Key Picture Error Concealment Using Residual Motion-Copy in Scalable Video Coding

Abstract

The current JSVM software only supports Frame- Copy and Motion-Copy Error Concealment (EC) methods for the concealment of key picture loss in Group of Picture (GOP). In the proposed method, we exploit motion retrieval information to conceal key frame in scalable video coding. The auxiliary motion vectors of only key picture are interleaved into the bit stream of previous key picture. To control bit-rate increased due to interleaving of supplement information, residual difference is calculated. This encoded residual difference is used to calculate the predictor so that concealment of each macroblock in a key picture will be performed. Consequently, the increase in bit-rate due to supplement information is compensated with increase in PSNR gain with correlated increase in GOP size. Moreover, we investigate the impact of key picture distortion propagation loss of hierarchical prediction in SVC. The proposed method provides drift-free substantial improvement without significant complexity.