Software Design Of AVS Video Encoder Based On DaVinci DSP Processor

With the widespread of the Internet, Multimedia has replaced text message to become the most popular information carrier in communication. Meanwhile, DSP performance has been upgraded dramatically in making real-time audio and video codec a reality, which greatly enriches our business and entertainment experience. The AVS national coding standard with independent intellectual property rights in China has guaranteed encoding quality to compete with the international mainstream standards while reducing computing complexity to an acceptable level. Considering its relatively low patent cost, AVS has a strong potential in various digital multimedia applications. At present, the AVS industrialization process is booming, to develop AVS audio and video codec software solution has become a new research focus in China. This thesis discussed the design of AVS encoder software based on DaVinci processor produced by Texas Instrument, and proposes a detailed framework design and its optimizing method.Chapter 1 reviews the development of video coding technology, and introduces the principle of video compression and the well known international video coding standard. Proceeded by an illustration of the task of this paper, a depiction of the development of DSP and its current situation will be presented.Chapter 2 discusses the hardware structure of TI DM6446 processor including DSP core framework, memory layer structure, video pre-processing module and instruction set features. A brief introduction of the integrated development environment CCS will be given afterwards.Chapter 3 provides the specification of the AVS-P2 national video coding standard, and all aspects of the standard including intra prediction, inter prediction, mode decision, transformation, quantization, entropy encoding and loop filtering.Chapter 4 develops the encoder software system on DaVinci DSP, describes the function and flow chart of the system initialization, video sampling, and video encoding modules. Then, realization and optimization strategies of key encoding modules such as motion estimation, motion compensation, loop filter, and entropy coding are depicted.Chapter 5 presents techniques of increasing video encoder's efficiency by utilizing various optimization methods:Firstly, it applies compiler options to improve code performance. Secondly, it allocates memory spaces and uses DMA to transfer data to accelerate memory access. Thirdly, it writes the linear assembly code to improve the critical functions'efficiency.Chapter 6 summarizes the system design and analyses advancing works to be accomplished.