Some of the features
are shown in Fig. 1. They mainly differ in the number of rectangles
and the orientation of the rectangles with respect to the template.
Jones and Viola [10] introduced diagonal features to capture diagonal
structures in object's appearance. Lienhart [13] enriched the features
of Viola and Jones [27] with efficiently computable rotated Haar-like
features. With the enriched set, they achieve a 10% lower false alarm
rate for a given true positive rate. Li et al. [12] proposed Haar-like
features with disjoint rectangles which were meant to characterize
non-symmetrical features of an object's appearance.