Project __ Writeup
Rudy Sandoval (rudy)
January 29, 2010
My project involves the alignment of different color channel images to produce a color image with as little chromatic abberration possible. This was done using a gaussian pyramid of the images and searching upward from smallest to largest image, in order to produce a refined search without exhaustively spanning the original image space. Sum of squared differences was used to evaluate alignment between the different images, and tended to work for most cases. While a smart cropping was not implemented, a constant percentage was cropped from the image to remove extremely high and low values that would throw off color-contrasting and white-balancing. These, along with the application of my algorithm to a randomly shifted image of my choosing make up the extra credit that was done for this project.
- Extra Credit Extension #1: Color-contrasting by manipulating the color range such that it spanned 0 to one, the full possible values for an image.
- Extra Credit Extension #2: White balancing by finding the brightest color pixel in the aligned image, and if variance between the color channels were low, adjust the image such that this pixel became white.
- Extra Credit Extension #3: Application of the algorithm to an image from flickr. The color channels were randomly shifted to test the quality of alignment.
Results Images
