Depth Estimation based on Stereo Camera Pairs of Color and Infrared using Cross-based Local Multipoint Filter

This paper proposes a method of acquiring threedimensional shape data about moving objects based on stereo camera pairs of the color and the infrared by projecting infrared dot patterns. Most of the methods that project infrared pattern estimate a single depth map by measuring the translation of a known pattern with an infrared camera. The aim of the proposed method is to make 3D models by integrating depth maps from multiple viewpoints for 3D video production and 3D imaging. The method utilizes the infrared pattern as a texture and estimates depth maps by stereo matching of an infrared image pair and a color image pair. The matching cost is integrated by the weighted sum of the matching cost of the infrared image and the color image. Before the disparity candidate with the lowest cost is selected, cost volume filtering is applied to each cost map by using the cross-based local multipoint filter (CLMF). We present experimental results showing the effectiveness of our proposed and describe to apply it to three-dimensional integral imaging. Keywordsdepth estimation; stereo matching; cross-based local multipoint filter; integral imaging; elemental images