The Fundamental Matrix encapsulates the geometric relationship between corresponding points in stereo images, providing essential information for rectification, stereo reconstruction, and 3D scene understanding.
The Eight-Point Algorithm is a robust method for estimating the Fundamental Matrix from point correspondences between stereo image pairs. With the complements of RANSAC, Eight Point Algorithm can effectively find the best fundamental matrix estimation to fit relate the stereo data keypoints.
| Left Image | Right Image | |
|---|---|---|
| Original |
|
|
| Key Points |
|
|
| Epipolar Line |
|
|
| RANSAC Epipolar Line |
|
|
The main code is in eight_point_algorithm.ipynb
The project uses python==3.12.2, and the dependencies can be installed by running:
pip install -r requirements.txt
- In Defense of the Eight-Point Algorithm
- Revisiting Hartley's Normalized Eight-Point Algorithm
- Intrinsic and Extrinsic Matrices | Camera Calibration
- Overview | Uncalibrated Stereo
- Problem of Uncalibrated Stereo | Uncalibrated Stereo
- Epipolar Geometry | Uncalibrated Stereo
- Estimating Fundamental Matrix | Uncalibrated Stereo
- Finding Correspondences | Uncalibrated Stereo