X-Git-Url: https://git.ucc.asn.au/?p=dja%2Fscandal.git;a=blobdiff_plain;f=scan2pages.sh;h=dbd66eea3ce681a831ce08aaf154e59b384a4b72;hp=ce588f75302475fb85c2c7a217928e0aa2a846c5;hb=88d39ee082b1599229efe259f04579e34a13a395;hpb=5769da90e7cef4fb5fe6ca0538925f74492a860d;ds=sidebyside

diff --git a/scan2pages.sh b/scan2pages.sh
index ce588f7..dbd66ee 100755
--- a/scan2pages.sh
+++ b/scan2pages.sh
@@ -2,66 +2,56 @@
 
 function usage {
 cat >&2 << __EOF__ 
-Usage: $0 [-vfpsmut] [-d depth] pdffile outdir
+Usage: $0 [-vfpsmutb] [-D DPI] [-d depth] [-F format] pdffile outdir
 Convert pdffile - a pdf of scanned facing pages - to a set of images in outdir,
 each with only one page.
 OPTIONS:
 	-v: Be verbose.
-	-d depth: Use given depth. Default is 1. Unpaper can only handle up to 8.
-	-s: skip masking/trimming. Overrides -m.
-	-f: Forceably redo everything.
-	-p: Forceably redo pdf conversion. Implies options below, equivalent to -f.
-	-m: Forceably redo masking/trimming and other preprocessing. Impiles options below.
-	-u: Forceably redo unpaper processing. Implies option below. IGNORED
-	-t: Forceably redo final trimming and cleaning. IGNORED
+	-d depth: Use given depth. Default is 1 for black and white, 8 for colour/gray.
+		Unpaper can only handle up to 8.
+	-D dpi: Use given DPI. Colour scans on library scanners are usually 300, bw 400.
+	-e extension: use the output format signified by the extension. Default is png
+	-b: Optimize for BeBook One.
+	-1: each physical page contains one logical page
+	-2: each physical page contains two logical pages
+		(neither specified: try to figure it out)
 __EOF__
 }
 
-## Setup and utilities ##
-# Unpaper
-UNPAPER_PATH="/home/wheel/dja/public-html/scandal/unpaper-0.3/bin"
-PATH=$UNPAPER_PATH:$PATH
-
-# Mac OS X doesn't have seq. It has jot instead.
-Linux_seq="seq"
-Darwin_seq="jot -"
-if [[ $(uname) == "Darwin" ]]; then my_seq=$Darwin_seq;
-else my_seq=$Linux_seq;
-fi;
 
+## Setup and utilities ##
+source $(dirname $0)/my_seq.sh
 
 ## Process flags ##
-forcepdf=
-forcemask=
-forceunpaper=
-forceclean=
 convertflags=
 unpaperflags=
-depthflags="-depth 1"
+depthset=
+depth=
+dpi=
 skipmask=
-while getopts 'vd:sfpmuc' OPTION
+verbose=
+extension="png"
+bebook=
+logperphys=
+while getopts 'vd:D:e:b12' OPTION
 do
 	case $OPTION in
-	v)  convertflags="$convertflags -verbose"
+	v)	verbose=1
+		convertflags="$convertflags -verbose"
 		unpaperflags="$unpaperflags -v --time"
 		;;
-	d)	depthflags="-depth $OPTARG"
-		;;
-	s)	skipmask=1;
+	d)	depth="$OPTARG"
 		;;
-	f)	forcepdf=1; forcemask=1; forceunpaper=1; forceclean=1
+	D)	dpi=$OPTARG
 		;;
-	p)	forcepdf=1; forcemask=1; forceunpaper=1; forceclean=1
+	e)	extension="$OPTARG"
 		;;
-	m)	forcemask=1; forceunpaper=1; forceclean=1
+	b)	bebook=1
 		;;
-	u)	forceunpaper=1; forceclean=1
+	1)	logperphys=1
 		;;
-	t)	forceclean=1
+	2)	logperphys=2
 		;;
-#	b)	bflag=1
-#		bval="$OPTARG"
-#		;;
 	?)	usage
 		exit 2
 		;;
@@ -76,63 +66,159 @@ if [[ $# != 2 ]]; then
 	exit 2
 fi
 
-file=$1
-dir=$2
+scanfile=$1
+outdir=$2
 
 filedir=$(dirname $1)
 base=$(basename $1 .pdf)
 
