Total Variation Regularized Turbo Decoder (TV-TD) for Reliable Image Transmission in Wireless Multimedia Sensor Networks

This paper presents a novel total variation regularization based turbo decoding (TV-TD) scheme for robust image transmission in wireless multimedia sensor networks. The proposed decoder comprises of constituent non-binary soft-output Viterbi algorithm (SOVA) decoders, where the decoder for the systematic bits is modified to a TV regularized non-binary SOVA decoder (TV-NSD) that additionally employs a TV cost- function in order to exploit the bounded variation property of the images for improved reconstruction. Further, a novel trellis structure is proposed for the TV-NSD, which evaluates the pertinent soft-outputs of the codewords corresponding to the coded image components. The expressions for various node/ branch metrics and soft-output values are determined for the proposed trellis structure followed by a description of the TV-TD scheme for image reconstruction. Simulations performed using various test images demonstrate the improved recovery achieved by the proposed decoder, thus making it well suited for image transmission in power-constrained sensor networks.