Project Summary

In this project, we leveraged Markov Random Fields (MRFs) to tackle the complex stereo correspondence problem, aiming to construct a disparity map from stereo image pairs. By framing the stereo problem as an MRF inference challenge, we developed a robust stereo energy function combining a unary cost function and a pairwise distance function. Our implementation progressed through two phases: starting with a Naive approach and evolving into a sophisticated 'Loopy' version to enhance accuracy and efficiency.