 # make the output dir
-mkdir -p $dir
+mkdir -p $outdir
+
+
+# guess at DPI, unless it's been set
+# the scans do not seem to accurately reflect their resolution.
+if [[ "$dpi" = "" ]]; then
+	# %k gives the number of colours/shades in the image.
+	# 2 for b/w, and the scanner seems to default to 400 dpi for that
+	# 256 for greyscale, which has to be selected manually (so I'm assuming 300 for no good reason)
+	# 31564 for colour, which defaults to 300
+	if [[ $( convert $scanfile[0] -format "%k" info: || exit 1 ) == 2 ]]; then
+		dpi=400;
+	else
+		dpi=300
+	fi
+	if [[ $verbose == 1 ]]; then
+		echo "DPI guessed at $dpi."
+	fi
+fi
+
+# guess at depth, unless it's been set
+if [[ "$depth" = "" ]]; then
+	if [[ $( convert $scanfile[0] -format "%k" info: || exit 1 ) == 2 ]]; then
+		depth=1;
+	else
+		depth=8
+	fi
+	if [[ $verbose == 1 ]]; then
+		echo "Depth guessed at $depth."
+	fi
+fi
 
 # figure out the number of pages
-dscname=$dir/${base}.dsc
-pdf2dsc $file $dscname || exit 1
+dscname=$outdir/${base}.dsc
+pdf2dsc $scanfile $dscname || exit 1
 pages=$(awk '$1 ~ "%%Pages" {print $2}' $dscname)
 echo "Got $pages page(s)."
 rm $dscname
 
 # process pages 1 by 1 to avoid convert gobbling all the memory
-for pg in `$my_seq 1 $pages`; do
-	echo "Processing page $pg."
+for scanpgnum in `$my_seq 1 $pages`; do
+	echo "Processing page $scanpgnum."
 	
-	pgn=$(printf '%03d' $pg)
+	scanpgnum=$(printf '%03d' $scanpgnum)
 	
 	# convert from pdf
-	origpnm=$dir/pg-${pgn}.pnm
-	if [[ ! $([ -e $origpnm ]) || $forcepdf ]]; then
-		convert $convertflags $depthflags -density 300 $file[$(expr $pg - 1)] \
-			$origpnm || exit 1
+	scanpg=$outdir/scanpg-${scanpgnum}.png
+	if [ ! -e $scanpg ]; then
+		convert $convertflags -depth $depth -density $dpi $scanfile[$(expr $scanpgnum - 1)] \
+			$scanpg || exit 1
 	fi;
 	
-	# preprocess it!
-	preppnm=$dir/pg-pp-${pgn}.pnm
-	if [[ ! $skipmask && ( ! $([ -e $preppnm ]) || $forcemask ) ]]; then
+done;
+
+# do ocr binarise
+[ -e $outdir/scanpgs ] && rm -r $outdir/scanpgs
+ocropus book2pages $outdir/scanpgs $outdir/scanpg-*.png || exit 1
+
+for scanpgnum in  `$my_seq 1 $pages`; do
+
+	binscanpg=$outdir/scanpgs/$(printf '%04d' $scanpgnum).bin.png
+	scanpgnum=$(printf '%03d' $scanpgnum)
+	scanpg=$outdir/scanpg-${scanpgnum}.png
+
+	# preprocess scanned page
+	cleanscanpg=$outdir/scanpg-clean-${scanpgnum}.pnm
+	if [ ! -e $cleanscanpg ]; then
 		# create mask: 
 		# ... downscale, blur,
-		convert $convertflags -resize 25% -depth 8 -blur 10 $origpnm $dir/pg-mask-${pgn}.pnm ||exit 1
+		convert $convertflags -resize 10% -depth 8 -blur 10 -median 2 $scanpg $outdir/scanpg-mask-${scanpgnum}.png ||exit 1
 
-		# ... get crop co-ords
-		cropcords=$(convert -resize 400% -trim -fuzz 90% -format "%wx%h%O" $dir/pg-mask-${pgn}.pnm info: || exit 1)
+		# ... get crop co-ords. They're off by ~2 as I don't know how to
+		# properly correct for the border.
+		cropcords=$(convert -border 1x1 -bordercolor '#000' -resize 1000% -trim -fuzz 90% -format "%wx%h%O" $outdir/scanpg-mask-${scanpgnum}.png info: || exit 1)
 		
 		# ... crop and despeckle? the final pre-prepared image
-		convert $convertflags -crop $cropcords $origpnm $preppnm || exit 1
-	elif [[ $skipmask ]]; then
-		cp $origpnm $preppnm
+		convert $convertflags -crop $cropcords $binscanpg  $cleanscanpg || exit 1
 	fi;
 	
