/* epstopdf-fixjpg.c -- Heiko Purnhagen (purnhage@tnt.uni-hannover.de) HP 20020825 20020826 20020827 20020830 20020902 20020903 20020921 Background: "get smaller PostSript/PDF files by using JPEG-compressed images" PostScript/EPS (Level 2 or higher) and PDF allow to include grayscale and color images in JPEG format, using the so-called DCTDecode filter. ImageMagick's convert tool can be used to convert image.jpg to image.eps: convert image.jpg eps2:image.eps If you want to avoid JPEG->JPEG transcoding, you could use a tool like jpeg2ps, which basically puts a wrapper around the orignal JPEG file: jpeg2ps image.jpg > image.eps However, jpeg2ps's binary mode (-b) seems to produce EPS files that are hardly includable in larger PostScript files. Furthermore, the EPS files generated by jpeg2ps don't work with the following epstopdf-fixjpg tool. Hence, I hacked a tiny AWK script, jpg2eps, that "wraps" the original JPEG file, generating a EPS file with the same structure as one generated by convert: jpg2eps image.jpg > image.eps For non-JPEG images, use: convert -compress JPEG -quality 75 image.gif eps2:image.eps Unfortunately, the ps/eps export doesn't work correct in (quite) some versions of ImageMagick, but at least ImageMagick-5.4.4-5.tar.gz seems to be OK. Conversion of such an image.eps to image.pdf, for example by the epstopdf script (which should come with any decent pdflatex distribution), is a bit more troublesome. epstopdf --outfile=image.pdf image.eps Depending on your version of ghostscript, you can get - large files using LZWDecode (gs 6.0) - small files using DCTDecode, but transcoded (gs 6.5, gs 7.0) While it is nice that the pdfwrite output device of ghostscript now produces compressed images by default, it sometime might be good to avoid transcoding for JPEG images. Hence, I wrote epstopdf-fixjpg as a quick hack to replace the images in image.pdf as generated by epstopdf by the "original" DCTDecode data in image.eps. This tool can also work for eps files that include more than one DCTDecode images, e.g. a figure exported from xfig containing several DCTDecode EPS images. Usage: gcc -o epstopdf-fixjpg epstopdf-fixjpg.c convert image.eps to image.pdf: epstopdf -o=image_tmp.pdf image.eps epstopdf-fixjpg image_tmp.pdf image.eps > image.pdf replace images in image_tmp.pdf by (concatenated) JPEGs in image.jpg epstopdf-fixjpg image_tmp.pdf image.jpg > image.pdf just re-generate xref table in PDF file: epstopdf-fixjpg image_tmp.pdf > image.pdf Links: http://www.tnt.uni-hannover.de/~purnhage/software/epstopdf-fixjpg.c http://www.tnt.uni-hannover.de/~purnhage/software/jpg2eps http://www.imagemagick.org/www/convert.html http://www.cs.wisc.edu/~ghost/ http://www-personal.umich.edu/~wsherman/eps_img/ http://www.pdflib.com/jpeg2ps/index.html http://www.tug.org/applications/pdftex/epstopdf http://www.xfig.org/ http://partners.adobe.com/asn/developer/acrosdk/DOCS/pdfspec.pdf */ #include #include #define MAXOBJ 1000 #define STRLEN 16384 #define KEYLEN 128 int getline(char *s, int *ic, int *oc, FILE *i, FILE *o) { int n=0; int t; char *ss=s; while ((t=getc(i))!=EOF) { *ss++ = t; n++; (*ic)++; if (t=='\n') return n; if (n==STRLEN-KEYLEN) { if (o) { fwrite(s,1,STRLEN-KEYLEN,o); (*oc) += STRLEN-KEYLEN; } memmove(s,s+STRLEN-KEYLEN,KEYLEN); ss -= STRLEN-KEYLEN; n -= STRLEN-KEYLEN; } } return n; } int extractjpg(char *s, int epsisjpg, int numjpg, FILE *i, int *oc, FILE *o) { int n; int t,t1,t2; rewind(i); if (epsisjpg) { t = getc(i); while (t!=EOF) { if (t!=0xff) t = getc(i); else if ((t=getc(i))==0xd8) { if (--numjpg == -1) { n = 2; if (o) { putc(0xff,o); putc(0xd8,o); *oc += 2; } t1 = 0xff; t2 = 0xd8; t = getc(i); while (t!=EOF) { n++; if (t!=0xff) { if (o) { putc(t,o); (*oc)++; } t1 = t2; t2 = t; t = getc(i); } else { if ((t=getc(i))==0xd8 && t1==0xff && t2==0xd9) { if (o) { putc('\n',o); (*oc)++; } return n-1; } else { if (o) { putc(0xff,o); (*oc)++; } t1 = t2; t2 = 0xff; } } } if (!n) { fprintf(stderr,"epstopdf-fixjpg: jpg parse error\n"); exit (-1); } if (o) { putc('\n',o); (*oc)++; } return n; } } } } else while ((t=getc(i))!