Personal Authentication Using Inner Knuckle Print Based on RGB Images

摘要

Biometric verification methods are reliable ways to recognize a person's identity. Over the past few years, it has been proved that inner knuckle print is one of the effective identifiers which has great ability for distinguishing different people. In this paper, a novel approach for inner knuckle print of ring and middle finger is proposed based on Contact-free 2D images. Instead of gray scale images, we proposed to use RGB images. After extracting region of interests, we considered each of them as three images in primary colors including red, green and blue. Then 2D Gabor filters with proper scales and orientations were utilized to extract features for each RGB component. Finally, these features were classified by using support vector machine. Experimental results achieved from our method can reach a considerable rate of accuracy. In this study, we have accomplished the equal error rate of 0.79±0.90 by using fusion of ring and middle fingers.