Kernel-Coefficient-Based Feature Method for Face Detection

A kernel-coefficient-based feature method is proposed to detect faces. The proposed method uses a mathematical expression and 26 different arrangements of kernel-coefficients of a kernel (testing region). The method manipulates the symmetric appearance of a face with respect to a rigid-kernel (fixed region). The expression, which is used to generate feature values, responds to pixels on edges of the image-objects only. For each distinct arrangement of kernel-coefficients, a feature-value is generated. The objective of the proposed kernel-coefficient-based feature method is to reduce the number of feature values required for face detection.