Finger vein image segmentation based on Hessian matrix
LIN Jian1, ZHONG Shuncong1,2, ZHANG Xiang3
(1.Laboratory of Optics, Terahertz and Nondestructive Testing & Evaluation, School of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, China; 2.Fujian Key Laboratory of Medical Instrument and Pharmaceutical Technology, Fuzhou 350108, China; 3.Zhejiang Academe of Mechanical Information Institute, Hangzhou 310009, China)
 

Abstract: Aiming at extracting the vein patterns in lowquality finger vein images, a finger vein images segmentation algorithm based on Hessian matrix was proposed. Hessian matrix was acquired by the convolution of the second derivative of Gaussian filter and image, and the pixels not belonging to finger vein region were firstly filtered out with the use of the property of matrix trace. Then the Hessian matrix eigenvalues of the rest of the pixels were calculated and the pixels not belonging to finger vein region according to the requirement of the eigenvalues were filtered again. Finally, the maximum of the eigenvalues at each pixel under different scales was chosen as the output, and a finger vein image was segmented effectively after image binarization and morphological filtering. The results indicate that the algorithm performs well in separating the vein region from the nonvein region, and the performing is 0.036 5 s faster than the algorithm not optimized for no need of going through all pixels to calculate the Hessian matrix eigenvalues.
Key words: finger vein recognition; image segmentation; Hessian matrix