This is a depth fusion technique using Gaussian Distribution-based pixel depths and Kalman updates. By leveraging Pose-derived baselines, it establishes epipolar lines, employs NCC for block matching, calculates depth using baselines, and fuses the data via the Kalman Equation.
This method offers a comprehensive approach to refining depth estimation from multi-angled scene images.
The Results that we obtain is as follows
Ground Truth | Result |
The helper functions for reading/writing were written by Dr. Gao. I just rewrote the implementation for learning purposes :)