On using Hough forests for robust face detection

Abstract

Face detection is one of the most important areas of research in computer vision due to its various uses in a wide range of human face-related applications. This paper proposes a method for detecting faces in uncontrolled imaging conditions using a probabilistic framework based on Hough forests. Hough forests can be regarded as task-adapted codebooks of local appearance that allow fast supervised training and fast matching at test time, codebooks are built upon a pool of heterogeneous local appearance features, a codebook is learned for the face appearance features that models the spatial distribution and appearance of facial components. The feasibility of the proposed method has been successfully tested on two challenging and widely used databases (i.e., CMU+MIT and FDDB) and the obtained results are encouraging.