Research On JPEG2000Image Compression Technology Based On GPU

As a new still image compression standard, compared with the traditional JPEG imagecompression standard, JPEG2000image compression standard can not only obtain goodquality images in the case of a higher compression ratio, but also has many advantages interms of performance, such as the region of interest encoding, supports both lossy and losslesscompression, progressive transmission, etc. Therefore, JPEG2000standard has been widelyused in the field of modern multimedia network technology. But the algorithm of JPEG2000itself has also increased the difficulties on the basis of JPEG standard.The current CPU with software processing is very difficult to fulfill the real-timerequirements of JPEG2000still image compression in some practical applications, or costly.However, with the development of computer science, the powerful parallel computingcapacity of programmable Graphics Processing Unit (GPU) has been followed with interestall over the world. Especially for the publication of CUDA which has changed the fate ofGPU parallel accelerated computing completely. Using CUDA technology implement theJPEG2000still image compression standard on the general purpose computing GPU caneffectively avoid the shortcomings such as the inflexibility or the long development cycleswhich is generated on hardware implementation.This paper firstly introduced the basic principles of the core coding system inJPEG2000still image compression standard,the hardware architecture of General-purposeGPU, CUDA heterogeneous programming model and storage model.Then based on the GPUparallel computing features, the paper analyzed the key part in JPEG2000algorithm whichcan be parallel implemented on GPU in detail. A program is proposed that the DWT part andthe Tier1part can be implemented by CUDA. And achieve the goal of improving the speed ofJPEG2000still image compression by parallel implementation on the GPU. Finally, the papercompared with the experimental results that JPEG2000standard is implemented on a generalpurpose GPU and concluded: JPEG2000image compression algorithm compression speedcan be substantially improved by using CUDA technology to parallel accelerated computing the JPEG2000algorithm.