Help:DjVu files/OCR with Tesseract

From Wikisource
Jump to: navigation, search
#!/usr/bin/perl

$lang = "eng";

if ( $#ARGV < 0 ) {
  print "give a DJVU file as 1st argument \n" ;
  exit 0 ;
}

$inputdjvu = $ARGV[0] ;
$imagetmp = "/tmp/temp.tif" ;
$djvutmp = "/tmp/outdjvu" ;

print "processing of $inputdjvu\n" ;

# calculate the number of pages
$nbpages = `djvused "$inputdjvu" -e 'n'` ;
print "number of pages: $nbpages" ;

for ($i=1 ; $i <= $nbpages ; $i++) {
  print "OCR de la page $i\n" ;

  # page extraction as an image
  `ddjvu -format=tiff -mode=black -page="$i" "$inputdjvu" $imagetmp` ;

  `tesseract $imagetmp /tmp/outocr -l $lang` ;
  print "OCR done\n" ;

  open(TXT, '/tmp/outocr.txt') ;
  open(TXTDJVU, ">$djvutmp") ;

  print TXTDJVU "(page 0 0 1 1\n" ;
  while ($line=<TXT>) {
    $line =~ s/\"/\\\"/g ;
    print TXTDJVU "(line 0 0 1 1 \"$line\")\n" ;
  }
  print TXTDJVU ")\n" ;

  close (TXT) ;
  close (TXTDJVU) ;

  # writing the text in the DJVU file
  `djvused "$inputdjvu" -e 'select $i; remove-txt' -s` ;
  `djvused "$inputdjvu" -e 'select $i; set-txt $djvutmp' -s` ;
}

# note: structure which works
# print TXTDJVU "(page 0 0 1 1\n" ;
#   print TXTDJVU "     (line 0 0 1 1 \"toto\")\n" ;
#   print TXTDJVU "     (line 0 0 1 1 \"toto la la\")\n";
#   print TXTDJVU ")\n" ;