Depth map enhancement method based on joint bilateral filter

Abstract

Considering the emergence of Microsoft Kinect has attracted attention not only from customers but also from researchers in the field of computer vision. Nonetheless, Kinect-generated depth maps usually contain various holes, especially at the edges of the front ground. In order to fill in the holes existing in the depth map, some methods that adopted color map guided in-painting or joint bilateral filter have been proposed to obtain missing depth pixels. However, these methods often caused edge blur. In this paper, the method is proposed that the holes and noises of the depth map are filled by the weighted joint bilateral filter and the fast marching algorithm based on the edge priority is further used to keep the edges smooth. To efficiently validate our method, we perform experiments on both the Kinect data and the Middlebury dataset, which provide qualitative and quantitative results. Experimental results show that our method has obtained the best performance of improvement of depth map for the smooth and edge regions and can perform real-time processing.