Recovering single lost description using DCT-MDC based on network coding

Abstract

Multiple Description Coding (MDC) is one of the source coding methods for providing several quantized, independent and identically distributed (iid) streams of data while Network coding (NC) is a method to recover the loss which occurs in the network. NC applies algebraic mathematic operation on independent data sequences in the intermediate nodes within a network. This paper attempts to transmit the higher value coefficients of image description to the channel and recover any loss happens in the network by joining the DCT-MDC and network coding. First, input image is zero padded and then DCT-MDC is performed on subimages. Next, downsampling block is used to create 4 multiple descriptions. These 4 descriptions are transmitted over the network and once the disconnection happens in the primary links, by using p-cycle NC lost data is reconstructed. In this method there is no need of retransmission (feedback). Results show that the PSNR and subjective evaluation factors for the proposed method are higher and clearer than the previous work which leads to have higher throughput.