A Robust Face Detection System for 3D Display System
Face detection is a kind of extremely useful technology in many areas, such as security surveillance, electronic commerce and human–computer interaction and so on. Face detection can be viewed as a two-class classification problem in which an image region
- PDF / 1,701,202 Bytes
- 5 Pages / 439.37 x 666.142 pts Page_size
- 29 Downloads / 211 Views
Abstract Face detection is a kind of extremely useful technology in many areas, such as security surveillance, electronic commerce and human–computer interaction and so on. Face detection can be viewed as a two-class classification problem in which an image region can be classified as being a ‘‘face’’ or ‘‘nonface’’. Detection and locating the position of the observers’ face exactly play a critical role in stereoscopic display system. Accuracy, speed and stability are some main standards to evaluate an object-tracking system. The face detection system presented in the paper with classifiers trained by AdaBoost algorithm can meet the specific requirements of stereoscopic display in detecting speed, accuracy and stability. After accurate face detection, we utilize a certain method to detect the pupil in the area of face which is obtained in the above process. At last, the active 3D display equipment will project corresponding images of the same scene to users’ pupil respectively to make sure the viewer can obtain the sense of depth. According to the experimental results, this system is highly accurate, stable and users can get well experience through this 3D display system. Keywords Face detection
3D display AdaBoost
Introduction Face detection is a kind of extremely useful technology in many areas, such as security surveillance, video search, electronic commerce, human–computer interaction and so on [1–3]. Face detection can be viewed as a two-class classification problem in which an image region could be classified as being a ‘‘face’’ or ‘‘non-face.’’ Face detection also provides interesting challenges to the pattern recognition and machine learning area. Y. Zhang (&) Y. Wang Department of Electronic Science and Engineering, NanJing University, NanJing, China e-mail: [email protected]
Y.-M. Huang et al. (eds.), Advanced Technologies, Embedded and Multimedia 1055 for Human-centric Computing, Lecture Notes in Electrical Engineering 260, DOI: 10.1007/978-94-007-7262-5_120, Ó Springer Science+Business Media Dordrecht 2014
1056
Y. Zhang and Y. Wang
After several decades of research, the existing face detection algorithms can be generally divided into four categories: knowledge-based methods [4], feature invariant approaches [5], template matching methods [6] and appearance-based methods [7, 8]. Accuracy, speed and stability are some main standards of evaluating an object-tracking system. The face detection system presented in the paper with classifiers trained by AdaBoost algorithm [7, 8] can meet the specific requirements of stereoscopic display in detecting speed, accuracy and stability. Afterwards, the accurate location of pupil will be detected in the next process with reference to the face area in the image which could reduce the computational complexity effectively and improve the detection accuracy in pupil tracking process. Subsequently, the stereoscopic display system will accurately project the corresponding images to the pupils of observers to achieve good user stereoscopic experiences without any
Data Loading...
