A New Feedback-Based Intra Refresh Method for Robust Video Coding
Real-time wireless video transmission systems must consider both error resiliency and low complexity. However, most error resilient features of recent video coding standards tend to increase computational complexity of the encoder. In this paper, we propose a low complexity error resilient joint source-channel adaptive intra-refresh rate scheme where the optimum number of intra-coded macroblocks is determined at frame level based on the minimum estimated end-to-end distortion. In this work, we propose source and transmission distortion models whose parameters are independent on sequence type which allows real-time video encoding. The source distortion model is based on residual information and quantization step using linear least square method. The residual information is estimated using the mean-absolute difference (MAD) prediction model based on the linear relationship between intra-refresh rate and MAD. The transmission distortion model is based on recursive model using reliable feedback channel. The proposed models are used to implement a joint source-channel video coding scheme using standard H.264/AVC encoder. Accurate estimate of the actual distortion at various refresh rates are achieved and able to estimate the distortion before encoding the frame. The proposed scheme is compared with random and periodic intra refresh schemes under wireless fading channel. Improvements in PSNR quality are measured which verifies the effectiveness of the proposed scheme especially in time varying channel conditions.