Simultaneous Estimation of Super-Resolved Image and 3D Information using Multiple Stereo-Pair Images

This paper examines a simultaneous estimation of both high resolution (high accuracy) 3D information and high-resolution image from multiple low-resolution stereo-pair images. The proposed method is based on the MAP (maximum a posteriori) framework and these two relevant problems (estimation of both high-resolution 3 D and intensity images) are incorporated into a unified cost function. Then the solution is obtained iteratively by a fixed point algorithm, which yeilds 3 D information while the high-resolution image is fixed, and vice versa. Each solution can be obtained by solving a linear equation. It can be expected that the proposed method improves accuracy of both 3 D information and high-resolution image simultaneously. Some experimental results are shown to validate the proposed method.