Abstract: As image processing algorithms are increasingly complex, increasingly require high performance to CPU, while traditional CPU-based image processing methods cannot meet demand. The graphics processing of CUDA-based graphic processing Unit (GPU) is researched and implemented. This paper full makes use of GPU prominent parallel processing capability, adapts CUDA, uses C++ language to implement image processing algorithm. By studying and designing of GPU parallel computing process of Gaussian blur processing algorithms, color negative processing algorithms, combined processing algorithms, and to achieve the same eflective performance comparison with CPU, this paper proves the efficiency of the GPU-based image processing algorithms.

Key words: digital image processing, unified computing device architecture, graphics processor, parallel computing