Robust Total Variation and Sphere Decoding Based Image Reconstruction for Wireless Sensor Networks

Provided by: Indian Institute of Technology Kanpur
Topic: Mobility
Format: PDF
In this paper, the authors propose a novel scheme for low-complexity robust recovery in the context of image transmission over Rayleigh fading wireless sensor links. The proposed scheme employs an optimal Maximum-Likelihood (ML) scheme for transmit vector detection in conjunction with a total variation regularization factor towards sparsity of the image gradient vector. It is demonstrated that while the optimal decoder for image recovery has an exponential complexity, a relaxed sphere-decoder based vector symbol detection for robust image recovery yields near optimal results. Simulation results show that their method has a superior performance compared to the conventional maximum-likelihood based symbol detection scheme for image recovery in terms of mean squared-error and visual quality of the reconstructed image at the cluster head.

Find By Topic