A basic definition of face averaging is taking the structure and texture average of multiple faces, which yields the average face.
Every individual has their own "individual set" of faces and the "crowd set" has faces from every individual, but any unique image of a face for any individual belongs to only one set.
|
Face Recognition Wants You Looking Your Best: Applying Face Averaging to Face Recognition
Advancements in the field of face recognition technology have turned toward a technique called face averaging. Face averaging is the process of creating an "Average Face" from a set of photos of a single person. Face recognition systems that use an average face instead of a regular photo to identify an individual can improve results by about 25%.
In my research paper, I explore a drastically faster approach toward face averaging which avoids the expensive procedures of the main face averaging method while maintaining an improvement on recognition results. In the process, I also create graphical software that allows for efficient testing and comparison of face averaging methods' improvement on face recognition performance. In the future, my fast alignment averaging method may be applied to real-time video of an individual person. A live-average can be kept and used for real-time recognition of that individual.
|
|
Matthew Thomas '09
|
|