[SOLVED] Neat image handling wheeze required.

business_kid · 06-11-2014, 11:37 AM

I have a fairly massive pdf (24M) of images which is poorly laid out and clunky to view on an intel GPU. I want to

1. Split the pdf into images
2. Batch process them through tesseract OCR which I have working here. The image quality is good enough that I can expect good OCR

How do I split that pdf into images? anybody done this with gs, or something else. Gimp is not an option speedwise.

Didier Spaier · 06-11-2014, 11:46 AM

I'd use pdfimages, shipped in the poppler package.

business_kid · 06-11-2014, 01:35 PM

That seemed to do it all right. They came out in some funny format. The book has 362 pages (=24 Megs) , but I'm up on image number 1050 & counting when I called halt with Ctrl_C >:-/. I have over 6 gigs there!

Thanks Didier - never knew it was there. It will be funny to see what it did.

Didier Spaier · 06-11-2014, 01:43 PM

Quote:

Originally Posted by business_kid

I have over 6 gigs there!

use the -j option: JPEG files are *much* lighter then PPM or PBM ones.

business_kid · 06-12-2014, 03:49 AM

I did! It requires the dct file base to be in the pdf, which it wasn't.

For each page I ended up with inverse text, .pbm white text on black, @~1Meg each, some background colour for each page, .ppm, @~2.7 Megs each and some rendered looking apparition based loosely on the text (Like the text melting and falling down), .ppm, @24 Megs each. I could only keep sanity in the home dir by running

rm *.ppm which left the pbms. Now to find some way of mass processing .pbm images (I feel a script using imagemagick is probably my only installed shot at that) but I'm not in scripting humour quite yet. Coffee is acting slowly this morning.

Didier Spaier · 06-12-2014, 03:58 AM

Quote:

Originally Posted by business_kid

Coffee is acting slowly this morning.

I will have coffee in a few minutes, so I can understand your feeling

Good luck.

business_kid · 06-13-2014, 03:14 AM

Well, I let tesseract loose in a mini script to do a batch job, and it converted pretty faultlessly, except that it recognizes 'fi' wrongly and I get a white square wherever in a word that letter combination occurs.

I currently have 350+ txt files, which I intend to remove, find a way of fixing the 'fi' thing in tesseract, and alter my small script slightly to send them all to the end of one big file.

business_kid · 06-13-2014, 07:43 AM

This nearly does it

Code:

#!/bin/bash
FILES=/home/dec/historical_books/temp/*.pbm
for f in $FILES
do
  echo "Processing $f file..."
  # take action on each file. $f store current file name
  tesseract $f  book
  # This second argument of that writes a text file book.txt. It refuses to
  # pipe even to stdout or write to a fifo
  cat book.txt >> Book.txt
  rm book.txt # not strictly necessary
done

I get the book. The one refinement I need now is a new page marker between each page. Any ideas?