Moving object segmentation using graph cuts
Jia Wang, Hanqing Lu, Gérard Eude, Qingshan Liu · 2005
This paper presents a new algorithm for segmentation of moving objects in image sequence. In this method, motion segmentation problem is considered as an image-labeling process in which every pixel in the image will be assigned a label describing its motion between consecutive frames. We formulate this vision problem in terms of energy minimization. Energy function is defined to measure the extent of smoothness and accuracy of motion vectors. Then, graph cuts algorithm is used to settle this energy minimization problem. Several experimental results illustrate its promising performance.