Real-time face detection using integral histogram of multi-scale local binary patterns

Abstract

This paper introduces a novel fast and efficient method for face detection using integral histogram of multi-scale local binary patterns (HMSLBP). Histogram of LBP and its variants are among the state-of-the-art descriptors in vision systems and is very competitive compared to well known Bag-of-SIFT. HMSLBP has the property to be mainly linearly separable and can be trained efficiently with a fast large-scale support vector machine (SVM) solver. The new proposed detector offers a framerate superior to 20 fps in VGA mode with modern computers. © 2011 Springer-Verlag.