Abstract:

Transmission error of video frequency always results in decoding mistake both in time and space and directly affects the quality of reconstructed image. This paper describes a new error concealment method for H.264. With the strong correlation of time and space between neighborhood macroblocks, the type and size selection of the lost block through the neighborhood blocks was estimated rightly. On the basis of the division result, a Shepard method constructed a surface was proposed to predict the motion vectors of the lost macroblock more exactly, and also improved the quality of the reconstructed image. The simulation result showed: this algorithm is better than usual boundary matching arithmetic. Under equal macroblock lost rate, the Y weight peak signal-to-noise ratio (PSNR) of video recovered by this proposed method could increase 0.75~1.5dB.

Key words: error concealment, type macroblock, surface fitting