Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Study of Video Format Conversion Algorithm Based on Motion Compensation

Author ZhangLiXiao
Tutor HouZhengXin
School Tianjin University
Course Signal and Information Processing
Keywords video format conversion de-interlacing frame rate up conversion 3-D recursive search block-matching (3DRS) bi-directional motion estimation
CLC TP391.41
Type Master's thesis
Year 2009
Downloads 41
Quotes 2
Download Dissertation

With the rapid development of video technology, there are many video formats coexisting in current video field, and the communication among these formats is becoming more and more frequently. So it is necessary to put forward a new topic, video format conversion, which is made up of scan rate conversion, de-interlacing and frame rate conversion. In this paper, the various existing algorithms of video format conversion are analyzed. Based on motion compensation technology; a new de-interlacing algorithm and a new frame rate up conversion algorithm are proposed separately.Firstly, with regard to motion estimation, traditional algorithms are discussed. As a dominating algorithm in video format conversion, 3-D recursive search block-matching (3DRS) with its unique superiorities in vector consistency and hardware implement are expounded. On this basis, a series of improvements are applied to the candidate vectors, scan order and error criterion of the original 3DRS. And in the following work, the improved 3DRS is used as the basic algorithm of motion estimation.Secondly, after briefly analysis of several typical de-interlacing algorithms and their characters, a new de-interlacing scheme based on motion compensation is explored. In this scheme, the improved 3DRS combined with bi-direction motion estimation is performed to get the motion vector. Then vector median filter is introduced to smooth vector field. After the motion state of current block is judged by the size of motion vector, different interpolate methods will be adaptively applied to static block and moving block to reconstruct the missing field. As the experimental results indicate, compared with several traditional methods, the proposed algorithm can gain higher objective quality and better subjective effect.Finally, on the basis of the studies to existing algorithms, a new frame rate up conversion system is proposed. Similar to de-interlacing, the motion estimation of frame rate up conversion is also performed with the improved 3DRS and bi-direction motion estimation. After getting motion vectors, vector median filter is applied to smooth motion vector filed as well. What’s more, stratified block erosion is introduced to reduce the‘block spot’. At last, the missing frame can be interpolated with the median filter motion compensation technology. After the contrast of different methods in objective quality, subjective effect and time complexity, the superiority of the proposed algorithm is proved.

Related Dissertations
More Dissertations