论文部分内容阅读
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.
To improve the development of TMS320C80 The decoding speed of the whole software decoding system reduces the on-chip RAM and off-chip data exchange. A variable length decoding (VLD) algorithm is proposed. According to the C80 high-speed parallel processing capability and only a small on-chip memory space and other characteristics of the VLD code table design. Using this algorithm, H can be stored in an on-chip RAM on the C80. 263 all VLD code tables. Compared with the RezaH method, which is commonly used to search the bit-by-bit decoding tree, the decoding speed of the MisAmerican image sequence is 70 times faster than that of the RezaH method. 4%, finally can realize the decoding speed of 12.5 frames / s in CIF format. This method is equally applicable to other systems in the analysis that require memory and decoding speed.