Enhanced Depth Estimation for Stereo Images using SIFT Features

Stereo vision, an essential aspect of computer vision, enables depth perception by analyzing two images of the same scene from different perspectives. This paper presents a comprehensive framework for stereo vision, encompassing calibration, feature detection and matching, fundamental matrix calculation using RANSAC (Random Sample Consensus), essential matrix computation, camera pose estimation, Pre-processing, Bi lateral filtering, edge detection, rectification, correspondence, and depth map generation. Key components include utilizing SIFT (Scale-Invariant Feature Transform) feature detection and a Brute Force/FLANN (Fast Library for Approximate Nearest Neighbors) matcher, employing RANSAC for robust fundamental matrix estimation, and implementing a disparity mapping technique for depth estimation. The proposed methodology is tested on three diverse datasets, demonstrating its effectiveness in accurately reconstructing 3D information from stereo image pairs. By addressing potential patent concerns and ensuring algorithmic robustness, this research contributes to the advancement of stereo vision techniques for diverse applications ranging from robotics to augmented reality.