This project is about estimating the disparity between two images when one of the image is taken by slightly moving towards the right. By knowing the focal length of the camera and the baseline distance, we can accurately determine the 3-d location of the point. This project aims at implementing some of those Stereo-Matching algorithms in Python.
Input images are Left and Right correspondingly.
LEFT Image | RIGHT Image |
---|---|
![]() |
![]() |
Block-matching Disparity | Energy-Minimization via Dynamic Programming |
---|---|
![]() |
![]() |
This section is about finding the Epipolar-Lines given a correspondence pair. Essentially, we estimate the FundaMental matrix such that x1.TFx = 0. The famous 8-point algorithm is used to find the F matrix. Link to paper- https://www.cse.unr.edu/~bebis/CS485/Handouts/hartley.pdf
The resulting epipoles are shown below.
VIEW-A | VIEW-B |
---|---|
![]() |
![]() |
Based on the fundamental-matrix estimated, we can rectify the images such that their epilines lie horizontally.
Run the StereoRectification/Stereo_rectify.py
to rectify both the images.
The resulting Rectified views are shown below.
RECTIFIED VIEW-A | RECTIFIED VIEW-B |
---|---|
![]() |
![]() |