Algorithms Reserach On MPEG-4 Coding And Its Optimized Implementation On DSP

Video Image is the most important information carrier with the biggest data quantity. Moreover, the image compression has been developed from the first generation encoding technology based on pixel to the second generation based on object. In order to standardize and promote the development of image coding technology, ISO (International Organization for Standardization) has constituted a series of coding standards.In the current image coding system, the image coding algorithms are very complex , and it is impossible to implement the real time coding using the software, so this kind of system are usually implemented by hardware. The TMS320C6000 series of TI Company has been used in many image compression systems for its good processing capacity and the high integration.MPEG-4 is an video coding standard based on object, using the second generation coding technology. Firstly, the thesis takes an overview of the specialities and new technologies of MPEG-4. Secondly, the key encoding algorithms in MPEG-4, including motion estimation, DCT, quantization, run-length encoding and Huffman coding, are presented, and then the algorithms which consume too much time are improved. Thirdly, in block matching part, the DS (diamond search) algorithm is improved by increasing the zero-vector judgment and the starting point prediction; In texture coding part, the Loeffler algorithm is improved; In quantization part, a new matrix avoiding division operation is constructed. Moreover, the thesis implements a software encoder which has basic encoding function in the environment of Visual C++ 6.0. Finally, the software system is been transplanted into TMS320C6416 DSP chip with the linear assembly codes instead of the C codes of the key algorithms, and the encoder is simulated in CCS.