Moving Object Segmentation in Video using Spatiotemporal Saliency and Laplacian Coordinates

This paper presents a new algorithm for automatic segmentation of moving objects in video based on spatio-temporal saliency and laplacian coordinates (LC). Our algorithm exploits the saliency and the motion information to build a spatio-temporal saliency map, used to extract a moving region of interest (MRI). This region is used to provide automatically the seeds for the segmentation of the moving object using LC. Experiments show a good performance of our algorithm for moving objects segmentation in video without a user interaction, especially on Segtrack dataset.