X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?a=blobdiff_plain;f=pctb%2Focr.c;h=f6104452dc1c1854e6ccd19aac77882caae0aba9;hb=ac65228e40fa375c829b46607fb4941ff11376e9;hp=bc8057725291a482fb7a31a9cf0428112e6442ad;hpb=83b1dc4a5d774abe3c0c4a89c67f78b7c5974c51;p=ypp-sc-tools.db-test.git diff --git a/pctb/ocr.c b/pctb/ocr.c index bc80577..f610445 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -40,31 +40,35 @@ typedef struct DatabaseNode { DatabaseLink *links; } DatabaseNode; +typedef struct { + OcrReader *rd; + OcrCellType ct; + int w; + Pixcol *cols; + int x; + unsigned ctxmap; +} FindCharArgs; +typedef struct { + DatabaseNode *match; + int rx; +} FindCharResults; + +enum { ct_Lower, ct_Upper, ct_Word, ct_Digit }; static const char *context_names[]= { - "Lower", - "Upper", - "Digit" + "Lower", /* bit 0, value 001 */ + "Upper", /* bit 1, value 002 */ + "Word", /* bit 2, value 004 */ + "Digit", /* bit 3, value 010 */ }; struct OcrCellTypeInfo { /* bitmaps of indices into context_names: */ unsigned initial, nextword, midword; int space_spaces; const char *name; -}; -const struct OcrCellTypeInfo ocr_celltype_number= { - 4,4,4, - .space_spaces= 5, - .name= "number" -}; -const struct OcrCellTypeInfo ocr_celltype_text= { - .initial=2, /* Uppercase */ - .nextword=3, /* Either */ - .midword=1, /* Lower only */ - .space_spaces= 4, - .name= "text" + int (*findchar_select)(const FindCharArgs *fca, + const FindCharResults results[]); }; - #define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0])) struct OcrReader { @@ -275,17 +279,8 @@ 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 +314,82 @@ 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(" unique"); + } else { + debugf(" ambiguous"); + match= !fca->ct->findchar_select ? -1 : + fca->ct->findchar_select(fca,results); + debugf(" resolved %s", match<0 ? "" : context_names[match]); + } + if (match<0) + return 0; + + *match_rx= results[ctxi].rx; + return results[ctxi].match; +} + +static int findchar_select_text(const FindCharArgs *fca, + const FindCharResults results[]) { + if (fca->ctxmap != 017) return -1; + + dbassert(! results[ct_Digit].match ); + if (results[ct_Word].match) return ct_Word; + if (results[ct_Lower].rx > results[ct_Upper].rx) return ct_Lower; + if (results[ct_Upper].rx > results[ct_Lower].rx) return ct_Upper; + return -1; +} + +const struct OcrCellTypeInfo ocr_celltype_number= { + 010,010,010, + .space_spaces= 5, + .name= "number", + .findchar_select= 0 +}; +const struct OcrCellTypeInfo ocr_celltype_text= { + .initial= 012, /* Digit|Upper */ + .nextword= 017, /* Digit|Upper|Lower|Word */ + .midword= 014, /* Digit|Lower */ + .space_spaces= 4, + .name= "text", + .findchar_select= findchar_select_text +}; + + +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.ct= ct; + 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 +418,30 @@ 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