Abstract
Facial image representation plays an important role in computer vision and image processing applications. This paper introduces a novel feature selection method, dominant LBP considering pattern type (DLBP-CPT), capable to capture, effectively, the most reliable and robust dominant patterns in face images. In contrast to the Dominant LBP (DLBP) approach, we take into account the dominant pattern types information. We find that pattern type represents essential information that should be included, especially, in facial image representation across illumination. We apply the proposed method with the conventional LBP and the angular difference LBP (AD-LBP) operators. It is shown in this paper, that the proposed DLBP-CPT and DAD-LBP-CPT descriptors are more reliable to represent the dominant pattern information in the facial images than either the conventional uniform LBP or other dominant LBP approaches.