Multispectral fusion for palmprint recognition

Abstract

In order to improve the accuracy of palmprint recognition, multispectral imaging has been employed with a fusion strategy to acquire more discriminative information. The PolyU multispectral palmprint images contain four kinds of palmprint images which are respectively captured under red, green, blue and near infrared illuminations and contain much information than the single band. However, there is also redundant information. How to extract the critical information among the multispectral palmprint images and to fuse them is an emerging problem. In this paper we address the problem of selecting bands from the original four bands, and use the extended general color image discriminant (GCID) model to generate three new color components for further improvement of the recognition performance. Experiments using the PolyU multispectral palmprints database show the effectiveness of the proposed models and fusion strategy.