Research On Hierarchical Video Coding Based On Four-Dimensional Matrix Model Thoery

The object of classical video coding is to maximize the quality of video reconstruction under certain rate. In this kind of system, encoder compresses the input video signal into a certain bit rate that is near or lower than the capacity of the channel, while the decoder will reconstruct the video image by using the bits that received from all the channels, which is called non-hierarchical video coding.In recent years, with the persistent expansion of Internet and the rapidly increasing number of consumers, the application of Internet is becoming wider and wider and the need of people for Internet in content increasing, too. Yet, the nature of Internet makes it only provide limited and unfixed bandwidth and unreliable service. So, the research and development of hierarchical video coding is becoming a hotspot, receiving more and more attention.Moreover, in order to express a color video as comprehensive as possible, we have advanced a new model called Four-dimensional Matrix Model to show a color video, which unify all the relativities in the color video into an integrated mathematics model using a four-dimensional matrix. By applying this theory to color video coding, we successfully solved the problem that classical color video coding algorithm doesn't take the relativity among channels into account, thereby can not fully wipe off all the kinds of relativities in color video and restrict the further improvement of compress ratio.Just under this kind of circumstances, this thesis did some work on the study of hierarchical coding of color video.In order to provide more strong flexibility to meet different needs, this thesis adopted a new hierarchical coding mechanism called Fine Granular Scalability which is introduced in standard MPEG-4. The encoding divides original video sequence into basis bit stream and enhancement bit streams which is expressed using bit-plane coding system. Also FGS encoding can control the enhancementbit stream with continuous rate. All of these are as following:The first part is about the four-dimensional matrix express of original color video and the division of submatrix. This thesis unifies color video into an integrated mathematics model using the four-dimensional theory our lab has introduced and fully wipe off the relativities among pixels, frames, thereby get the high compress ratio under the circumstances of high PSNR. In the hierarchical video coding system based on 4DM-DCT theory, original color video is represented a four-dimensional matrix, then divided into some 8x8x3x3 four-dimensional submatrixes, which is a preparation for the following 4DM-DCT.The purpose of the second part is to perform 4DM-DCT as well as qutization to the above mentioned submatrixes. From the classical matrix theory, we have represented the multi-dimensional matrix theory systematically. Combining four-dimensional matrix and traditional 2D-DCT theory, the 4DM-DCT and its inverse transform are given and proved as lemmas and the orthogonal performance of the transform is proved at the same time. So, by performing 4DM-DCT to the submatrixes redundant information among pixels, frames in color video sequences can be effectively removed. Moreover, in order to further improve compress ratio, quantization technology is also introduced to quantify the DCT coefficients.The third part is to encode the quantization coefficients using run length coding and get the basis bit stream. Originally, MPEG-4 introduced threetechniques to FGS: bit-plane coding about DCT coefficients, wavelet coding about image remains,matching coding about image remains. After some experiments, bit-plane coding is selected as the coding standard of FGS in MPEG-4. So, this thesis also adopted this technique here.In the fourth part, the new coefficients which are gotten by using the DCT coefficients minus those gotten by run length coding the quantization coefficients above mentioned are encoded with the bit-plane coding technique, consequently,...