Hello, I want to build a simple OCR program and I was wondering what library to use. The images are saved on my hdd as colour .png files, i won't be using any effects on them but I will be working with pixels a lot, so direct access to them would really be great.

So I was curious if there's a lib that provides direct access to individual pixels and pretty much nothing else (libpng, devIL or QImage from qt are all good but each has it's downfalls).

Perhaps imagemagick would do?

http://www.imagemagick.org/script/ap...f31u7hmljf7vq5

1 members found this post helpful.

There are various imaging libraries available. libPng is one of the most used. if you are using Python try PIL.

I don't know if the library classifies as ''simple'', but did you have a look at GIL from Adobe? It can read png files and you may access single pixels... However, there is a lot more included in that library, so I don't know if it fits your needs.

The URL is
http://www.boost.org/doc/libs/1_46_1...doc/index.html

Cheers
Tobias

I would suggest PIL if you are using Python, I haven't worked with PNGs with any other language so that is all the help I can offer.

My 1st thought, too.

I hesitated to suggest cimg, because it doesn't fit the "almost nothing else" requirement, but I'll throw out out anyway.

So what did you feel were the "downfalls" of the libraries you listed?

You should explain what exactly you didn't like about libraries you listed. QImage, for example, is pretty decent and allows manipulation of individual pixels. There's also SDL_image, but it is supposed to be used in SDL applications only.

thank you for your replies!
After searching a bit a decided to settle with imagemagick.

and as for the other libraries, it's just a personal thing:
libpng has a pretty hard-to-work-with API, so png++ would've be better since I was going to build a wrapper for libpng anyway
QImage, required QT and I don't think it would've been the best decision to add libqt as a dep for just it's image library
devIL was my second choice after png++ but even though it has an easier API, it's still in C from what I saw, so again I would've needed to build a wrapper (not that I don't like C or anything, it's just I feel like code is more organized when you use OOP)
PS: I don't have anything against any library, but I wanted something simple for my first image-manipulation program
cheers!