Accessing the Data
The Wayback Machine was our source for the images. Initially, we downloaded the link to every screen-capture we wanted to use, but then our data also included the text. The image processing software that we used to make the histogram, ImageJ, does not like analyzing text so we needed another way to extract just the pictures.
We ended up going through and taking screenshots of all of the images instead. In the recent years, there have been up to 100 screen-captures saved on the Wayback Machine every year, so instead of going through each one, we chose one screen-capture per month, per year. We decided to collect our data on a per-month basis because we assumed that the images are probably not replaced any more frequently than that. Furthermore, because we were doing everything manually, we limited our data to only the images that were on the front page, not the ones that appeared on subpages of carleton.edu. Even with these restrictions, we built a dataset with a total of 136 images. We then added all of them into a desktop folder.
Processing the Data
We used the software called ImageJ to perform color analysis and to create our graph. ImageJ is a Java-based processing software that uses macros to organize, analyze, and edit image datasets. We downloaded an ImageJ bundle file, which included both the actual software, as well as the txt files “ImageMeasure” and “ImagePlot.”
First, we used ImageMeasure to open our dataset. ImageMeasure is an image-analyzing macro that produces a txt file containing numerical values for the brightness, saturation, and hue of each image that you open in ImageJ. With the txt file that we got from opening our dataset in ImageMeasure, we then ran it through the macro ImagePlot. ImagePlot allows you to make a histogram with images as the data point icons. So, we created a histogram with the images in our dataset as the icons, the saturation values from the txt file on the y-axis, and the year in which the image appeared on the front page on the x-axis.
We also created a histogram with brightness values on the y-axis, but we did not see any notable patterns. Based on this, we think that Carleton has not placed as much emphasis on the brightness of the pictures they use, and instead focused on the content and tone of the images.
