Dense depth estimation using adaptive structured light and cooperative algorithm

In this paper we propose a new depth estimation approach using adaptive structured light. A random noise adaptive structured light pattern is projected onto objects and then two cameras capture stereo images. The adaptive colors are acquired using principle component analysis in the RGB color space of the image of the scene. By using inverse principle component analysis on the images with structured light, the desirable structured light information can be maximally retrieved. By combining the original three RGB channels of the scene under adaptive structured light with a fourth channel generated using inverse principle component analysis we can use the cooperative algorithm to generate a dense depth map. In order to keep clear depth discontinuities and alleviate noise in the depth map, we aggregate the local match score with shiftable windows. Experimental results show our approach performs well on images of real-world objects with strong colors and complex textures that have been captured under ambient light conditions.