Enhancement And Segmentation Of Low Quality Hand Dorsal Vein Images

Identification is essential to everyone since ancient times, It's more crucial to the requirements of identification in modern society. Due to the development of social science and technology, the traditional identification technology has not been able to fit the requirements of increasing security. Biometrics recognition is a new identification technology. Because biometrics recognition as compared with the traditional identification has more security, more privacy and greater convenience, the biometric technology has been studied more deeply and been widely applied to various fields of social life. Hand dorsal vein recognition technology is a biometric technology that has the advantages of unique, non-contact, hardly forged properties and high levels of security, etc., so it has being a hot research topic in recent years. Hand dorsal vein recognition technology that is similar to the some other biometrics recognitions such as face recognition, fingerprint recognition, etc. faces many problems to be solved in the realization of the commercial systems.Generally, the contrast of the collected hand dorsal vein images is very low, and those images are not fit for the segmentation and feature extraction of hand dorsal veins. Thus, it is necessary to enhance the low-contrast hand dorsal vein images before recognition. Although many researches on the hand dorsal vein recognition technology have been made and the vein images were enhanced before recognition, the existing enhancement methods are too simple and the enhancement effect is not so obvious.In addition, the quality of the segmentation of hand dorsal vein images directly affects the feature extraction of the hand dorsal veins. Many researchers mainly use the method of threshold segmentation to segment the hand dorsal vein images, but it is not so good because using this method can lead to losing some characteristics of the hand dorsal veins.In this dissertation, firstly, the collected low-contrast images of hand dorsal veins are normalized in size and intensity and filtered, and then, the hand dorsal vein images are enhanced by using the contrast limited adaptive histogram equalization (CLAHE) after the study of the histogram equalization processing method and the local adaptive histogram equalization. In addition, this dissertation presents an improved method of hand dorsal vein image segmentation. First of all, the enhanced vein images are processed by using the method of mathematical morphology; and then, the enhanced vein image and the vein image processed by mathematical morphology operations are combined by weighting operation; finally, the images that operated by the weighting operation are segmented by using the improved threshold segmentation method, and we obtain the final segmented images. The improved image segmentation method is compared with some classical threshold segmentation, and the experimental results show that the improved vein segmentation method in this dissertation has better segmentation effect.