Sveinung Myhre, s.myhre@berkeley.edu
The project utilizes the historic photos taken by Sergei Prokudin-Gorskii,
a visionary who captured the Russian Empire in color long before modern
color photography existed. Using his method of capturing
three images through red, green, and blue filters, one can reconstruct
his black-and-white images into color photos.
The program loads an input image and divides it into three separate color channels: red, green, and blue. Image adjustment is dune using pyramid. Each channel is adjusted to achieve proper color alignment by using an image pyramid (a series of images with gradually reduced resolution). Some different metrics vere used but the l2_norm was found to be sufficient. After alignment, borders are removed using the remove_borders() function. The adjusted color channels are assembled into a color image that is saved and displayed. Translated with DeepL.com (free version)
Below are the results of the algorithm applied to various images:
The next 13 images are based off of source images provieded at https://inst.eecs.berkeley.edu/~cs180/fa24/hw/proj1/data/