A fast detection model for Omni-directional faces

In order to solve the problem that Omni-directional faces, which was in images with complex context, couldn´t be detected, an eye-core based face detection model was proposed. In the proposed model, the technique of HSI based skin detection combined with eye-core detection was used to detect eyes, and then image rotation, features extraction from images and neural network based classification were applied to implement face detection and face confirmation, and obtained the goal that Omni-directional faces could be detected. At last, an experiment was conducted to verify the proposed model with MatLab 2011Ra, and the result that detection accuracy was 95% proved that the model was effective.