Depth Estimation with Multi-camera Array

The reduction of camera cost makes it possible and necessary to use camera array in complex scenes. The depth estimation method with multi cameras is proposed and a verification system of four camera array is designed in this paper. The projection geometry of the camera is described and the depth estimation based on least squares algorithm is derived. The random sample consensus (RANSAC) algorithm based on multi-views geometry is used to choose inliers. The result shows that the depth estimation errors of the multi-views stereo are much smaller than the binocular stereo method. More importantly, the multi-views stereo solves the spatial aliasing problem. The redundancy configuration of the camera array makes the depth estimation more robust and accurate.