A moving objects extraction method based on true motion information
Hui Yuan, Yilin Chang, Yanzhuo Ma · 2008
In this paper, a moving objects extraction method based on true motion information is proposed. First of all, three dimensional recursive search algorithm (3DRS) combined with median filter is used instead of full search in searching for motion vectors (MVs) which reflect the true motion of objects in video sequences. After that a new parameter segmentation factor (SF) is constructed to describe motion feature of each block in a picture. Thus, moving objects extraction with the fuzzy C-means clustering algorithm (FCM) over SFs is implemented, during which gathering degree is used to distinguish the foreground which refers to the moving objects from the background. Experimental results show that this method is effective for still scenes or scenes with consistent global motion.