Throughout this exercise, I developed a complete stereo vision pipeline capable of reconstructing a 3D scene from two camera images. Starting with the capture and visualization of the stereo images, I applied preprocessing techniques like gamma correction to improve visibility and enhance edge detection using the Canny algorithm.
Once the interest points were extracted, I implemented a reproducible selection method to work consistently with the same inputs. From there, I created a custom patch-matching function to establish correspondences between both images, which laid the foundation for triangulating points in 3D space.
After refining the triangulation logic and incorporating the true position of the cameras, I achieved a final visualization where the depth information is correctly represented, and each point is colored based on the images. The resulting 3D output clearly reflects the structure of the observed scene and confirms that the reconstruction process works effectively from start to finish.