CS194-26 Project 6A: Image Warping and Mosaicing Part 1
Recovering Homographies
Before warping our images into alignment, we needed to compute the homography, H. We first picked points from 2 images that corresponded to each other. Then, we computed H by solving a least squares problem, with H being a 3x3 transformation matrix. H is as shown in this image, with i as 1.
Image Rectification
I rectified images by picking the corners of what I wanted to straighten out and 4 arbitrary points that I wanted to morph the object into (in this case, a more straightened out version of the object).
Original Notebook
|
Warped Notebook
|
Original Sign
|
Warped Sign
|
Mosaics
In this section, I created some mosaics. I padded the images that I wanted to use and I had 1 fixed image in the center and warped the second image around that. Then, I used alpha blending to stitch them together.
Yard Image 1
|
Yard Image 2
|
Yard Mosaic
|
TV Image 1
|
TV Image 2
|
TV Mosaic
|
House Image 1
|
House Image 2
|
House Mosaic
|
Summary
I really enjoyed this project because it showed me that you can create pretty seamlessly mosaic-ed images with just some linear algebra and some alpha blending. I also learned that even being slightly off when picking points can result in a blurrier image than what you anticipated.