Computer Vision News - June 2016

From the above figure we can see a periodic pattern repeating itself every 11 points. These 11 points correspond to the 11 image per person we have in the Yale face database. In addition, you can see two dominant peaks in almost every period. What/why are they? Let’s go to the corresponding image in the database to see what happened - for example image number 69. Computer Vision News Tool 15 Image number 69: a face with a big left light (shadow) on it. Examination of the other peaks will reveal the same left or right light image faces with (shadow). In other words, those peaks we observe in the A13 correspond to the shadowed images. Oh, wait a second! We see that near point 57 one peak is “missing”! What happened? Inspection of this image will reveal the following: a face image with a very small shadow - almost like no shadow at all.