W. Yang and J. Gilbert. A real-time face recognition system using custom VLSI hardware. In , December 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....runs at 100 Hz (four times frame rate) on a Sun SparcStation 10. Several frames taken from a typical image sequence, along with a drawing pin representation of the estimated facial orientation, are shown in Figure 11. Correlation based algorithms (which can run in real time on dedicated hardware [16]) could track the nose tip, allowing the 3 D method to be used for near frontal views of the face. The facial normal, and an estimate of the gaze direction, can be extracted from a single, monocular view of a face, making minimal assumptions about the underlying facial structure. The method for ....

W. Yang and J. Gilbert. A real-time face recognition system using custom VLSI hardware. In , December 1993.

....need a learning set which densely sampled the continuum of possible lighting conditions. Second, correlation is computationally expensive. For recognition, we must correlate the image of the test face with each image in the learning set; in an effort to reduce the computation time, implementors [12] of the algorithm described in [3] developed special purpose VLSI hardware. Third, it requires large amounts of storage the learning set must contain numerous images of each person. 2.2 Eigenfaces As correlation methods are computationally expensive and require great amounts of storage, it is ....

J. Gilbert and W. Yang. A Real--Time Face Recognition System Using Custom VLSI Hardware. In Proceedings of IEEE Workshop on Computer Architectures for Machine Perception, pages 58--66, 1993.

....by using filtered images of model faces. In template based systems, the simplest pictorial representation, faces are represented either by images of the whole face or by subimages of the major facial features such as the eyes, nose, and mouth (Baron[3] Brunelli and Poggio[7] Yang and Gilbert[42], Burt[8] Bichsel[6] Template images need not be taken from the original grey levels; some systems use the gradient magnitude or gradient vector field in order to get invariance to lighting. An input face is then recognized by comparing it to all of the model templates, typically using ....

....transform, optical flow, and correlation. On our unoptimized CM 5 implementation, it takes about 10 seconds for the templatebased recognizer to run since we can distribute the data base so that each processor compares the input against one person. Specialized hardware, for example correlation chips[42], can be used to further speed up the computation. 5 Conclusion In this paper we presented a view based approach for recognizing faces under varying pose. Motivated by the success of recent template based approaches for frontal views, our approach models faces with templates from 15 views that ....

Woody Yang and Jeff Gilbert. A real-time face recognition system using custom VLSI hardware. In IEEE Computer Architectures for Machine Vision Workshop, December 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC