Local directional number (LDN) pattern operator adopts Kirsch edge mask s to calculate eight directional edge response values of each pixel in an image.In order to improve the feature extraction capacity of LDN operator,16-neighbor and 24-ne ighbor edge masks are proposed in this paper.These two kinds of masks also use more pixels′ gray information to calculate edge respons es of each pixel in an image and can achieve more edge response values.Therefore,more than 64feature values can be extracted from an i mage to express the image.Experimental results on the ORL face database,the Yale face database B and Cohn-Kanade face database s how that when 16-neighbor and 24-neighbor edge masks are used in the LDN operator,these two operator′ average recognition rate s are separately 11.15% and 13.04%.higher than those of LDN operators with Kirsch edge masks.This demonstrates that the improv ement of edge mask can greatly promote the feature extraction capacity of LDN operator.In addition,the experimental results also sh ow that the anti-noise capability and robustness to illumination variation of LDN become stronger by changing the edge masks.