From fca6360e428cfd63c0d8f9fc30b83d27a611bf4a Mon Sep 17 00:00:00 2001
From: Nick White <git@njw.me.uk>
Date: Mon, 7 Nov 2011 20:09:55 +0000
Subject: Dramatically improve mkocrpdf script

---
 extras/mkocrpdf.sh | 19 +++++++++++--------
 extras/mkpdf.sh    |  1 +
 2 files changed, 12 insertions(+), 8 deletions(-)

(limited to 'extras')

diff --git a/extras/mkocrpdf.sh b/extras/mkocrpdf.sh
index 165059a..3a01559 100644
--- a/extras/mkocrpdf.sh
+++ b/extras/mkocrpdf.sh
@@ -1,22 +1,25 @@
 #!/bin/sh
+# See COPYING file for copyright and license details.
 #
 # Makes a PDF with embedded text extracted by tesseract
-#
-# Requires imagemagick, pdftk, hocr2pdf and tesseract
-#
-# Also requires this tesseract configuration:
-# echo 'tessedit_create_hocr 1' > /usr/local/share/tessdata/configs/hocr
+# Requires imagemagick, pdftk, hocr2pdf and tesseract 3
+
+echo 'tessedit_create_hocr 1' > hocr
 
 for i in `ls *png`
 do
 	a=`basename $i .png`
 	echo processing $a
 
-	tesseract $i $a hocr 2>/dev/null
+	# unfortunately tesseract seems to work much better with a
+	# resized larger image
+	convert $i -geometry 1000x $a.big.png
+	tesseract $a.big.png $a hocr 2>/dev/null
 
 	# hocr2pdf has a habit of segfaulting, so fall back to convert
-	hocr2pdf -i $i -o $a.pdf < $a.html || convert $i $a.pdf
-	rm -f $a.html
+	hocr2pdf -i $a.big.png -o $a.pdf < $a.html || convert $a.big.png $a.pdf
+	rm -f $a.html $a.big.png
 done
 
 pdftk *pdf cat output book.pdf
+rm -f [0-9]*pdf hocr
diff --git a/extras/mkpdf.sh b/extras/mkpdf.sh
index 80e776f..cbfcd1b 100755
--- a/extras/mkpdf.sh
+++ b/extras/mkpdf.sh
@@ -1,5 +1,6 @@
 #!/bin/sh
 # See COPYING file for copyright and license details.
+#
 # Requires imagemagick
 
 convert *.png book.pdf
-- 
cgit v1.2.3