Overview

In this project we utilize black and white photos taken with red, green, and blue filters to reconstruct these images into color photos. The main challenge here was to take these separate images and align them well so that the overlap would produce the appropriate picture. The smaller pictures were easier, because one could take a simple approach of looking through a certain range of pixels one by one and comparing the similarity or "success" of these alignments using sum of squared differences of the image matrices. The main problem I ran into was working with the large images. Specifically, lady.tif and emir.tif. Lady was difficult to align at first, because there was a lot of whitespace that could have been easily misaligned. Therefore, when actually comparing the different alignments using SSD, I made sure to only comparing the parts of the photo that were distinctive around the actual lady herself. Emir was difficult because the colors would simply not match up, probably because of the bland background and the complicated clothing. So, to combat this problem, instead of aligning the other colors to blue, I tried to align the different colors to green, which promptly fixed the problem.

Result of Algorithm

All photos have offsets in the format of x, y offsets for green and blue and then red and blue (except "Emir.tif" which has green and blue switched).

cathedral

2,5 3,12

emir

-20,-50 20,60

cathedral

20,60 20,130

cathedral

20,40 20,90

cathedral

10,60 10,120

cathedral

2,-3 2,3

cathedral

1,3 0,7

emir

30,80 40,180

cathedral

0,7 -1,14

cathedral

10,50 10,110

cathedral

0,40 30,90

cathedral

20,60 30,120

cathedral

10,60 20,140

cathedral

-1,-3 -2,-2

cathedral

0,2 1,7

Bells and Whistles

Hello World! This is my first website and I am very excited. I've always wanted to learn how websites work and I'm well on my way to finding out!