Error concealment methods in video communication are divided to two categories. One is temporal error concealment, and the other is spatial error concealment. Temporal method uses the characteristics of temporal correlation between current frame and previous one, while spatial method uses the spatial correlation inside the frame. In general, temporal method is good to conceal the corrupted image block. But, in the case of fast moving image or scene change image, spatial method is better than the temporal one. This thesis proposes the algorithm which selects the better one between temporal method and spatial method when some blocks are corrupted.
The proposed algorithm has three steps. First, the corruped image blocks are divided into fore classes; scene change, moving camera, moving object, and no motion. Second, we decide which classes the corrupted block is included. In the decision, two cost functions are used. The first cost function has the meaning of temporal difference of neighboring pixels, and the second, distribution of neighboring motion vectors. At last, we select the better error concealment method for the corrupted image block.
The performance of proposed algorithm is tested by computer simulation. The proposed algorithm is compared with other error concealment algorithms by PSNR and subjective image quality.
The proposed algorithm gives remarkable performance, because it can select the better error concealment method for each set block corrupted. And it is particularly effective in the case of fast moving or scene change.
