Inner-Knuckle-Print Acquisition And Preprocessing

As a new biological characteristic, Inner-Knuckle-Print (IKP) has received more andmore attention in recent years for some specific characters such as prominent line feature andrich information content. However, the existing IKP image acquisition and pre-processingmethods have some disadvantages. First, the direction of the fingers are not fixed, hence theremay exist severe rotation transformation between the regions of interest (ROI) acquired fromthe same finger at different times. Severe rotation transformation can reduce the performanceof the IKP identification systems greatly. Second, the existing methods extract the ROI fromthe whole hand image. The ROI, which is much smaller than the hand image, is alow-resolution image. To resolve these problems, this dissertation presents a new imageacquisition and preprocessing method in which a small region around two knuckles is imagedand two ROI sub-images with higher resolution are extracted from it. Moreover, the rotationtransformation between the ROIs is reduced to the minimal by fixing the direction of thefingers being imaged.The main contents of the paper are as follows:1. The designing of image acquisition equipment. This equipment can capture a smallregion which contains two knuckles. The resolution of the image obtained is higherthan that obtained with traditional methods with the same kind of camera. Moreover,we use two pegs to control the position and the direction of the fingers being imaged,such that the rotation transformation between the ROIs is reduced to the minimal.2. An efficient ROI extraction algorithm is designed. In this algorithm, the fingers in theimage are segmented based on lines between the fingers, and the ROI sub-images areextracted from the finger regions based on the line features of the IKP.3. The establishment of an IKP database. We collected2000images from100volunteers.The images were collected on two separate occasions, at an interval of around twomonths. On each occasion, the subject was asked to provide20images of the right hand. The ROI sub-images of the middle and ring fingers are extracted from theseimages to establish the IKP database.Experimental results show that our work can improve the performance of IKPrecognition system.