 |
 |
 |
 |
 |
 |
 |
 |
 |
|
To align the images I used a sum of squared differences algorithms to compare the red and green images with the blue one. Initially I cropped the original image to remove the white and black edges around it. The amount that I cropped it was relative to the size of the image. Then after splitting the image into red, green, and blue images, I cropped the top and bottom of these again to remove any overlap between images or black edges due to unexact splitting. Then I used these images to perform the alignment. I used a pyramid method that halved the resolution of the image until it was less that 400 pixels wide. On the 400 pixel image I used a window of size [-10,10] to compare the two images and found the closest offset. Then on each increasing size of the image I used the previous offset as a starting point and searched a window of size [-4,4]. Here are the results of my algorithm:
|
|
|
|
|
|
|
|
|
|
|
 |
 |
 |
 |
 |
 |
Select images from database:
 |
 |
 |
 |