_name(OcrCellType ct) { return ct->name; } + OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { int nspaces; - unsigned ctxmap; - int ctxi, x; + int x; + + FindCharArgs fca; + fca.rd= rd; + fca.w= w; + fca.cols= cols; + fca.x= -1; restart: nspaces=- w; - ctxmap= ct->initial; + fca.ctxmap= ct->initial; rd->nresults=0; debugf("OCR h=%d w=%d",rd->h,w); for (x=0; xspace_spaces) { debugf("OCR x=%x nspaces=%d space\n",x,nspaces); - ctxmap= ct->nextword; + fca.ctxmap= ct->nextword; } continue; } @@ -356,52 +395,30 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { add_result(rd," ",x-nspaces,x+1,0); nspaces=0; - FindCharArgs fca; - fca.rd= rd; - fca.w= w; - fca.cols= cols; fca.x= x; - - DatabaseNode *uniquematch= 0; - int uniquematch_rx=-1; - - debugf("OCR lx=%d ctxmap=%x ",x,ctxmap); - - for (ctxi=0; ctxicontexts[ctxi], &uniquematch_rx); - if (!match) continue; - - if (uniquematch && strcmp(match->str, uniquematch->str)) { - debugf( " ambiguous"); - uniquematch= 0; - } else { - uniquematch= match; - } - } - - if (uniquematch) { + int match_rx=-1; + DatabaseNode *match= findchar(&fca, &match_rx); + + if (match) { debugf(" || YES"); - add_result(rd, uniquematch->str, x, uniquematch_rx, ctxmap); - x= uniquematch_rx+1; - if (uniquematch->match) ctxmap= ct->midword; + add_result(rd, match->str, x, match_rx, fca.ctxmap); + x= match_rx+1; + if (match->match) fca.ctxmap= ct->midword; else debugf(" (empty)"); - if (uniquematch->endsword) { + if (match->endsword) { nspaces= ct->space_spaces; debugf("_"); - ctxmap= ct->nextword; + fca.ctxmap= ct->nextword; } debugf("\n"); } else { int rx; debugf(" || UNKNOWN"); for (rx=x; rx