Dense depth recovery based on adaptive image segmentation

Stereoscopic vision has been a hot research topic in recent years. As one of the most important steps for 3D reconstruction from image sequences, dense depth recovery has attracted much attention in computer vision recently. A lot of efficient algorithms based on stereo matching have been proposed. But the difficulties in depth recovery are not overcome completely yet. In this paper, we aim at the main problems of depth recovery and propose our solutions. And an effective approach for dense depth recovery based on adaptive image segmentation is also presented. Experiments show that accurate depth results can be achieved through the proposed approach.