Optimal Bit-Allocation for Wavelet Scalable Video Coding with User Preference

Abstract

• Fundamentals of Wavelet-Based Scalable Video Coding : The purpose and the general concept of scalable video coding will be introduced in this section. We also give a brief comparison between the major two scalable video coding methods, which are the wavelet-based scalable video coding and H.264/SVC. In addition, we introduce the structure of wavelet-based scalable video coding in this section. A wavelet scalable video encoder consists of two steps. The first step is to decompose each GOP (group of pictures) into multiple subbands, and the second one is to perform entropy coding, which is usually implemented by (embedded zero block coding) [1]. We are going to discuss why and how these two steps achieve video compression and provide universal scalability in a theoretical style. • The Objective Function of Scalable Video Coding : The objective of the scalable video coding will be discussed in this section. In the the discussion, the essential elements that affects the performance of scalable video coding will be considered. These essential elements are the status of network transmission, subscribers’ preferences, and the video quality in terms of the conventional Peak-to-Noise-Ratio (PSNR). And then according to the discussion, an objective function that considers these elements simultaneously is established for the optimization of the scalable video coder. • The Rate Allocation of Wavelet-Based Scalable Video Coding : Since the entropy coding procedure needs to know the number of bits allocated (which is usually referred to as rate) to each subband, the encoder should decides the rates of the subbands before the entropy coding applied to the subbands. In this section, we are going to introduce how to perform rate allocation that optimize the SVC coder with respect to the proposed objective function and compare its performance with those of the existing rate allocation methods. We will also discuss several issues related to the performance of the rate allocation, such as the concept of rate-distortion curve and the inequivalent energy between the pixel and transform domains caused by non-orthogonal filters.