A Discussion of Optimization about Stereo Image Depth Estimation Based on Multi-baseline Trinocular Camera Model

The huge computational complexity and occlusion problems make stereo matching a major challenge. In this work, we use multi-baseline trinocular camera model to accelerate the stereo matching algorithms and improve the accuracy of disparity estimation. We propose a special scheme named the trinocular flexible disparity searching range (FDSR) to accelerate the stereo matching algorithms. In this scheme, we optimize stereo matching by reducing the disparity searching range. Based on FDSR, we proposed the FDSR-MCCNN for trinocular stereo matching. According to the evaluation results, the FDSR-MCCNN could not only reduce the computational complexity but also improve the accuracy. Moreover, the optimization schemes we designed can be extended to other stereo matching algorithms that possess pixel-wise matching cost calculations and aggregation steps. We proved that the proposed optimization methods for trinocular stereo matching are effective and that trinocular stereo matching is useful for either improving accuracy or reducing computational complexity.