A Human Face Recognition Approach Based on Spatially Weighted Pseudo-Zernike Moments

摘要

A new modified pseudo-Zernike moments feature, namely, "spatial weighted pseudo-Zernike moments" (SWPZM) is proposed for face recognition in this paper. Since different facial region plays a different important role for face recognition, the new modified pseudo-Zernike feature is weighted with a weight function derived from the spatial information of the human face; hence the most important regions such as the eyes, nose, and mouth regions are intensified for face discrimination. Experimental results based on the AT&T/ORL, Yale, and their combined face database show that SWPZM can obtain 95.7%, 92.3%, and 92.5% recognition rates with the nearest neighbor rule and have better identification power than other methods.