在 TMS320C80 上实现的一种可变长解码算法  

A New Huffman Decoding Method Based on TMS320C80

作　　者：谭径微[1] 周璇[1] 方勤 余松煜[1] 

机构地区：[1]上海交通大学图象通信与信息处理研究所

出　　处：《上海交通大学学报》1998年第9期1-5,共5页Journal of Shanghai Jiaotong University

摘　　要：为提高在ＴＭＳ３２０Ｃ８０上开发Ｈ．２６３全软件解码系统的解码速度，减少片内ＲＡＭ与片外数据交换，提出了一种可变长解码（ＶＬＤ）算法．该算法根据Ｃ８０的高速并行处理能力和只有很小片内存储空间等特点，对ＶＬＤ码表进行设计．使用该算法，在Ｃ８０的一个片内ＲＡＭ中就能存放Ｈ．２６３的所有ＶＬＤ码表．为与目前较常用的逐位查找解码树方法和ＲｅｚａＨ方法比较，对ＭｉｓＡｍｅｒｉｃａｎ图象序列的７０帧图象进行实验，其解码速度是逐位查找解码树方法的２倍，比ＲｅｚａＨ方法快９．４％，最终可实现ＣＩＦ格式的１２．５帧／ｓ的解码速率．该方法同样适用于分析中需考虑内存及解码速度的其他系统中．A developed method for variable length code decoding (VLD) is proposed which is used in the wholly software H.263 decoding system by TMS320C80 for improving the decoding speed and reducing the frequency of the data exchange between on chip and off chip memory. Based on the characters of C80——highly processing speed and parallel processing performance but only with limited on chip memory, this method makes a design of VLD table. By this method, all of the H.263's VLD tables can be stored in one on chip RAM of C80. Compared with the common bit detection code tree method and the VLD method proposed by Reza H, the experimental result on 70 frames of Miss America shows that the decoding speed is 1 time higher than that of the bit detection code tree method and 109.4% as high as the method proposed by Reza. The result shows that the whole decoding system's high performance——12.5 frame per second for CIF format can be met. This method is also useful for other systems needed to consider both memory space and processing speed.

关 键 词：图象处理 可变长解码 视频解码 TMS320C80 

分 类 号：TN919.8[电子电信—通信与信息系统] TN941.1[电子电信—信息与通信工程]