Cs1942-26 Aaron Chen Project 5

Part A: image warping

In this part, we were stitching together two overlapping images. This was done by manually selecting points that were the same in both images. Then, we used these to calculate a homography by using a system of equations. Once we had the homography, we could then warp an image into the other images system in order to stitch them together. With linear blending, you can see below that the stitch is pretty good.

Part B: Auto stitching

In this part, we first used Harris detection to find interesting points in both images. You can see it below;

Once we have this, we use ANMS to pick the most interesting points. Then, we auto match features by sampling a window around each point and checking to see if the features error is below some threshold, where the error is the ratio between the best and second best nn error.

Unfortunately I could not produce a reasonable output, but I attempted feature matching & the RANSAC algorithm.