Facial Landmarks Detection Based on Correlation Filters

The problem of locating facial landmarks is important in many applications such as security, 3D modeling and expression recognition. In this paper, we present a new facial landmarks detection system. The core of the proposed system is a cascade of a new detector based on correlation filters. This detector inherits from the correlation filters the tolerance to small variations of the desirable pattern. This detector is refereed to as IPD (Inner Product Detector) and, different from the correlation filters, is suitable for features with a small number of dimensions. In our experiments we use cross-validation of 503 images from BioID database. We verify that the proposed method provides competitive performance when compared to Support Vector Machines.