Enhancement of stereoscopic depth estimation by the use of motion information

The method presented in this paper uses motion information to complement the state-of-the-art depth estimation technique. The use of motion field correspondence between different views in the matching cost function for disparity computation is proposed. In addition, edges from motion maps are used to modify smoothing cost function in the algorithm in order to preserve sharp edges of moving objects and smooth transitions in the background or steady areas of depth map. Experimental results show that our modified algorithm significantly increases quality of depth maps in local areas. The technique is particularly interesting for applications where motion estimation is also needed for other purposes.