Stereo Vision

Stereo vision, aiming to reconstruct 3D scenes from two or more 2D images, is a crucial area of computer vision research. Current efforts focus on improving depth estimation accuracy, particularly in challenging scenarios like occlusions and low-light conditions, often employing deep learning architectures such as transformers and Siamese networks, along with classical methods like Semi-Global Matching (SGBM). These advancements are driving progress in diverse applications, including autonomous driving, robotics (especially for manipulation and navigation), and medical imaging, where precise 3D scene understanding is critical for improved safety, efficiency, and diagnostic capabilities.