The idea for FaceBlender stems from a discovery channel documentary I saw somewhere in the late 90ies. It stated the theory that our brain averages all the faces we see in our brain, and subsequently compares a new face to the average it has created. If the new face resembles the average more, then it is considered more beautiful.
In response, I decided to write a computer program that can give me an idea of what this average would look like. In 2000 I wrote a C++ Windows program that I lated ported to an iPhone app called FaceBlender that has been in the app store for a few years until I discontinued it due to lack of time to keep it up to date and modern. The unique feature of the app is that it is designed to blend huge numbers of images instead of what you will usually find in face blending software, namely only 2.
The latest version of the software works by feature detecting the mouth and eyes of multiple portrait photos and averaging the distance between the eyes and the eyes and the mouth. Then it automatically transforms all the images so that the eyes and chin overlap perfectly.
This gallery shows several examples of output of my program.
In the image above you also see a resuling self-organizing Kohonen map which classifies the types of portaits.