基于多角度的JPEG算法优化的硬件实现  

Hardware Realization of JPEG Algorithm Optimization Based on Multi-angle

作　　者：周焱江 李锦明[1] ZHOU Yanjiang;LI Jinming(School of Semiconductor and Physics,North University of China,Taiyuan 030051)

机构地区：[1]中北大学半导体与物理学院,太原030051

出　　处：《舰船电子工程》2024年第12期126-131,214,共7页Ship Electronic Engineering

基　　金：基础研究群项目(编号:514010504-308)资助。

摘　　要：针对传统JPEG压缩算法需要占用大量的CPU资源,且传输速率较慢的问题,该设计提出了一种从多角度对JPEG算法进行优化的方案并使其适合于用硬件来实现。从算法角度使用近似DCT算法和剪枝理论来对传统DCT算法进行优化并提出了新的量化表,从系统架构上采用并行流水线的方式且将量化与Z排序相结合来进一步优化传输速度,从硬件方面使用Altera公司的Cyclone Ⅲ芯片对设计进行实现,充分利用片上资源,满足JPEG的快速实现,最后在Matlab环境中完成解码且对图像质量进行评判。经过测试,该实现方法的PSNR值均在30 d B以上,SSIM值在0.84左右,图像压缩比也可达到26.7:1,压缩率大大增加且在进行压缩转换后人眼视觉效果仍然良好,符合设计要求。Aiming at the problem that the traditional JPEG compression algorithm needs to occupy a large amount of CPU re⁃sources and the transmission rate is slow,this design proposes a scheme to optimize the JPEG algorithm from multi-angle and makes it suitable for hardware implementation.From the perspective of algorithm,the approximate DCT algorithm and pruning theo⁃ry are used to optimize the traditional DCT and a new quantization table is proposed.From the system architecture,the parallel pipe⁃line method is adopted and the quantization and Z sorting are combined to further optimize the transmission speed.From the hard⁃ware aspect,Altera's Cyclone Ⅲ chip is used to implement the design,making full use of on-chip resources to meet the rapid imple⁃mentation of JPEG.Finally,the decoding is completed in the Matlab environment and the image quality is judged.After testing,the PSNR value of the implementation method is above 30 dB,the SSIM value is about 0.84,the image compression ratio can also reach 26.7:1,the compression rate is greatly increased,and the human eye visual effect is still good after compression conversion,which meets the design requirements.

关 键 词：近似DCT 去块效应 FPGA SSIM 并行流水线 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]