Design And Implementation Of Intraframe Prediction And Deblocking Filter For AVS Decoder

With the rapid development of the modern technology, the digital information processing has to meet more strict demands. Meanwhile, the information storage devices and the network communication are also faced to meet more demands due to the large amount of the digital information. Therefore, the ISO and the MPEG establishes series of coding/decoding compression standards for videos to satisfy the requirement of the data compression, such as the popular H.264.As the audio/video standard owned by China, AVS standard is originated from H.264 standard and adopts amounts of new technologies including of intra prediction, adaptive block size motion compensation, quarter-pixel interpolation, integer cosine transformation, context-based adaptive binary arithmetic coding and so on. Because of these new technologies, the AVS standard can be used in some specific area like SD/HD coding/decoding with the higher compression efficiency and the better decoding quality. The AVS standard has a complete standard system of system, video, audio and copyright management for media; therefore, the AVS could provide comprehensive solution for audio/video coding/decoding. Theses advantages has made AVS become the best choice for the new generation compression standard.However, these new technologies also bring more complexity both in arithmetic and implementation, and limit the widespread use and application. Many excellent researchers have worked a lot on it to improve the arithmetic and the architecture. To solve these problems further, I propose some optimal methods for the design of AVS decoder in this paper.To be specific, this paper includes the following contents:First, I formulate the basic principle and the whole architecture of the AVS codec especially in the key technology and the functional modules, after having given a brief introduction about some relative background as well as software and hardware development platform.Second, to improve the decoding performance of the intra prediction further, I analyze all modes of the intra prediction arithmetic and then propose a feasible and sharable architecture of prediction. This architecture is such a kind of prediction modules which can be shared by luminance and chroma, and the accordant pre-processing unit can be shared by different prediction modules as well. I also propose a flexible and effective storage methodology which can renew dynamically, thus it could effectively decrease the access time for out-chip data frequently. Finally, I would show the feasibility and validity of the design on the strong support of the simulation results.Last but not least, on the purpose of improving the decoding performance of deblocking filter, based on both filter arithmetic and procedure, I propose some design methods, like filter sequence and storage architecture, which can manage and process temporary data efficiently. Furthermore, a complete architecture is designed, and the results of simulation and verification could show its reliability and feasibility.