| Summary: | 碩士 === 國立東華大學 === 電機工程學系 === 91 ===
In recent years, human face detection is becoming more and more popular. Automatically detecting human faces is becoming a very important task in various applications such as video surveillance, human computer interface, face recognition and face image database management. In the face recognition application, the human face location must be known before the processing. The face tracking application also needs a predefined face location at first. In the face image database management, the human faces must be discovered as fast as possible due to the large image database. Although numerous methods are currently used to perform the face detection, there are still many factors that make the face detection more difficult, such as scale, location, orientation, occlusion, expression and wearing glasses. Various approaches of face detection are proposed in recent years, but rare of them take all of the factors above into account. However, a face detection technique that can be used in any real time application needs to satisfy the factors above.
In this thesis, we propose a novel method to deal with the above difficulties. The objective is to detect the face region for video sequences. Therefore, the face pose should not be laminated. We propose a fast algorithm of face detection based on color, motion and facial feature analysis. Firstly, we use a set of chrominance values to obtain the skin color region. Secondly, we propose a novel method for segmenting the motion region by the enhanced frame difference. Then, we combine the skin color region and the motion region to locate the face candidates. We propose a robust eye detection method to detect the eyes in the detected face candidates region. Finally, we verify each eye pair to decide the validity of the face candidate.
According to the experiment results, the user need not be restricted when detecting the face. In general condition, the user could have wide range of face activity such as different position, size, orientation, view and facial expression. Besides, the proposed algorithm also has a satisfied detection rate even if the user is wearing glasses. The detection speed can achieve 30 frames per second for CIF sequence and 120 frames per second for QCIF sequence. Consequently, the proposed method is robust, practical and efficient.
|
|---|
