From e1fada376c4b883d53d6293babb899aa1c510b65 Mon Sep 17 00:00:00 2001
From: Nick White <git@njw.me.uk>
Date: Thu, 2 Feb 2012 00:34:03 +0000
Subject: Add djvu creation scripts

---
 extras/mkocrdjvu.sh | 41 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 41 insertions(+)
 create mode 100644 extras/mkocrdjvu.sh

(limited to 'extras/mkocrdjvu.sh')

diff --git a/extras/mkocrdjvu.sh b/extras/mkocrdjvu.sh
new file mode 100644
index 0000000..c1207f5
--- /dev/null
+++ b/extras/mkocrdjvu.sh
@@ -0,0 +1,41 @@
+#!/bin/sh
+# See COPYING file for copyright and license details.
+#
+# Makes a DjVu with embedded text extracted by tesseract
+# Requires imagemagick, djvulibre and tesseract 3
+#
+# Note that this doesn't use bounding box info, so that text
+# reflows much better.
+
+test $# -ne 1 && echo "Usage: $0 bookdir" && exit 1
+cd "$1" || exit 1
+
+for i in `ls`
+do
+	echo "$i"
+
+	# create djvu compressed version
+	convert "$i" "$i.ppm"
+	c44 "$i.ppm" "$i.djvu"
+
+	# create a much bigger version of the page image, for better
+	# tesseract accuracy
+	width=`identify "$i" |awk '{print $3}'|awk -F x '{print $1}'`
+	height=`identify "$i" |awk '{print $3}'|awk -F x '{print $2}'`
+	bigwidth=`expr $width \* 4`
+	convert "$i" -geometry ${bigwidth}x "$i.tif"
+
+	tesseract "$i.tif" "$i" 2>&1 | sed '/Tesseract Open Source OCR Engine/d'
+
+	# convert tesseract output into djvused input
+	(printf "(page 0 0 $width $height \"";sed 's/"/\\"/g;'"s/\'/\\\'/g" < "$i.txt";printf \"")\n") > "$i.djvutxt"
+	djvused "$i.djvu" -e "select 1; set-txt $i.djvutxt" -s
+
+	rm -f "$i.ppm" "$i.tif" "$i.txt" "$i.djvutxt"
+done
+
+djvm -c book.djvu *.djvu
+
+rm -f [0-9]*djvu
+
+echo "$1/book.djvu"
-- 
cgit v1.2.3