X-Git-Url: https://git.ucc.asn.au/?p=dja%2Fscandal.git;a=blobdiff_plain;f=scan2pages.sh;fp=scan2pages.sh;h=5fbdab5ffdeba68b5684bdad7a60c59fb525574d;hp=b2bad107ea4837582e6ce27de538b75929c0a568;hb=8fe732e83a703d38e483d0925f0c3bc1b2683910;hpb=2312058a395f4cedf83c5c19434a9c9b400c9e56

diff --git a/scan2pages.sh b/scan2pages.sh
index b2bad10..5fbdab5 100755
--- a/scan2pages.sh
+++ b/scan2pages.sh
@@ -125,7 +125,28 @@ for scanpgnum in `$my_seq 1 $pages`; do
 			$scanpg || exit 1
 	fi;
 	
-	# preprocess it!
+done;
+
+# do ocr binarise
+if [ -e $outdir/scanpgs ]; then
+	#assume (for the sake of speed in repeated runs) that
+	# if the last file exists, the process completed successfully
+	if [ -e $outdir/scanpgs/$(printf '%04d' $pages).bin.png ]; then
+		[[ $verbose ]] && echo Binarisation already complete
+	else
+		rm -r $outdir/scanpgs
+	fi
+fi
+[ -e $outdir/scanpgs ] || ocropus book2pages $outdir/scanpgs $outdir/scanpg-*.png || exit 1
+
+# process each binarised scan page
+for scanpgnum in  `$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 $cleanscanpg ]; then
 		# create mask: 
@@ -135,11 +156,9 @@ for scanpgnum in `$my_seq 1 $pages`; do
 		# ... get crop co-ords. They're off by ~2 as I don't know how to
 		# properly correct for the border.
 		cropcords=$(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 $scanpg $cleanscanpg || exit 1
-	elif [[ $skipmask ]]; then
-		cp $origfile $preppnm
+		convert $convertflags -crop $cropcords $binscanpg  $cleanscanpg || exit 1
 	fi;
 	
 	# check it hasn't mostly disappeared - e.g. if the scan was all black
@@ -152,7 +171,11 @@ for scanpgnum in `$my_seq 1 $pages`; do
 	#unpaper it
 	physpgbase=$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
+		# most of unpaper's processing is redundant given ocropus, and somewhat too agressive
+		unpaper $unpaperflags --layout double --overwrite -op 2 \
+			--no-processing \
+			$cleanscanpg ${physpgbase}-%01d.pnm || exit 1
+
 	fi;
 	
 	for physpgnum in $($my_seq 1 2); do
@@ -167,13 +190,22 @@ for scanpgnum in `$my_seq 1 $pages`; do
 			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=$(convert -trim -fuzz 90%\
+			# binarise is so effective, try something << 90%
+			cropcords=$(convert -trim -fuzz 50%\
 			 -resize 1000% -format "%wx%h%O" $outdir/physpg-mask-${scanpgnum}-${physpgnum}.png info: || exit 1)
 		
+			[[ $verbose ]] && echo Crop co-ords: $cropcords
+
 			# ... crop and despeckle? the final pre-prepared image
 			convert $convertflags -crop $cropcords $physpg $cleanphyspg || exit 1
 		fi;
-		
+
+		# check it hasn't mostly disappeared, warn viciously if it has!
+		if [[ $(convert  $cleanphyspg -format '%[fx:s.w*s.h>1000]' info:) = "0" ]]; then
+			echo "Warning: discarding physical pg ${scanpgnum}-${physpgnum}: not enough remains after masking."
+			continue;
+		fi;
+
 		#detect if the page is 2-up
 
 		if [[ $logperphys == 2 ]] || ( [[ $logperphys != 1 ]] && $(dirname $0)/detect2pages.sh ${cleanphyspg} ${scanpgnum} ${physpgnum} ); then
@@ -187,18 +219,28 @@ for scanpgnum in `$my_seq 1 $pages`; do
 			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 -colorspace Gray -median 1 $outdir/logpg-${scanpgnum}-${physpgnum}-?.pnm -trim -fuzz 80% -resize 1200x1600 $outdir/final-${scanpgnum}-${physpgnum}-%01d.${extension} || exit 1
+		if [[ $bebook ]]; then #1200x1600
+			convert $convertflags $outdir/logpg-${scanpgnum}-${physpgnum}-?.pnm -trim $outdir/final-${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
 
+#don't try full ocr, it's a waste of time.
+#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