Improved Gradient-based Methods For Fingerprint Orientation Field Extraction

With the development of science and the progress of the society, the fingerprint identification technology has been very mature. But as the improvement of the living standard, the requirement of people for fingerprint identification technology is more and more strict. As so far, there are still many shortcomings and loopholes in the fingerprint identification technology,fingerprint identification technology remains to be improved. The extraction of orientation field is one of key step in the fingerprint identification technology,the accurate extraction of orientation field is very important to fingerprint enhancement, feature extraction and fingerprint match. Although there are many methods for the extraction of orientation field currently, there are still a lot of problems to be solved. Two kinds of methods based on gradient were considered to extract the orientation field of the fingerprint image. The mainly innovation in the extraction of orientation field:(1) A neighborhood operator gradient-based method for the fingerprint orientation field extraction is considered. As the orientation field tend to be continuous in the neighboring regions, and the orientation field is closely related with neighborhood information, the consistency can reflect the reliability of the orientation field, combined the local fingerprint image with itsneighboring image, selected the most suitable neighboring image structure by comparing the consistency of the local image to calculate the orientation.Experiments were made to compare and analysis in three ways :the orientation field histogram, the mean absolute difference and the computation time of the orientation field.The result showed that the method is able to better estimate fingerprint orientation field.(2) An adaptive gradient-based method for the fingerprint orientation field extraction is considered. As the orientation field tend to be continuous in the neighboring regions, and the orientation field is closely related with neighborhood information. Calculate the fingerprint image based on gradient,and then divide the image into a series of non-overlapping inner blocks.Distributing a common outer block for the inner block with different size, then use the outer block to calculate the gradient vectors and the coherence of the inner blocks. Selecting the most suitable outer block by comparing the coherence of the inner blocks. Confirming the threshold of the coherence and the number of effective points. For every inner block, if its coherence or number of effective points is smaller than the threshold, then set its gradient vectors as zero, and calculate the blocks with their neighboring blocks.Experiments were made to compare and analysis in three ways :the orientation field histogram, the edge detection and the computation time of the orientation field. The result showed that the method is able to better estimate fingerprint orientation field.