Fast hierarchical face detection

We describe a new face detection algorithm based on a hierarchy of support vector classifiers (SVMs). Some preprocessing steps reduce the number of candidates, which should be verifying by SVMs. Hierarchical SVMs are designed for coarse-to-fine search of human face. In the SVMS, eye classifiers with different complexity (measured by the number of support vector) are used to find the eye candidates. The gain for efficient computation is enormous. The proposed face detection system can also be used to detect multiple faces embedded in complicated background. In addition, this approach can be applied to different kinds of image variations such as face sizes, lighting conditions, and image qualities.