There are methods, using tools like Photoshop, to manually select the foreground of an image and decolor the background, drawing greater attention to the in-color foreground (as shown in the image below).
However, we are interested in implementing an automatic program that distinguishes between the background/foreground of an image and converts the background to grayscale.
Motivation
UW-Madison academic calendar for the month of November.
Inspiration
Approach
We determined methods using defocus blur to detect the foreground to be too complex given the time constraints of this project, so we decided to use k-means clustering based on color.
We had considered face-detection techniques, as well, but did not want to limit the object of the foreground to people. Also, allowing the user to select the subject would defeat the point of an automatic method.