Object segmentation using image disparities and intensities

This paper presents a hybrid algorithm for joined disparity segmentation and intensity segmentation in object segmentation for stereoscopic vision system. The disparity maps of stereo images are firstly estimated by using the foreground-based disparity estimation algorithm. Then the intensity stereo image and its corresponding disparity map are segmented based on region-growing technique separately. The errors of the two. segmentation procedures can be eliminated by combination processing on the two segmentation results, thereby reliable object boundaries can be extracted. Computer simulations show that object segmentation referring depth information is worthy to be studied especially for various three-dimensional applications.