CS 129 Project 1 Writeup

[Aaron Silverman] (ajs9)
September 2012

For this project, we had to line up three images in order to make a full color image. I solved this problem by blurring the image with a gaussian blur before shrinking i down by a factor of two. I did this blurring and shrinking process four times. After finishing that process, I compared the images in a square of eight pixels around the center. Using this information, I figured out how much I needed to move the image at this coarsest level by using the sum of squared differences and then passed up the displacement value to the less coarse image above, making sure to factor that the displacement is twice as big in the finer image. This is repeated until a final displacement value is obtained at the topmost level.

Results Images

The higher resolution images were less accurate. My program can't detect a higher resolution image, and so it still maintains a depth of 4, when it really should use a much higher depth.

Note: High resolution images scaled down in size for convenience. Some may be slightly warped because of this