X-Git-Url: https://git.ucc.asn.au/?p=dja%2Fscandal.git;a=blobdiff_plain;f=scan2pages.sh;h=cbfe52f06bd4af008225c65a3426c9e3b0527f49;hp=42b3bfbf8a6d091394e881876c35f2f98767abe1;hb=88f2331a5a94726bdf7c2b28f11c203a3f9ae15a;hpb=aeb663e7da50cdad5b60dcb5d7a1f78fc83f36b4

diff --git a/scan2pages.sh b/scan2pages.sh
index 42b3bfb..cbfe52f 100755
--- a/scan2pages.sh
+++ b/scan2pages.sh
@@ -128,14 +128,23 @@ for scanpgnum in `$my_seq 1 $pages`; do
 done;
 
 # do ocr binarise
-[ -e $outdir/scanpgs ] && rm -r $outdir/scanpgs
-ocropus book2pages $outdir/scanpgs $outdir/scanpg-*.png || exit 1
+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-${scanpgnum}.png
-	binscanpg=$outdir/scanpgs/$(printf '%04d' $scanpgnum).bin.png
 
 	# preprocess scanned page
 	cleanscanpg=$outdir/scanpg-clean-${scanpgnum}.pnm
@@ -162,7 +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
@@ -197,18 +210,28 @@ 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
+			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
 
+#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