Stereoscopic Video Object Segmentation Based on Disparity Map

Stereoscopic video object segmentation plays an important role in stereo vision analysis and application. In this paper, a new method for disparity image segmentation derived from the extended mean shift theorem was proposed. Disparity estimation based on adaptive support weight approach was firstly introduced to get reliable depth information. Then an extended mean shift algorithm was exploited to achieve initial segmentation of depth image and coarse object areas were obtained. Finally, accurate objects were segmented according to the marking of other adjacent pixels. The experimental results show that the proposed method can segment the semantically meaningful objects successfully.