Performance Analysis of Inter-Layer Prediction in Scalable Video Coding Extension of H.264/AVC

Scalable video coding (SVC) is a good approach for video services over heterogeneous networks. To achieve efficient video broadcasting systems, a good understanding on the performance of coding tools in SVC is necessary. In this paper, the efficiency of inter-layer prediction tools in SVC extension of H.264/AVC is thoroughly investigated by simulations. First, it is shown that inter-layer prediction is more efficient for fast/complex sequences than for slow/simple scenarios. Second, among the three inter-layer prediction methods, inter-layer residual prediction contributes most gains in simulations suggested by SVC standardization group. Nevertheless, all of them are comparatively less efficient for sequences with many details in the environment of spatial scalability. In addition, medium-grain quality scalability (MGS) outperforms coarse-grain quality scalability (CGS): MGS is able to provide up to 1 dB gain over CGS while keeping an even higher flexibility. However, MGS introduces a relatively large PSNR fluctuation which impacts visual quality.