Image segmentation using temporal-spatial information in dynamic scenes

The goal of image segmentation is to identify homogeneous regions in images. A common method for segmentation of moving regions in image sequences involves "background subtraction", namely, thresholding the error between an estimate of the image without moving objects and the current image. The numerous approaches to this problem differ in the type of background model used and the procedure used to update the model. One possible approach to this problem is to model the attributes of the pixels. In order to segment a larger image in reasonable time and obtain better results, an algorithm of segmenting background based on temporal spatial information of pixels is proposed in this paper. Firstly, adaptive pixel models are modeled to describe the recent history of color at each observed pixel. Then each pixel is classified as background or foreground according to pixel models and the spatial relation between pixels. Finally, parameters of pixel models are updated using on-line EM algorithm. Experimental results show that our approach is suitable for segmenting foreground from background in dynamic environments.