For this project, multiple photos were taken from one location, and after that, were stitched together into a single image.
The first step is to compute homographies, and to write a function to morph images with the recovered homographies.
This step is shown independently below. The images were taken at an angle, and then adjusted to have square things be square.
GHC 3k light buttons
GHC 3k 'Thanks Red Hat!' sign
A particularly interesting result with floor tiles
Some actual image morphing. I haven't gotten the image blending yet, so for now it is just the two images at 50% on top of each other, with one odd bug in the middle.
Sixth floor of Gates: This worked fairly well.
Panther Hollow bridge: There are fewer source points, because it is hard to match trees, so the images don't line up nearly as well.
The top of my dresser at home. Once again, fewer points, so a worse homography.
The following image was done with automatic data point selection, and the stitch NaN bug fixed:
