A Bit Allocation Method for Smoothing Temporal Picture Quality of Wavelet Video Coders

Three-dimensional (3D) wavelet transforms have been used in video codecs (WVC) to generate quality scalable bitstreams for communications over heterogeneous networks. However, reconstructed pictures of 3D-WVC suffer temporal quality fluctuations due to unequal error distributions in the code/decode process. Theoretical WVC coding models had been proposed to smooth temporal picture quality fluctuations. However, in practical coding processes, signal properties do not comply with assumptions in the ideal WVC model. We propose to bridge this gap between ideal and practical signal properties to improve the temporal quality smoothness of WVC. The relation between distortions of subbands and qualities of reconstructed pictures of one WVC was exploited, from which the TS algorithm was developed. The most important features of the proposed TS-WVC method are: 1) low computation complexity; 2) largely reduce the temporal picture quality variations; and 3) make it feasible for real-time WVC applications