Real-Time Face Detection Using Illumination Invariant Features

Abstract

Abstract.A robust object/face detection technique processing every frame in real-time (video-rate) is presented. A methodological novelty are the suggested quantized angle features ("quangles"), being designed for illumination invariance without the need for pre-processing, e.g. histogram equalization. This is achieved by using both the gradient direction and the double angle direction (the structure tensor angle), and by ignoring the magnitude of the gradient. Boosting techniques are applied in a quantized feature space. Separable filtering and the use of lookup tables favor the detection speed. Furthermore, the gradient may then be reused for other tasks as well. A side effect is that the training of effective cascaded classifiers is feasible in very short time, less than 1 hour for data sets of order 10 4 . We present favorable results on face detection, for several public databases (e.g. 93% Detection Rate at 1 × 10 −6 False Positive Rate on the CMU-MIT frontal face test set).

Verifying liveness by multiple experts in face biometrics

“…For step 1, face detection, we suggest to employ the method introduced in [6]. It is an image-based technique for object detection that utilizes quantized angle features ("quangles").…”

An integrated sign language recognition system

“…There are currently a wide variety of face detection techniques [58,103]. The most popular of these techniques are those that are knowledge-based [5,74], feature-invariantbased [1,19,64], template matching-based [41,78,124], appearance-based [36,50,102] and part-based methods [81,110]. Of all these techniques, the appearance-based ViolaJones face detector is the most popular [32,34].…”

Soft biometric trait classification from real-world face videos conditioned on head pose estimation

“…Face classification from unconstrained environments is not a trivial task considering the challenges presented by real-world environments (Figure 1). Despite the wide literature on soft biometric trait classification [30,22,9,24,13,28,4,16,32,29,6,23,5,10,19,33] and head pose estimation [25,31,1,26,3,8], most of these approaches are not built for unconstrained environments (see Section 2 for details). Humans, on the other hand, are good at such classification/estimation tasks in real-world environments since they take into consideration not only the facial features, but also the conditions under which these features are collected, such as the head pose for the case of biometric trait classification.…”