The goal of the project was to get familiar with warping and stitching images together into panoramas.
Took two pictures of my kitchen and marked some 10 corresponding points so as to facilitate the warping.
The next step was to find the unknowns - 8 degrees of freedom and create a homography matrix that would be used for warping.
The next step was to warp and rectify the image into a more compatible field of view.
Finally we blend the warped image with the original image to create a mosaic.
First we find the Harris corners- as we can see we have several harris corners
Then we select the best 500 using ANMS and use feature descriptor matching to match the points in both the images
Then we use the RANSAC algorithm to sample pairs and get the largest subset of inliers and use that to compute the homography. Then we blend the images and get the result. Below is a comparison between the manual and automatic blend - they look quite similar, neat!
Conclusion: This project taught me some very great image processing skills. Learning the math and the code behind how to create panoramic pictures from separate images with different views was all quite interesting!