Effect of hough forests parameters on face detection performance: An empirical analysis

Face detection as one of the most challenging tasks in computer vision has received a lot of attention in recent decades due to its wide range of use in face based image analysis. In this paper, we propose an efficient approach for face detection that efficiently combines generalized Hough transform within random decision forests framework. In this approach, we train random decision forests that directly maps the image patch appearance to the probabilistic vote about the possible location of the face centroid; the detection hypotheses then correspond to the maxima of the Hough image. The random decision forests construction and prediction abilities depend on setting some parameters, which in turns affects the performance of the method. Therefore, the impact of these parameters that most influence the behavior of the forest for detecting faces is studied through experiments on the widely used CMU+MIT database. Moreover, a comparison with some published methods is presented.