Finger Vein Recognition in Radon Space Using Local Entropy Thresholding and Common Spatial Pattern

One of the most suitable biometric methods for identifying individuals is finger veins. In this paper we have proposed a new algorithm for finger vein recognition with a high accuracy level. First we extracted veins from finger vein images using entropy based thresholding. The method extracted veins well, but the images were very noisy. It means that the extracted veins may appeare as broken lines. So we applied Radon transformation to segmented images. The Radon transform is not sensitive to noise due to its integral nature. So in comparison with other methods, it is more resistant to noise. This transformation does not require the extraction of vein lines accurately. Also with using this method, the recognition accuracy and speed increased. For extracting dominant features from finger vein images, the common spatial pattern (CSP) algorithm was applied to the blocks of Radon transformation. To improve the classification accuracy, redundant features were eliminated using genetic algorithm. Finally we used 1-NN classifier to identify people using their finger veins. The study was performed on the Peking University finger vein data set. Experimental results showed that the proposed method had good performance in recognizing individuals and obtained the recognition rate of 100%.