Color face recognition under various learning scenarios

In this paper, the usage of the YCbCr transformation in the input stage of a Face Recognition (FR) system is examined. Chromatic sampling ratios used in MPEG standards have been applied on the input face images to examine a possible integration of video and FR systems. Experiments were conducted in a number of learning scenarios and imaging conditions, and it was observed that almost all the YCbCr inputs lead to good FR systems. In addition, the conditions in which color inputs enhance the FR performance over gray scale were evaluated both for supervised and unsupervised FR systems since the computational time and storage space involved in color FR systems is high.