Font Size: a A A

Research On Macroblock-level Adaptive Frequency Weighting Of H.264

Posted on:2011-10-13Degree:MasterType:Thesis
Country:ChinaCandidate:Q Q NiuFull Text:PDF
GTID:2198330332975550Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
With the advent of high-definition (H-D) digital video age, H-D has become the inevitable tendency of the development of a variety of digital devices. The demands for high fidelity with no change and upgrade to existing networks increase greatly. How to raise video coding compression ratio, remove all redundant information and reduce the rate to conserve bandwidth as much as possible with a high fidelity is a very hot research field.Today, the frequency weighting technique has been used in MPEG-2 and further developed in H.264 High Profile (HP), in which quantization matrix is updated picture by picture without considering the different characteristics of different parts. However, the subjective quality and compression ratio are not so good. Based on the picture-level adaptive frequency weighting of H.264, the Macroblock-Level adaptive frequency weighting algorithms for H.264 are proposed.First, in the report, the structure and decoding principle of H.264 are analyzed, and then the advantages and key technology are concluded comparing with the previous video coding standard.Second, the characteristic of human visual system is presented and employed. Considering the shortcoming existing in H.264 picture-level adaptive frequency weighting algorithm, an improved Macroblock-Level adaptive frequency weighting algorithm based on human visual system is proposed in this paper to improve the subjective quality.Third, by the analysis of the region of interest detecting technique, the importance of the region of interest is used. In order to ensure the efficiency of algorithm, a better approach to detect the region of interest is proposed.Moreover, the improved region of interest detect algorithm is applied in the Macroblock-Level adaptive frequency weighting, combing with the improved quantization matrix, a Macroblock-level adaptive frequency weighting based on region of interest is presented.Finally, the whole article is summarized and the inadequacy of the current work is pointed out. Also, the direction for future work is specified at the end of the paper.
Keywords/Search Tags:H.264, Human Visual System, Region of Interest, Frequency Weighting
PDF Full Text Request
Related items