| Video compression is the key technology of multimedia communication, which not only affects the bandwidth, but also affects the image quality. With the development of multimedia technology and network applications, data compression become more important, in which the video compression is the kernel. During these several decades, there are more excellent techniques invented, for example, the technology based wavelet. After we studying and comparing those practical technologies, we find that all of those make progress under Shannon's Information Theory, whose essence, however, is to remove data redundancy and approach information entropy.This paper is divided into two parties. The first studies how the Information Theory provides the theory basis and the technical approach. As all known, the Information Theory is "A Mathematical Theory of Communication", and only when you understand this deeply, you may perfectly understand a variety of compression technique in all standards, and may find new algorithm. So we have perfectly study the Information Theory applied in video codec. Of course, we also discussed affection of human vision system, which is the basis of loss compression.The second introduces how to realize a high efficient coder for MPEG-4. Here, we also briefly introduce the technologies accepted by MPEG-4 standard and the characteristic of DSPs.Real time is a key character of video coder, so the pivotal resolution is to improve its performance efficiently. After we studied the C compiler provided by ADI and its assemble language characterized as VLIW and SIMD, we proposal several ways to optimize the video compression algorithm using C and assemble language. These algorithms typically include DCT, SAD and motion estimation. Now, we have achieved a favorable result that the MPEG-4 code rate has arrived at 30 f/s. All things we have done above will make a good reference for other DSPs to implement CODEC.
|
PDF Full Text Request