Colorizing Old Photographs

As a student of photography, the colorization of old photographs is both fun and interesting while making me cringe at the blatant abuse of the code of ethics. I have been studying the history of photography and the photographic processes for a little over five years now, and I cannot help but think back to what my photography professor taught us about the National Press Photographer’s Association’s Code of Ethics. I am specifically thinking of the tenet that says “Editing should maintain the integrity of the photographic images’ content and context. Do not manipulate images or add or alter sound in any way that can mislead viewers or misrepresent subjects.”1 I know that those rules do not necessarily apply to this particular assignment, but I could not help but bring it up. This is why intent and citations are important, otherwise what I have done here could be considered blatant plagiarism and unethical on the grounds that I have edited a photo beyond what it considered appropriate.

For this assignment we looked at a variety of old photographs from the Library of Congress or the Smithsonian Archives and were tasked with picking one to colorize, then putting that same photo through an AI to determine which appears more accurate.

The photo I chose was from the Library of Congress. It was taken between 1899-1900 by Thomas E. Askew, and was a part of a collection by W.E.B. Du Bois. The photo has the descriptive title “Mamie Westmorland, school teacher; half-length portrait, with left hand to cheek, facing front.”2

For the colorized photo I did, I used Pixlr, which is an online program that is nearly identical to Photoshop that you can use in your web browser for free. I chose to use a variety of overlayed color-dodging and burning methods to colorize this photo due to the high contrast of the original. For anyone not familiar with the technical aspects of photography, dodging is a technique used to lighten a particular area of a photograph, while burning darkens an area. Looking at the original, you can see the high contrast between her dark clothing and the photo background and her skin tone. To include more detail in the photograph I had to employ dodging techniques with her clothes, while burning the shadows towards the bottom of the frame. I also had to use dodging on her hair, otherwise it would not have appeared with as much detail as it did. The color I chose for her blouse is called Copenhagen Blue (HEX #313C45). I used this color because according to the Fashion Institute of Technology, both darker and lighter colors were popular. I also did not want to force color into the photo, because her clothing is very dark. The only other color I had to look up was Russet Brown (HEX #8D5524) for her skin tone, which I simply overlayed onto the original picture. The rest of the picture actually does not contain that many colors, and is instead burned or dodged variants of the original black and white photograph. I also used a light colored gradient to lighten the background to a pale cream color used more often in commercial photography settings.

The colorized photo using the Algorithmia AI looks slightly similar yet vastly different from the one I did. I find it interesting that the AI also decided to use a dark blue for her blouse. However, the photograph appears to have been put through a sepia-like filter. This effect does little to differentiate her skin tone from her hair and the surrounding background. Moreover, the AI apparently could not figure out where the photo began and ended, because the sepia effect looks like a halo outside the border. Artistically speaking, the AI did a relatively good job of colorizing the photo. However, it looks a little fake. It appears to have randomly inserted color into the background instead of keeping it constant. It looks almost like the shadows bleed blue, as if someone spilled cyanotype chemicals on the paper and left it exposed for too long.

I may be biased because I take pride in the colorized photo I made, but I think the colorized photo I made looks more accurate. I have been doing film photography for awhile and based on my knowledge in that area along with the techniques I employed make it more consistent with actual photographic techniques. This type of photography would be classed as commercial photography, because of the stagnant nature of the photograph. The camera technology of late nineteenth century would mean still-figure portraits would have to be taken in this style to avoid blurriness. The environment would have to be controlled to greatest extent, which is what I tried to demonstrate in my photo. What I was going for was not accurate in terms of colorization, but accurate in terms of methods and presentation. The AI appears to have lightened and darkened portions of the picture at odd intervals, which is not something that happens when a photo is exposed. It is possible that the AI has been trained on more modern photographs and cannot comprehend how the colorization should be implemented in an old photographic print. Either way, I believe my photo to be better in terms of all these features. I enjoyed doing this project because it allowed me to put my photography skills to use in a very different medium. I hope to use these methods again one day, and plan to use Pixlr again for future projects in other classes.

  1. National Press Photographers Association, “Code of Ethics,” NPPA (National Press Photographers Association, November 28, 2017), https://nppa.org/code-ethics</a>.[]
  2. Thomas E. Askew, Mamie Westmorland, School Teacher; Half-Length Portrait, with Left Hand to Cheek, Facing Front, 1899, Library of Congress, 1899, https://www.loc.gov/pictures/item/99472113/.[]

Leave a Reply

Your email address will not be published. Required fields are marked *

css.php