Abstract: Edge response in terms of direction and number will vary from person to person and expression to expression and can be used for face and expression recognition. Encoding these face features in a compact form is a must to obtain the discriminant information. This is a hot research topic and many works have already been done on the same. This paper proposes a novel algorithm based on Local Directional Number Pattern (LDN) for encoding edge responses and directional information of an image in an effective manner. Choice of appropriate edge detection method is very important for finding the absolute gradient magnitude for edges. Kirsch Compass Mask (KCM), a popular edge detection operator which uses the derivative approximation to find edges is used in this work. This paper also discusses two variants of edge detection namely Slow KCM (SKCM) approach which is computationally more complex and Fast KCM (FKCM) approach which is less complex. Analysis of edge responses in different expressions namely normal, smile, surprise, disgust, and sad, using the proposed algorithm is presented in this paper. The results show that LDN coded image can be used as unique feature descriptor in face and expression recognition.

Keywords: Local Directional Number Pattern, gradient magnitude, Kirsch Compass Mask, derivative approximation, Face and expression recognition.