Project 6 - Image Stitching

Ankit Mathur

In this project, we focused on stitching images together by finding homographies to map points from the space of one image to the space of another image. Then, we transform these images to lie in the same plane and are able to show these side by side after performing some blending.

Homographies

In order to compute the matrix for the homography, we can start by noting that the equation for the transformation is p' = Hp. Since our points are only in 2 dimensions, we can ignore the fact that we have a third row by using the last element of that row as a scaling factor. Then, we need to solve for the other eight elements in the matrix.

To do that, we expand the equations here:

Rectification

Here, we need to test the homographies that we have made so far. Therefore, we test this by picking a bounding box over an object and calling the warp function in order to change the perceived shape/orientation of the object. We do this simply by calling warp.

Here are some examples:

Mosaics

GBC

Gardens

Telegraph

As an additional note, in my code, I included a horizontal blending function which allowed me to get rid of black elements in the image that showed up from the stitching.

In this project, I learned about the usefulness of homographies. In addition to that, I learned that mapping across the space of images to create stitched images is really hard. I also learned that I really enjoy the art of putting images together to create a beautiful final product.

( Made with Carrd )