Image Stitching
Alex Hills (ahills)
This project stitches together images by automatically detecting features and picking the best matches using the following process:
1)Detect the harris points
2)Take only the strongest ones (ones with only weak neighbors)
3)Get feature sets from each point (a sampling of the nearest points) and find the best matches between the images
4)Using a RANSAC algorithm, find the best projection matrix from the first image to the second (pick four harris points in each, their matches in the other images, and generate the matrix. Then check all the points and check how many transformed points match the points they're "supposed" to).
5)Compose the final images together.
I also implemented a rectifying method that will take in an image and four points, and project the image such that the specified points form a square in the center of the image.
Results Images
Failed images
