Abstract
We proposed a new normalization method for iris recognition, which is different from the conventional one in which the annular iris region is unwrapped to a rectangular block under polar coordinate. In this method, we investigate the effect of interpolation and decimation in conventional normalization method to recognition rate for the first time. We used the original texture to fill the pupil area, then a novel normalized image can be obtained with the geometric structure and directional information of original iris image well preserved, which enables us to choose simpler features than before. Subsequently, we extracted the multi-direction and multi-scale information feature of normalized iris image by contourlet transform, and adopt SVM to classify the features. Experimental results validate the improvement of recognition rate. required. Thus the geometric structure and the directional information of iris image are altered, and the recognition performance may be affected if texture and directional information are adopted as features. However, the effect has not been studied previously. The purpose of the letter is to study the distortion of the polar coordinate transform to the recognition rate. In order to validate and attempt to alleviate this effect, we proposed another different normalization method without adopting the polar coordinate transformation, thus the original geometric structure and directional information can be preserved. We also propose to use a simple feature that is derived from multi-scale and multi-direction local information so that we can highlight the effect of the normalization step to the recognition rate. After this preprocessing procedure, we use contourlet transform to extract the multi-direction and multi-scale information feature of normalized iris image, and then use support-vector-machine (SVM) to classify the features and give the classification result (8). Contourlet transform is a typical multi-scale and multi-direction system (7). It contains Laplacian Pyramid to achieve multi-scale property and Direction Filter Bank to achieve multi-direction characteristic. SVMs belong to a family of generalized linear classifiers. A special property is that they simultaneously minimize the empirical classification error and maximize the geometric margin. A comparison of our proposed method with existing traditional method is presented, which confirms that interpolation and decimation operations will indeed affect recognition accuracy when texture and direction information are used as features. Experimental results show that the recognition rate can be improved by approximately 30% compared with conventional methods with simple feature being used.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.