A hardware acceleration unit for face detection

In recent years the use of real-time face detection and face recognition for surveillance, human-machine interfaces and other applications has increased and thus the need for high power, low cost implementations has been posed. In embedded implementations, the computing power needed for face detection system calls for a custom designed processor. In this paper we will discuss implementations of alternatives for accelerating functions used in face-detection algorithms of the OpenCV library. A comparison between computer-based Haar face detection algorithms of OpenCV and custom computer vision hardware will be presented.