Programming Project #4-A - Atsunobu Kotani (akotani@berkeley.edu)

First, I am showing the results for image rectification. The following images were used for this set of experiments.

And the keypoints are selected as it follows.

The rectification results are:

Next, I used the following images for the homography experiments.

We do two warping. Left image + Middle image and Middle image and Right image. Definining the key points:

With the correct computation (or least square optimization), we have homography matrix, and the warped images are shown below.

Blending them together:

Adjusting the alpha values for each image a little bit, we have:

Other case:

I also wanted to study if this thing works with the close-up views.

There are several lessons that I learned.

The consistency in lighting condition does matter to maintain the quality of the generated image. I would say I was not successful in maintaining the same lighting condition when I was taking the picture.
The simple blur does not work.
It works quite well in close-up images as well. (it makes sense as we have a nice projection matrix.)