A SMART CAMERA FOR FACE RECOGNITION Martijn Reuvers , Richard Kleihorst , Harry Broers , Ben Kr¨ ose University of Amsterdam, NL Philips Research Laboratories, Eindhoven,NL Philips CFT, Eindhoven, NL E-mail: mreuvers@science.uva.nl ABSTRACT There is a rapidly growing demand for using smart cameras for various biometric applications in surveil- lance. Although having a small form-factor, most of these applications demand huge processing per- formance for real-time processing. Face recognition is one of those applications. In this paper we show that we can run face recognition in real-time by im- plementing the algorithm on an architecture which combines a parallel pixel processor with a digital sig- nal processor. Everything fits within a digital camera, the size of a normal surveillance camera. I. INTRODUCTION Recently, face recognition is becoming an impor- tant application for smart cameras. Face detection and recognition requires lots of processing perfor- mance if real-time constraints are taken into account [1]. We want to show in this publication that it is pos- sible using thought-over smart camera architectures to achieve good, real-time face recognition results. A “smart camera” is hereby defined as a stand-alone device which is preferably programmable with a size not bigger than a typical video surveillance camera. The platform we suggest for face recognition is the Intelligent Camera (INCA ) produced by Philips CFT [2] as shown in Figure 1. This camera houses a CMOS sensor, a parallel processor for pixel crunch- ing and a DSP for the high level programs. We will show in this paper that this platform is ideal for face recognition. The contents of the paper is as follows: In Sec- tion II we explain about the architecture of the cam- era, In Sections III and IV respectively, we explain about the algorithms that we used for recognition. The results are given in Section V and conclusions are drawn in Section VI. Fig. 1. INCA camera II. MOTIVATION OF THE ARCHITECTURE “Face recognition” consists of a face detection and a face recognition part. In the detection part faces are detected in the scene and their Region of Interest (ROI) are forwarded to the face recognition process where the found faces are matched to a database in order to recognize and identify them. The detection part is face oriented (high-level im- age processing). It finds face candidates in the scene. In order to reduce the amount of work, the image needs to be preprocessed by a number of low-level operations. These low-level operations are at pixel level, simple and equal for each pixel. This allows massive data-level parallelism. So the detection part involves low-level as well as high-level image pro- cessing. The recognition part uses high-level image pro- cessing and only works on a few faces per second, but it has a high amount of operations in an iterative way while a database is “scanned”. Because of the higher complexity of the instructions and the combi- nation with an operating system, this part of the algo- rithm is best mapped on a task-parallel architecture.