Image Averaging

I taught an undergraduate course on digital history methods at Huron University College in the fall of 2013. We focused on topics from American popular culture as source materials to experiment with those digital tools. Students had a great deal of freedom in the topics they chose for their larger independent projects in that class. Quite a few students chose to work with sports-related topics.

In particular, one student was interested in the Super Bowl and its marketing over time. I helped her find digital source materials for that component of her work, and one small set we found quite interesting were the Super Bowl program covers over at the NFL website. But in what ways could a series of images be analyzed beyond just laying them out one-by-one for examination?

Image averaging is a useful technique where individual pixel values are averaged across images and the results constructed into a new image. An excellent introduction to the technique and its uses can be found in Form+Code, by Casey Reas, Chandler McWilliams and LUST. They describe that, “By repeatedly combining related images, one can expose behaviorial norms, reveal expectations, and find connections that were less obvious when the images were viewed as a series separated in space,” (p. 83) and they point out some excellent examples by artists Jason Salavon and Michael Najjar.

We decided to try the technique across the Super Bowl program covers by just grouping them into 10-year spans. The results are below. In her work, the student chose to focus on the apparent shifts towards the imagery of the trophy from the images of players and places. The trophy took central prominence, with a more uniform focus in the final series.

superbowlcoveraverage1-9 superbowlcoveraverage10-19superbowlcoveraverage20-29 superbowlcoveraverage30-39 superbowlcoveraverage40-47

To accomplish this averaging, we used ImageMagick. There’s an “evaluate-sequence” function with a “mean” option that will allow you to take a group of images, average them, and create a file of the results.

Imagery can be a very useful source for analysis by historians, and as more collections of digitized imagery become available to historians, I’m looking forward to seeing what we can do with those collections.