B. Fröba, A. Ernst, and C. Küblbeck (Germany)
Object Detection, Real Time Detection, Edge Orientation Matching
In this paper we present a real-time algorithm for detection of frontal faces in grey images. It extends our former work on edge orientation matching (EOM) for upright face de tection. Here we introduce an additional verification stage to the EOM which leads to a more robust detection sys tem compared with our former detector [6]. Further on we present current developments like the detection of in-plane rotated faces and a new training procedure for the EOM using AdaBoost [5]. Unlike many approaches that model the grey level appearance of the face our approach is com putationally very fast. It takes about 0.05 seconds on an Athlon 1000MHz for a ¿ ¢¾ imageto be processed us ing a multi-resolution search with 10 resolution levels. We demonstrate the capability of our detection method on an image database of 18525 images which show frontal faces. The variations in head size, lighting and background are considerable. The obtained detection rate is more than 96% on that database.
Important Links:
Go Back
