论文部分内容阅读
由于DPCM解码器的递归结构,当出现信道误码时,误差以一种特殊的误差图案形式在复原的图象中扩散。这种误差图案的初始幅值取决于量化特性、码字分配及出现差错的比特位置。由于量化后予测误差的不等概率分布及量化器的非线性特性,不同的码字分配方法对于初始的误差幅值有不同的影响。本文在每个取样中的错误比特不多于1个的条件下,比较了四种码字分配方式。结果表明,符号-幅值码在这四种码字分配方式中性能最好,采用这种码字分配方式可以得到最小的平均误码幅值,并可以减小信道误码的可见性。
Due to the recursive structure of the DPCM decoder, when a channel error occurs, the error spreads in the reconstructed image in the form of a special error pattern. The initial amplitude of this error pattern depends on the quantization characteristics, codeword allocation, and the bit position at which the error occurred. Due to the unequal probability distribution of the measured error after quantization and the non-linearity of the quantizer, different codeword allocation methods have different effects on the initial error amplitude. In this paper, the error code in each sample no more than one condition, compared to four kinds of code allocation. The results show that symbol-amplitude codes have the best performance among the four codeword allocation schemes. With this codeword allocation method, the minimum average error code amplitude can be obtained and the channel error code visibility can be reduced.