Detection, Recognition, and Expression Analysis of Human Faces

Automated processing of facial images is a problem of interest since the beginning of computer vision. Recently the interest has revived for biometrics and human-computer interaction. I am going to talk about our recent progress in face image analysis: detecting faces in an image with arbitrary size, pose and background, understanding facial actions units of FACS to understand facial expression, and recognizing people by face images. Also, I will talk about "recovering" a higher resolution image of the face from an extremely low resolution image (the process we nicknamed "hallucination"). Common to these methods is a combination of probabilistic approach and critical care of registration in order to obtain better results than previous attempts.