A New Motion Compensation Algorithm for Video with Camera Zooming Motion

• Main Authors: Wen-Han Wu, 吳文瀚
• Other Authors: Long-Wen Chang
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/90332923045735701364

碩士 === 國立清華大學 === 資訊工程學系 === 94 === Camera zooming motion is very common in both commercial movies or home made videos. However, since the property of the moving motion caused by camera zooming, block based motion estimation method in current video coding standard couldn’t process video sequences with camera zooming motion well. Base on property of camera zooming motion, we proposed a zooming motion estimation method which is different from present global motion analysis method. If we estimate that there is a camera zooming motion, we can further estimate the camera zooming speed. We propose a camera zooming motion compensation algorithm which can estimate the motion vector of each pixel by the camera zooming speed to reconstruct whole frame. We further propose a combined method which combines our camera zooming motion compensation algorithm with traditional block based motion estimation method. The combined method not only has high visual and image quality, but also improves the coding efficiency compared to block based motion estimation. Key words: camera zooming motion; global motion analysis; block based motion estimation; motion vector; motion compensation.