Briana Brownell

Colorization of Historical Photographs 


History of Color Photography


Black and white photography captures only a single channel of information - brightness. Meanwhile, color photography uses multiple channels of information to reproduce how the human eye perceives color. Renowned physicist John C. Maxwell published one of the early theories for how color photography could be achieved in 1855 and the first color photograph using this method was taking in 1861. 

Despite these advances in chemistry, color photography remained rare for another hundred years, through the 1950's. Many of the historical images we have were captured with a single channel of information - not enough to infer color directly. Artists began trying to colorize black and white photographs by hand, which was a tedious and completely manual process. 

However, we now have access to millions of color photographs and can use deep learning to infer what color might fit a black and white image automatically. This can bring new vibrancy to some of the most important historical moments.

This method, created by Iizuka and Simo-Serra in 2016, uses a convolutional neural network with both prior information about colors of various elements in images combined with local features on the input image to give an intelligent guess for what the output color should be.