Model Based Object Contour Detection For Image Sequence

• Main Authors: Yu Yung Lan, 藍昱詠
• Other Authors: Yong-Ren Huang
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/14621787028400946053

碩士 === 樹德科技大學 === 資訊工程系碩士班 === 103 === This paper presents an object contour detection using the object shape model. First, we detect the change of background luminance ratio outside the bounding box of object. Second, the temporal segmentation is obtained by mesh based motion activity detection. A new combined mask is composed of temporal segmentation result and shape model by OR operation. Finally, for modification of the mask, we detect the non-matching points with local gradient along the boundary of combined mask in a counting window. And the MSE is calculated for the detection of the matching points. The video object contour will be obtained by the iterative algorithm.