Adaptive frame based flexible macroblock ordering for error resilient H.264 video coding and transmission

Abstract

The transmission of video signal over wireless environments is becoming more common today as advances of telecommunication and multimedia convergence market make it possible. This enables the seamless mobile multimedia applications such as mobile video class and wireless video streaming. However, to achieve good video quality from such transmissions is still a challenging issue for practical video applications. This is due to the high error prone characteristic of wireless channels due to multipath effects and deep fades. Thus, research approaches to combat channel errors for better video quality especially error resilience video coding are one of the key research topics nowadays. In this dissertation, we propose two methods for error resilient in wireless video coding based on adaptive flexible macroblock ordering (FMO) for wireless video streaming applications. In the first method, effect of error propagation between frames is used as an indicator to generate FMO map frame-by-frame. A technique for generating FMO map is proposed to adapt with predicted error bursts of channel. In addition, a suitable intra refresh rate is selected to reduce the effect of error propagation. Therefore, the number of undecodable important macroblocks (MBs) is decreased. In the second method, we investigate the cross-layer approach between Application and Media Access Control (MAC) layers. In traditional approaches, video packets are classified and mapped into queues with different priorities at MAC layer. However, these approaches cause the unnecessary dropping packet at low priority queues. In the proposed approach, FMO map for each frame is generated by adaptive FMO and queuing overflow rate. Based on the estimated queuing overflow state information at the MAC layer, encoder maps MBs into slice groups in such a way that the arrival rate of packets to the full queue is reduced and arrival rate of packets to empty queue is increased. Hence, the number of dropped packet at Mac layer is minimized. The results in experiments showed that the proposed methods gain an improvement in terms of peak signal-to-noise rate (PSNR), number of undecodable MBs and number of dropped packets as compared to the previously proposed methods. The contributions gained from this dissertation would benefit the video communication community in future adoption of technology in practical wireless video transmission applications.