=EOF) { /* the following line might fail sometimes ... */ if (t=='\n' && getc(i)==0xff && getc(i)==0xd8) { if (--numjpg == -1) { n = 2; if (o) { putc(0xff,o); putc(0xd8,o); *oc += 2; } while (t=getline(s,&n,oc,i,o)) { if (!strncmp(s,"%%EndData\n",t)) return n-10-1; if (o) { fwrite(s,1,t,o); *oc += t; } } fprintf(stderr,"epstopdf-fixjpg: eps parse error\n"); exit (-1); } } } return 0; } int main (int argc, char *argv[]) { FILE *pdff,*epsf,*outf; int pdfc,epsc,outc; int numjpg; int numobj,maxobj; int objoff[MAXOBJ+1]; char pdfs[STRLEN+1]; char epss[STRLEN+1]; char outs[STRLEN+1]; int epsisjpg; int n,obj,i; char *s; int ti; char ts[5]; if (argc!=2 && argc!=3) { fprintf(stderr, "epstopdf-fixjpg (HP20020921)\n" "usage: epstopdf-fixjpg image_tmp.pdf image.eps > image.pdf\n" " or: epstopdf-fixjpg image_tmp.pdf image.jpg > image.pdf\n" "fix xref: epstopdf-fixjpg image_tmp.pdf > image.pdf\n"); exit (-1); } if ((pdff=fopen(argv[1],"r"))==NULL) { fprintf(stderr, "epstopdf-fixjpg: can not open pdf \"%s\"\n",argv[1]); exit (-1); } if (argc==2) epsf = NULL; else if ((epsf=fopen(argv[2],"r"))==NULL) { fprintf(stderr, "epstopdf-fixjpg: can not open eps/jpg \"%s\"\n",argv[2]); exit (-1); } if (epsf) epsisjpg = (getc(epsf)==0xff && getc(epsf)==0xd8); outf = stdout; pdfc = epsc = outc = 0; numjpg = 0; maxobj = numobj = 0; *epss = '\0'; while (n=getline(pdfs,&pdfc,&outc,pdff,outf)) { *(pdfs+n) = '\0'; if (strstr(pdfs," 0 obj") && sscanf(pdfs,"%d %d %4s",&obj,&ti,ts)==3 && ti==0 && (!strcmp(ts,"obj") || !strcmp(ts,"obj "))) { if (obj>=MAXOBJ) { fprintf(stderr,"epstopdf-fixjpg: obj error\n"); exit (-1); } numobj++; if (maxobj0) { n = getline(pdfs,&pdfc,NULL,pdff,NULL); s = pdfs; } n = extractjpg(epss,epsisjpg,numjpg,epsf,NULL,NULL); fprintf(stderr,"epstopdf-fixjpg: jpg%d: %d\n",numjpg,n); if (!n) { fprintf(stderr,"epstopdf-fixjpg: no jpg found\n"); exit (-1); } sprintf(outs,"/Filter/DCTDecode/Length %d>>stream\n",n); fprintf(outf,"%s",outs); outc += strlen(outs); extractjpg(epss,epsisjpg,numjpg,epsf,&outc,outf); do { n = getline(pdfs,&pdfc,NULL,pdff,NULL); *(pdfs+n) = '\0'; } while (strcmp(pdfs,"endstream\n")!=0 && n>0); fprintf(outf,"%s",pdfs); outc += strlen(pdfs); n = 0; numjpg++; *epss = '\0'; } else if (!strcmp(pdfs,"xref\n")) { ti = 0; objoff[0] = 0; for (i=maxobj; i>=0; i--) if (!objoff[i]) { objoff[i] = -ti; ti = i; } sprintf(outs,"xref\n0 %d\n%010d 65535 f \n",maxobj+1,-objoff[0]); objoff[0] = outc; fprintf(outf,"%s",outs); outc += strlen(outs); for (i=1; i<=maxobj; i++) if (objoff[i]>0) fprintf(outf,"%010d 00000 n \n",objoff[i]); else fprintf(outf,"%010d 00000 f \n",-objoff[i]); outc += numobj*20; for (i=0; i6 string JFIF \b, JFIF standard >6 string Exif \b, EXIF standard od -N 10 -t x1 -a image.jpg 0000000 ff d8 ff e0 00 10 4a 46 49 46 del X del ` nul dle J F I F 0000000 ff d8 ff e1 88 45 45 78 69 66 del X del a bs E E x i f od -t x1 image.jpg | tail -2 0141660 3e ba f2 28 8c ff d9 0141667 eps - convert 5.4.4: userdict begin %%BeginData: 24154 BINARY Bytes DisplayImage 0 0 438 297 12.000000 438 297 1 0 1 0 8 ÿØ*** %%EndData pdf - gs 6.0: <> stream ÿØ*** endstream pdf - gs 7.0: /ColorSpace/DeviceGray /Width 438 /Height 297 /BitsPerComponent 8 /Filter/DCTDecode/Length 21152>>stream ÿØ*** endstream grep -ab '^[0-9]\+ [0-9] obj' a75gc750.pdf 15:5 0 obj 225:6 0 obj 244:4 0 obj 415:3 0 obj 474:1 0 obj 522:7 0 obj 591:8 0 obj 21891:9 0 obj 45408:10 0 obj 45438:11 0 obj 45477:2 0 obj grep -ab '^xref$' a75gc750.pdf 45638:xref tail -25 a75gc750.pdf 2 0 obj <>endobj xref 0 12 0000000000 65535 f 0000000474 00000 n 0000045477 00000 n 0000000415 00000 n 0000000244 00000 n 0000000015 00000 n 0000000225 00000 n 0000000522 00000 n 0000000591 00000 n 0000021891 00000 n 0000045408 00000 n 0000045438 00000 n trailer << /Size 12 /Root 1 0 R /Info 2 0 R >> startxref 45638 %%EOF */