Reaserch Into GPU High-Speed Optimization Method Of Wavelet Transformation In Image Compression

In today's society, digital image has become important means to access and send information because of its large amount of information and the feature of vivid. Due to limitations of storage space and transmission bandwidth, image compression has come to the key technology in transmission, processing and storaging of digital image. As the features of "multi-resolution analysis" and "adaptive", Wavelet Transform has become the most widely used time-frequency analysis tool in the field of image compression, so many excellent image compression algorithms are based on it. With the rapid development of Internet technology,the number of images need to be instantly stored and real-time processed is exponentially increasing. The existing compression technologies have difficult to meet the needs of processing speed, so the research on fast implementation of wavelet transform in image compression has important practical Significance.In this paper, on the CUDA, methods of GPU high-speed optimization to wavelet transform are studied.According to the computing characteristics of wavelet. We made a specific program to optimize 5 / 3 wavelet transform on the CUDA platform, and then implemented and improved the programs in steps. Next we tested the performance in numbers of hardwares, and finally got an optimization program of strong portability and high value. Experimental results showed that under the premise of the quality of image can be guaranteed, the optimization process can obtain speedup of an order of magnitude to CPU. The high-speed wavelet transform is achieved successfully on GPU. This research laid the foundation for high-performance image compression based on Wavelet Transform, and provided inspiration to processing of large-capacity image in the further study on the CUDA platform to make full use of GPU parallel computing features.