While human perception is fine-tuned to detect and recognize face images with great precision, existing computer algorithms for face recognition still perform worse than humans. Can we learn from human perception to improve the performance of face identification by a computer? Or, is it that such a biologically-inspired approach, or biomimetic, is not a good idea at all? In this talk we will start by introducing some interesting facts in human perception of faces. We will then present examples of image analysis techniques that are inspired by human perception. Comparing these techniques with face recognition algorithms based on conventional pattern recognition techniques, we will outline some promising research directions.
Tuesday, April 15, 2008
Free and open to the public