Implementation of 2-D Discrete Cosine Transform Algorithm on GPU
Shivang Ghetia1, Nagendra Gajjar2 and Ruchi Gajjar3
|Related article at Pubmed, Scholar Google|
Discrete Cosine Transform (DCT) is a technique to get frequency separation. When DCT is applied on an image, it will give frequency segregation of an image since it is composed of DC value and range of low frequency values to high frequency values. DCT is very useful in image compression. When high frequency values are eliminated from image, it will give efficient compression at the cost of little degradation of image quality. But, the bottleneck is that when 2-Dimentional DCT is carried out on CPU, it takes much time since there is very high order of computation. To overcome this problem, Graphics Processing Unit (GPU) has opened the door for parallel processing. In this paper, we have implemented 2-D DCT with parallel approach on NVIDIA GPU using CUDA (Compute Unified Device Architecture). By applying here presented 2-D DCT algorithm for image processing has narrowed down the time requirement and has achieved speed up by factor 97x including data transfer timing from CPU to GPU and again back to CPU. So, parallel processing of 2-D DCT algorithm on GPU has fulfilled the purpose of fast and efficient processing of an image.