Fast Detection of Multi-View Face and Eye Based on Cascaded Classifier

In our multi-view face and eye detection, we use a cascaded classifier trained by gentle AdaBoost algorithm, one of the appearance-based pattern learning method. Specifically, in order to detect multi-view face, we propose a special cascaded classifier using coarse-to-fine search, simple-to-complex search, and parallel-to-separated search. In order to detect eye, we propose a four-step eye detection method. Using proposed methods, we got face and eye detection ratios as 99.5%, 88.3%, respectively for 7 different DBs including 17,018 various multi-view faces.