Low complexity adaptation of H.264/MPEG-4 SVC for multiple description video coding

Abstract

The scalable video coding (SVC) extension of the H.264/MPEG-4 AVC standard provides a feature-rich partition of the video that is amenable to multiple description coding (MDC) for transmission over multiple unreliable channels. In this paper we propose a novel low-complexity MDC scheme where we enforce compliance with H.264/MPEG-4 SVC and restrict our processing to the network adaptation layer (NAL) level. In our scheme video is encoded with two spatial and multiple temporal layers. The lower resolution spatial layer is coarsely encoded using a high quantization parameter (QP) value, while the higher resolution spatial layer is encoded with finer QP. The MDC is performed on an already encoded video stream by partitioning the NAL units from the encoded video stream into the two descriptions, where the coarse lower spatial resolution is duplicated in the two descriptions while finely encoded enhancement layer information is symmetrically distributed between the two descriptions. Experimental results show that compared with single description coding (SDC) the performance of proposed scheme results in gains of up to 3 dB.