基于AVS视频解码的错误检测方法

An error detection method for AVS video decoding

针对AVS视频传输过程中出现的码流错误,提出了一种基于健康度记录与双阈值判别的错误检测方法。在解码过程中,对每一个宏块都用邻域相关性来计算其健康度,并记录健康度较差的宏块号,健康度判别中使用双阈值,出现首个大于较大的阈值一的宏块后则改为使用较小的阈值二进行判别;若出现语法/语义错误,则再利用宏块内部相关性对出错宏块号进行微调,最后把所有被记录下的宏块都标记为出错。实验结果表明,该错误检测方法能较准确定位出错宏块位置,特别是对于出错宏块不连续和相关性增幅不够大的错误类型,与传统的逐个向前检测的方法相比,能有更好的检测效果。

Concerning the stream errors that appear during AVS video transmission, an error detection method combining recording health degrees with double thresholds is presented. In the decoding process, each macroblock＇s health degree is calculated by average intersample difference across boundary, and then the numbers of macroblocks whose health degrees are bad are recorded. Double thresholds are used in distinguishing health degrees. The smaller threshold will replace the larger one after the health degree which is larger than the larger threshold first meet. When syntax/semantics error occurs, average intersample difference of macroblock is used to fine tune the numbers of error macroblocks, and all the macroblocks which are recorded are marked as corrupted finally. The experimental results show that the proposed method can accurately locate the error macroblocks＇ positions, and it have a better detection effect in comparison with traditional method that detects macroblocks forward one by one, especially in error conditions that error macrob：ocks are incontinuous or average intersample difference across boundary don＇t become large enough.