-	#unpaper it
-	#names go a bit funny here
-	#also, ignore flags starting here
-	unppnm=$dir/upg-${pgn}-%01d.pnm
-	unpaper $unpaperflags --layout double --overwrite --no-blackfilter -ni 10 -op 2 $preppnm $unppnm || exit 1
-
-	# final convert and clean
-	convert $convertflags $dir/upg-${pgn}-?.pnm $dir/upg-${pgn}-%01d.tiff || exit 1
+	# check it hasn't mostly disappeared - e.g. if the scan was all black
+	# (e.g. forgot to put the book down when you first hit scan)
+	if [[ $(convert  $cleanscanpg -format '%[fx:s.w*s.h>1000]' info:) = "0" ]]; then
+		[[ $verbose == 1 ]] && echo "Discarding pg ${scanpgnum}: not enough remains after masking."
+		continue; 
+	fi;
 	
+	#unpaper it
+	physpgbase=$outdir/physpg-${scanpgnum}
+	if [ ! -e ${physpgbase}-1.pnm ] || [ ! -e ${physpgbase}-2.pnm ]; then
+		unpaper $unpaperflags --layout double --overwrite -ni 10 -op 2 $cleanscanpg ${physpgbase}-%01d.pnm || exit 1
+	fi;
 	
+	for physpgnum in $($my_seq 1 2); do
+		echo Processing physical page ${physpgnum}.
+		
+		physpg=${physpgbase}-${physpgnum}.pnm #$physpgbase has $outdir/ already
+		
+		cleanphyspg=$outdir/physpg-clean-${scanpgnum}-${physpgnum}.pnm
+		
+		if [ ! -e $cleanphyspg ]; then
+			# remask and retrim. FIXME: may throw off 2page detection
+			convert $convertflags -resize 10% -depth 8 -gamma 0.01 -median 2 $physpg $outdir/physpg-mask-${scanpgnum}-${physpgnum}.png ||exit 1
+
+			# Trim #-border 1x1 -bordercolor '#fff' -trim -fuzz 30% 
+			cropcords=$(convert -trim -fuzz 90%\
+			 -resize 1000% -format "%wx%h%O" $outdir/physpg-mask-${scanpgnum}-${physpgnum}.png info: || exit 1)
+		
+			# ... crop and despeckle? the final pre-prepared image
+			convert $convertflags -crop $cropcords $physpg $cleanphyspg || exit 1
+		fi;
+		
+		#detect if the page is 2-up
+
+		if [[ $logperphys == 2 ]] || ( [[ $logperphys != 1 ]] && $(dirname $0)/detect2pages.sh ${cleanphyspg} ${scanpgnum} ${physpgnum} ); then
+			if [[ $verbose == 1 ]]; then
+				echo "Resplitting physical page ${physpgnum}."
+			fi
+			unpaper $unpaperflags --pre-rotate 90 --layout double --overwrite \
+			 -op 2 --no-blackfilter --no-grayfilter --no-noisefilter \
+			 --no-blurfilter $cleanphyspg $outdir/logpg-${scanpgnum}-${physpgnum}-%01d.pnm || exit 1
+		else
+			cp $cleanphyspg $outdir/logpg-${scanpgnum}-${physpgnum}-1.pnm
+		fi;
+
+		#prepare for ocr
+		convert $convertflags $outdir/logpg-${scanpgnum}-${physpgnum}-?.pnm $outdir/logpg-${scanpgnum}-${physpgnum}-%01d.png || exit 1
+
+		#final convert and clean w/ bebook optimisation
+		if [[ $bebook ]]; then
+			convert $convertflags $outdir/logpg-${scanpgnum}-${physpgnum}-?.pnm -trim -fuzz 80% -resize 1200x1600 $outdir/final-${scanpgnum}-${physpgnum}-%01d.${extension} || exit 1
+		else
+			convert $convertflags $outdir/logpg-${scanpgnum}-${physpgnum}-?.pnm $outdir/final-${scanpgnum}-${physpgnum}-%01d.${extension} || exit 1
+		fi
+
+	done;
+
 done
 
-mkdir $dir/pages
-mv $dir/upg-*.tiff $dir/pages
+#try full ocr
+rm -rf $outdir/logpgs
+ocropus book2pages $outdir/logpgs $outdir/logpg-*.png
+ocropus pages2lines $outdir/logpgs
+ocropus lines2fsts $outdir/logpgs
+ocropus fsts2bestpaths $outdir/logpgs
+ocropus buildhtml $outdir/logpgs > $outdir/out.html
+
+mkdir -p $outdir/pages
+mv $outdir/final-*.${extension} $outdir/pages