A GOP-Adaptive Priority-Based Rate-Distortion Optimization Bitstream Extraction Algorithm for Scalable Video Coding

Abstract

Bitstream extraction is an important issue in scalable video coding (SVC). In this paper, we propose an effective GOPadaptive priority-based Rate-Distortion optimization extraction algorithm. The main feature of this scheme is being able to achieve the global optimal path. The extraction procedures, mainly including the calculation of PSNR and the allocation of priority, are carried out with Group of Pictures (GOP) as a unit. Firstly, we analyze the structure of SVC stream, then present PSNR score mechanism, based on the R-D curve's characteristics and bandwidth distributions, we propose a novel performance evaluation principle for comparing the R-D performance of different extraction paths. Finally, our proposed algorithm is introduced in detail. Experiments are performed to verify the proposed scheme with JSVM, and the experiment results show that the proposed algorithm which won't bring too much computation costs offers better performance as compared with default and SNR-based methods.

Optimized State-Distortion Extraction for Scalable Extension of H.264/AVC

“…In the state model of the fully scalable bit streams, let p j , j = 1,…,M, denotes the set of U leaves in the j th path, j k π denotes the leaf L ε p j in level k in the j th path for k=1,…,U. The M [7] and U which is the number of path and the number of level, can be calculated as…”

“…By using the rate-distortion model of quality layer, the extraction algorithm with quality layer has been addressed [4]- [5], and both of them may not practical since there is a lack of temporal and spatial extraction. A greedy priority algorithm is proposed to select the best EL which has the best rate-distortion performance in a GOP [6]- [7], but the algorithm doesn't consider the bandwidth limitations. A bit extraction algorithm that evaluates the rate distortion performance of frames has recently proposed as a frame-based scheme [8], but the complexity of distortion estimation increases with GOP size.…”