| Summary: | 碩士 === 國立高雄第一科技大學 === 電腦與通訊工程系 === 89 === In video coding, motion estimation is an efficient predictive coding for successive digital image sequence. Motion estimation was adopted by many international standards for video coding such as H.261, H.263, MPEG-1, MPEG-2. Motion estimation is a predictive technique which explore temporal redundancy between successive frames of video sequence. Block matching is an efficient and simple method to perform motion estimation. By dividing each frame into rectangular block, motion vectors are obtained via block matching algorithms（BMA）. The kind of method is the most common one.
In this thesis, two new methods are presented. One is the Threshold Two-path Search block matching algorithm（TTPS）. The TTPS algorithm improves the traditional BMAs significantly. In the traditional BMAs, they follow one searching path that defined by its search rule to search the best motion vector. The TTPS algorithm has two search paths. It can get better motion vector than the traditional BMAs. There is a threshold in each search procedure, and the threshold is set by designer advantageously. When the compared value higher than the threshold, another search path is added to search motion vector. It gives much performance than tradition and just increases little computational complexity.
Another motion estimation method based on a genetic algorithm to perform motion estimation processing. A novel algorithm is proposed in this paper. It combines genetic algorithm（GA） and diamond search algorithm（DS）. Initial operation and mutation operation uses the characteristics of DS. The proposed method takes advantage of GA and DS.
|
|---|
