An improved scheme for multiple description image coding with quincunx segmentation

In this paper, an improved scheme for multiple description image coding with quincunx segmentation is proposed. For each description, this scheme processes one of the quincunx-subsampled images as the base layer information and the prediction error for the other quincunx-subsampled image as the enhancement layer information, based on multiple description layered coding. In addition, an improved interpolation scheme for quincunx-subsampled images is employed in the interpolation prediction of coder and decoder to predict the other quincunx-subsampled image. Experiments show that preserving the central rate-distortion performance basically; this improved scheme can achieve better side rate-distortion performance than the original scheme, especially at moderate or high bit rate.