CS 194-26: Image Manipulation and Computational Photography
Project 1: Images of the Russian Empire -- colorizing the Prokudin-Gorskii photo collection
Brief Overview
Color photography by Sergei Mikhailovich Prokudin-Gorskii involves alignment of exposures onto glass plates using red, green and blue filters. We can produce colored imagery via image processing from misaligned photographs of the three difference color channel images.
My Approach
I took each image and first split the singular image of the three different color channels into 3 different images for each R, G, B color channel. Then I used an alignment algorithm which first cropped each image by 85% (I used the blue color channel as the base image) and compared the normalized cross correlation with different alignments of each image to look for an ideal translation to align the two images. I tried the sum of squared difference but the normalized cross correlation ended up with better results overall - mainly owing to the fact that it was less affected by the difference in color intensity in differing color channels.
For the larger tif files, I rescaled the image using skimage.transform.resize down to below 3000 pixels, performed. Note that for emir.tif I used the green channel for the base and aligned the red and blue channels accordingly - this was because the result of using the base channel as blue was not very good as the color intensity varied greatly (there is a lot of blue in the middle of the photo).
