Abstract:

The error resilience of AVS-M can not achieve desirable results because the adopted error control technology is too simple. In order to solve this problem, the end-to-end error propagation distortion in AVS-M video encoding and transmission is investigated at a precision level of pixel, and a non-feedback key reference picture selection （KRPS） algorithm is realized, based on the rate-distortion optimization model which combines the general transmission distortion estimation model with the rate-distortion optimization （RDO） rule. Then, a feedback KRPS algorithm is proposed based on the non-feedback KRPS. Experimental results indicate that the adopted distortion estimation model can accurately estimate the end-to-end distortion, and that, as compared with the algorithm of AVS-M, the non-feedback and the feedback KRPS algorithms respectively result in the Y-PSNR gains of 1.9 -6. 4dB and 2. 7 ～ 10. 9dB, at a packet loss rate varying from 3% to 20%. It is thus concluded that the two proposed algorithms effectively improve the performance of error resilience in AVS-M video encoding.

Key words: video coding, error resilience, rate-distortion optimization, end-to-end distortion, reference picture selection