An efficient spatial domain error concealment method for H.264 video

This paper presents an efficient spatial domain error concealment method for the forthcoming video coding standard H.264. In H.264, a frame is divided into 4×4 blocks during the encoding procedure. For natural image signal, the blocks are smoothly connected with each other. Based on this property, a linear smoothness constraint equation that describes the connection of the lost block and its neighboring blocks can be constructed. By solving this equation, the coefficients of lost block can be recovered. Because the reconstructed high frequency coefficients may be affected by noise, the recovered center pixel may have obvious error. To eliminate the error, we use the recovered pixels that are on the boundaries of the lost block and average pixel difference to interpolate the center pixels. The implementation is simple and is suitable for real-time video application. Experimental results show our method has better recovery result than conventional approach.