Project 5

Part A

Sorry, I’ve run into a bit of a time crunch. I mostly finished part 5, with my two images being


The combined image is

still to do:

  1. linear blending between images. Currently I am using just a max() between them… but it doesn’t look too bad
  2. automatically finding bounding boxes of images. I just added 1000 pixels to each side of the image and called it a day, but I can do better than that by resizing appropriately by checking the box of the source image after transformation.

Part B

Identifying the features was done by using the staff solution:

But, to properly choose the strongest features, I used ANMS and ordered the features by their strength. The 200 strongest feature points are these:

The features were then extracted using the 40x40 box around each point. This box was resized to 8x8 to make computation easier and normalized. This is an example feaature: