基于CUDA技术的DCT并行算法研究与实现  

Research and Implementation of DCT Parallel Algorithm Based on CUDA Technology

作　　者：张金霜 黄旭彬 ZHANG Jin-shuang;HUANG Xu-bin(Guangdong Perschool Normal College In Maoming,Maoming 525000,China)

机构地区：[1]广东茂名幼儿师范专科学校,广东茂名525000

出　　处：《电脑知识与技术》2020年第33期198-201,共4页Computer Knowledge and Technology

摘　　要：JPEG有损压缩算法以DCT变换算法为核心,但DCT变换算法较为耗时,为提升图像压缩效率,提出利用基于GPU平台的CUDA技术对DCT算法做并行优化。通过分析DCT变换算法的原理,发现DCT算法具有很好的并行性,优化空间较大,于是利用CUDA技术实现高速DCT并行算法。实验结果表明,在一张2592×2592分辨率的图像做DCT变换,加速比能达到124.5,算法效率得到明显提升,且压缩效果无显著差异。The core of JPEG lossy compression algorithm is DCT transform algorithm,but DCT is time-consuming.In order to improve the image compression efficiency,the parallel optimization of DCT algorithm is proposed by CUDA technology based on GPU platform.By analyzing the principle of DCT transform algorithm,it is found that DCT algorithm has good parallelism and large optimization space,so CUDA technology is used to realize a high-speed DCT parallel algorithms.The experimental results show that the acceleration ratio of DCT transformation in a 2592×2592 resolution image can reach 124.5.The efficiency of the algorithm was significantly improved,and the compression effect was not significantly different.

关 键 词：CUDA GPU 并行程序设计 DCT算法 图像压缩 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]