H.264中变换和量化的SIMD优化  

SIMD Optimization of Transform and Quantization in H.264

作　　者：魏芳[1] 李学明[1] 

机构地区：[1]北京邮电大学信息工程学院,北京100876

出　　处：《计算机工程与应用》2004年第17期24-27,共4页Computer Engineering and Applications

基　　金：国家自然科学基金项目(编号:60172050)资助

摘　　要：H.264是一个新的基于运动补偿+变换+量化+熵编码框架的视频编码国际标准。H.264中采用了大量的新技术,这些技术在提高编码效率的同时,也极大地增加了算法的复杂度。为此H.264在保证性能的前提下也做了一些优化,如变换和量化可以在16比特精度下完成,并且除了量化中需要少许乘法外,其余可以只用加法和移位实现。这些特点使得H.264中的变换和量化可以很好地使用支持单指令多数据(SIMD)的MMX技术进行进一步优化。该文首先介绍了H.264中变换和量化的实现过程和特点,接下来重点讨论了利用MMX指令对变换和量化中的关键部分进行优化的方法,最后给出了计算机仿真的结果,并对结果进行了分析。仿真结果表明:该文提出的方法可以使H.264的变换和量化模块的运算速度提高到原来的3.5～5.2倍,优化效果十分显著。H.264is a new international video coding standard based on motion estimation,transform,quantization and entropy coding framework.H.264adopts plentiful new technologies,which greatly improve coding efficiency but hugely adding complexity.Therefore some simplifications with no performance loss such as low complexity transform and quantization are introduced into H.264.The transform and quantization in H.264requires only additions and shifts except a minimum number of multiplication in quantization and can be computed in16-bit arithmetic,which make it possible to be magnificently optimized with MMX technology supporting SIMD(Single Instruction Multiple Data)parallelism.This paper firstly introduces the implementation process and characteristic of the transform and quantization in H.264,further the key modules of which are optimized by making full advantage of MMX.At last,conclusions can be made after analyzing the presented results of computer simulation,that the proposed algorithms are prominent while3.5～5.2times calculation speed can be achieved after optimization.

关 键 词：H.264 整数余弦变换 哈达玛变换 周期量化 SIMD MMX 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]