Computational Photography -- Project 1 -- Kelleker Guerin

Overview:
This project delt with creating a composite RGB image from a glass plate containing three seperate exposures, each taken with a different color filter. The challenge lies in correctly aligning the three channels to one another so the resulting image looks correct. There were two types of images to be processed using this technique. Small jpegs and very large .tiff files. For the larger files it was necessary to create an image pyramid consisting of smaller and smaller versions of the original image. The alignment algorithm was then applied to each of these pyramids, starting with the smallest first. This avoided a large scale search of the full sized image, instead searching in progressively smaller areas as the image size increased, and updating the position from the previous size. These images were cropped before processing to improve matching. Note, the bells and whisles implemented in this assignment was the cropping of the smaller images. This same method would have worked on the larger images, but the cropping algorithm i developed turned out take a prohibitively long time to run on them.

Small scale Jpegs
----------------------------------Unaligned----------------------------------------------Aligned---------------------------------

00125v

00149v

00153v

00154v

00163v

00270v

00398v

00564v

01167v

31421v

Additional Jpegs found from previous year's data file (Note 56 is pretty interesting with the non uniform degredation of the film, but you can still make out the color image)
----------------------------------Unaligned----------------------------------------------Aligned---------------------------------

00017v07

00056v07

00084v07 Large Tiffs, resized to smaller size, with jpg compression. The pyramid system had alot of trouble working with these images, and the alignments were much poorer than the smaller images.
----------------------------------Unaligned----------------------------------------------Aligned---------------------------------

01861a

01657u

01047u

01043u

00911u

00458u