Image Processing and Communications Challenges 8

Abstract

This paper uses different algorithms to build a hybrid system for detecting human-face and tracking unrestricted. The system uses the face detection algorithm, Kalman filter [1]. The architecture is as follows: it locates the face in the image and get a sub image from the region of the head, face patterns are determined as the eyes, the center of the face, the border of the head, these parameters are used in the Kalman filter to takes the final decision on the direction in which the face in the image and reduce the error when more than one person in the picture, especially when there is no face but we know that still another position. In face recognition [2], the algorithm takes the detection phase, cutting image of detected face, which is divided in 9 subsections [6], where histogram comparison process [8] and phase correlation are made [7], where given results are processed by a decision tree which makes the decision if face is known or not. The experimental results show that the system is stable when it is saturated field of view with many faces or people. © 2013 Springer-Verlag.