| With the development of network and multimedia technology, the storage and transmission of digital multimedia information become a hot topic research. Because digital multimedia information, especially video, is enormous, it must be compressed to store and transmit efficiently. Taking an example for transmitting the original video of HDTV, information content per second is almost 1G bits, which is highly beyond communication channel's capability. It is however a huge waste of communication bandwidth to send the original video data through communication channel. Thus the research and development of video compression have a very important practical significance.In actual image compressing systems, because of the very complex algorithms of video compression and large sum of computing, software method can't satisfy real-time requirement. Accordingly these systems mostly work by using hardware method. In hardware implementing methods, video processing chip for special purpose is powerful but is not flexible, and its application is limited. However, general high-performance Digital Signal Processing (DSP) is applied widely due to flexible software design and powerful adaptive capability. In market of chips for general DSP, TI company's series of C6000 which have strong processing capability, flexible external interfaces and completed development tools are adapted by most embedded image real-time systems,Because its application environment in this paper is very peculiar, the video encoder must satisfy many kinds of bit rate (2048Kbps and 256Kbps), image resolution ratio (720x576 and 352x288), and can process color image and gray image, and can change its frame ratio at random. Obviously, the chips for special purpose can't meet these needs. Therefore, general DSP Chip is only selected to implement multi-requirements video encoder. Surely, there exists certain gap between general DSP chip and the chips for special purpose in the capacity of video processing. If general DSP want to get the performances which can be attained only by special video processing chip, video encoding structure and software compressing algorithms need to optimize and improve on to make up the shortage in processing ability of general DSP. Besides, due to the nature of variable length codes (VLCs), they are very sensitive to errors and vulnerable in error-prone environments. Once errors are produced, they can propagate in successive frames because of the mechanism of prediction, which can badly affect the subjective quality of images decoded. Thus, it is necessary for the video encoder to have its ability to tolerate source error simply and effectively.Considering the above factors, the paper probes in three aspects as follows:1. The traditional video encoder regards a frame as one unit coded. Although its structure is simple, it affects the performances of DSP. The high-speed memory inside DSP is restrictedin size so that the whole image data of one frame cannot be stored in interior memory, but just be placed in exterior memory. Owing to the difference of interior and exterior memory in speed, DSP have to be often clogged in accessing the exterior memory, and its efficiency is bad. So the paper presents an encoding scheme based on GOMB (Group Of Marco Block). It regards a GOMB as one unit encoded and makes the best use of the high-speed memory restricted in size inside DSP, and thus greatly improves the efficiency of encoding.2. Quantization in video encoder accounts for about 47 percent of the whole coding period during coding, so it's very important for improving the encoder's efficiency to study the quantization algorithm suitable to DSP. There exist two problems with the algorithms for quantization of the MPEG video compression standards realized on the DSP (Digital Signal Processor). Firstly, the efficiency of the traditional quantization based on division operation is very low, owing to lacking of a division component inside DSP. Secondly, the complicated structure of the quantizers cannot improve the quality of images... |
PDF Full Text Request