Title:A single scale retinex based method for palm vein extraction

Abstract:Palm vein recognition is a novel biometric identification technology. But how to gain a better vein extraction result from the raw palm image is still a challenging problem, especially when the raw data collection has the problem of asymmetric illumination. This paper proposes a method based on single scale Retinex algorithm to extract palm vein image when strong shadow presents due to asymmetric illumination and uneven geometry of the palm. We test our method on a multispectral palm image. The experimental result shows that the proposed method is robust to the influence of illumination angle and shadow. Compared to the traditional extraction methods, the proposed method can obtain palm vein lines with better visualization performance (the contrast ratio increases by 18.4%, entropy increases by 1.07%, and definition increases by 18.8%).