Scale and translation invariant face detection and efficient facial feature extraction

Abstract

In this paper, improved and fully automatic face detection and facial feature extraction algorithms are introduced. The face detection algorithm is based on skin color, head shape and additional information such as orientation and characteristics of the segmented face region. The proposed face detection technique is position, scale, shape and skin color invariant. The more challenging task of facial feature extraction involves the identification of the sub-images containing the facial features. Each sub-image is analyzed by an efficient combination of a number of techniques such as: circle fitting, edge detection, and intensity based adaptive clustering. Our simulations indicate that the performance especially for the detection of facial features is significantly improved.