Processing image with CUDA

I know that my question may be a bit stupid but here is my problem.
I am starting with CUDA programming and I have to work on images (creating filtering algorithms etc…). But I do not know how to charge an image to work on in my program, so how do I do to define an image (which libraries, what to include etc…)

I have never work on processing pictures before ^^

( I work with cuda 7.5 on windows 10)

thanks for helping

One of the easiest way is to go with PGM (grayscale) or PPM (RGB color) images in ASCII format.

Reading and writing the format can be done in about 10-20 lines of code each way, no external libraries needed.

Command line tools such as ImageMagick would allow you to convert any other format into PGM or PPM as needed.

You can typically read this into a 2D array of unsigned bytes, and for processing with CUDA you way want to store this data in a texture reference or texture object. That gives you the benefit of cached data access, bilinear interpolation, conversion to floating point on access, etc… (these features can be turned on as needed).