Depth Estimation of Stereo Matching Based on Microarray Camera

Depth information affects greatly the accuracy of image measurement, 3D reconstruction and image recognition. In general, the methods of obtaining depth information are from 3D laser scanners, structured light, and depth cameras. The traditional method using binocular camera to obtain the depth information of images is based on the disparity between the left and right views, but it also brings some problems such as the occlusion area and mismatched points. To improve the accuracy, we proposed a novel method of depth estimation of stereo matching based on microarray camera. First, each lens in the camera is calibrated to compute the intrinsic and extrinsic parameters, which are used to rectify the captured images respectively. Then stereo matching between images is modeled by a Markov random field, and the energy cost function for the MRF is built and the minimization of it is solved with Graph-Cuts algorithm. Finally, the matching result is obtained with depth information simultaneously. In the minimization procedure, level division of depth is incorporated, and the gradient information of the reference image is utilized to refine the depth layers of the corresponding image. Experimental results showed the efficiency and accuracy of the proposed algorithm.