Parallel Image Processing Algorithms on GPU Environment

The use of GPUs for general purpose applications is not a recent approach, but it was quickly becoming widespread with NVIDIA's CUDA (Compute Unified Device Architecture) architecture based on C programming language in 2007. Algorithms suitable for parallel operation, such as image processing applications, can be implemented much more quickly with the GPU. First of all, this paper summarizes the GPU and CUDA. Furthermore, we implement several conventional image processing algorithms on GPU hardware. The CPU and GPU versions of the implemented algorithms will be compared in terms of application speed. For testing purposes, four basic image processing algorithms are implemented using both CPU and GPU. We have chosen image convolution, histogram equalization, color conversion and median filter. As a result, the GPU version executes much faster than CPU version, especially when the image size is bigger.