An effective side information generation scheme for Wyner-Ziv video coding

Abstract

Distributed Video Coding (DVC) is a video coding archetype that explores the source statistics at the decoder and hence reducing the encoder complexity. The Rate-Distortion (RD) performance of DVC strongly depends on the quality of the side information (SI) generation. So, efficient techniques to generate reliable SI are therefore essential to obtain a better quality of decoded video. In this paper, a SI generation technique based on radial basis function neural network (RBFNN) is proposed. RBF networks are widely used in various applications including function approximation and pattern recognition. Compared to other feed-forward neural networks, it has many advantages which makes it more suitable for nonlinear system modeling. The proposed model is trained and tested with different standard video sequences. The proposed scheme is merged with Transform Domain Wyner-Ziv (TDWZ) architecture and different experiments are performed to derive an overall conclusion. The overall experimental results demonstrate that the proposed technique produces an improved result in terms of Peak Signal to Noise Ratio (PSNR), bit rate, number of parity requests, decoding time complexity, etc. as compared to the existing state-of-art techniques.