Rate Allocation of Equal Image Quality for MPEG-4 FGS Video Streaming

Xuejun Zhao,Yuwen He,Shiqiang Yang,Yuzhuo Zhong

2002-01-01

Abstract:This paper proposes a new rate allocation method for MPEG-4 fine granularity scalable (FGS) video streaming, with which the received image quality can be smoothed during the streaming period. Although average bit rate allocation is quite simple, the image's quality changes greatly between frames, which will result in quality fluctuation and degrade the streaming quality. In this paper the bits are allocated to those frames according to their rate-distortion curve in order to obtain the same objective quality. Fast rate allocation algorithm of equal image quality is implemented based on Newton search method, which is much faster than linear search method. It is important to realize fast rate allocation for bit-stream truncating in practical video streaming applications. To validate the performance of proposed rate allocation method, other three relevant methods are compared with different sequences at different bit-rate, such as average bit rate allocation, rate allocation with equal slope of rate-distortion curve, grouping rate allocation. In comparison, the proposed rate allocation method achieves the best performance among four tested methods. The fluctuating of distortion between two adjacent frames is reduced to about 0.3dB, which is much less than other methods. The quality's variation is less than 1.0dB over the buffered video frames in one shot. The speed of proposed fast rate allocation algorithm is also measured, which is about 60 times faster than that of real time, thus this method is feasible in video streaming application.