Research On Motion-Compensated 3D Wavelet Scalable Video Coding

Video coding, which is important branch of digital image processing, has not only obtained the significant success in theory, but also obtained great progress in applications. A series of video coding international standards have appeared. However, coming with the emergence of video applications under the network environment, the heterogeneous issue makes video coding confront with the challenge in efficiency and flexibility. Therefore, the research target of video coding has been shifted from a pure compression viewpoint to designing a coding system that allows efficient transmission. The compressed stream should be capable of accommodating a variety of applications with diverse constraints in network bandwidth or receiver complexity. In past decade, scalable video coding, allowing partial decoding at a variety of resolution and quality levels from a single compressed stream, is widely considered as a promising resolution for efficient signal representation and transmission in a network environment, such as Wireless-net, Internet.This thesis introduces a motion compensated 3D wavelet scalable video coding system. And three key techniques of the coding system have been thoroughly researched, namely MC temporal filtering, 3D wavelet transform structure and 3D wavelet subband coding.Firstly, traditional wavelet transform theory and wavelet lifting scheme are illustrated. By using Haar wavelet lifting scheme, MC temporal filtering used for removing temporal redundancy is improved. This method can maintain the superiority of traditional wavelet transform to effectively remove temporal redundancy. In addition, owing to the feather of the integer form of wavelet lifting scheme, the computational complexity and memory requirement are reduced, which can enhance the performance of the wavelet video coding system.Secondly, two kinds of typical 3D wavelet transform structure have been studied. It could further remove temporal redundancy by organizing multi-level MC temporal filtering. According to the motion characteristics of the video sequence, an adaptive GOP structure is proposed. The experimental results indicate that, in addition to guarantee the coding performance, this method may especially improve the compression effect on the video sequence with high motion. Meanwhile, it can greatly reduce computational complexity and memory requirement.Wavelet subband coding method has advanced considerably in 2D image compression. Many effective embedded wavelet image coding methods have already emerged. We improve the SPIHT algorithm by expanding the coding method used for gray image to color image of the video. Quadtree split is a simple but efficient method for organizing wavelet coefficients. We proposed a wavelet image coding method based on quadtree split, which exploits energy clustering in frequency and in space. Using the fixed rate allocation method, we expand these two methods to 3D spatio-temporal subbands coding. The experimental results show that, compared with the method of subband finite state scalar quantization, both of the two wavelet coding methods can achieve better coding performance.