Depth map enhancement based on adaptive joint trilateral filter

A new generation of active 3D range sensors, such as time-of-flight cameras, enables recording of full-frame depth maps at video rate. Unfortunately, the captured data are typically contaminated by noise and have only a rather limited spatial resolution. We therefore present a novel method for depth map enhancement. Our work is based on an adaptive joint trilateral filter which performs on each slice of cost volume and then the final solution is determined by seeking the minimum cost in a WTA framework. Thus given a high resolution color image and its corresponding low resolution depth map, we can improve the quality of depth map by increasing its spatial resolution, suppressing noise and preventing unwanted artifacts like texture copying into smooth geometry. Experimental results show that depth map obtained by the proposed method provides satisfactory quality. © 2013 ISSN 1881-803X.