Dissertation > Industrial Technology > Radio electronics, telecommunications technology > Communicate > Image communication, multimedia communication > Image coding

Research and Implementation on Key Techniques of GPU Based AVS Video Codec

Author MaSiShuo
Tutor QuJunHua
School North China Electric Power University
Course Applied Computer Technology
Keywords GPU AV S Video Coder Parallel Coding
CLC TN919.81
Type Master's thesis
Year 2012
Downloads 70
Quotes 0
Download Dissertation

Nowadays, video is effecting and changing the whole society in kinds of ways, no matter the area is political, economic, culture, sports, education or entertainment. Since the digital video began to replace analog video, digital video applications became the driving force of information industry’s development. As the most challenging branch, the research motivation of video coding is to get rid of the digital video redundant, saves the costs in transferring and storing video. Video coding standard is one of the most important ways to promote the development of video coding, as a video coding standard drafted by our country independently and with independent intellectual property, AVS standard has a breakthrough in the coding efficiency, the important indexes have reached the advanced level of other standards around the world.On the other hand, the development of graphic processing unit (GPU) provided another way to develop video coding rapidly. GPU is different from CPU; it adopted a many-core microprocessor system structure. The parallel encoding theory which is emphasized by GPU has achieved widespread success in many areas, in this paper, GPU is adopted to implement the video coder which is based on AVS standard. In the year of2007, NVDIA Company formally released computer unified device architecture (CUDA), CUDA is a general development environment and software system using class C language for general calculation, people could code parallel with GPU using CUDA conveniently. In this paper, we made a large number of experiments and theory research, from that, we put forward the four levels parallel theory which is architecture level parallel, frame level parallel, macroblock level parallel and search window level parallel, we implemented it on GPU with the structure of Fermi GF100which is made by NVIDIA, we improved the encoding speed of AVS video encoder greatly with no loss in video quality.

Related Dissertations
More Dissertations