Research On Image - Based Palmprint Extraction Method

Effective extraction of biological characteristics by using image processingtechnology has particularly important since the biometric technology was used inindustry and living widely. Palmprint who is one of the biological characteristicsattracted much attention according to its rich details, clear texture and specificity.Two branches applications of palmprint processing technology are palmprintrecognition and palmprint diagnosis expert system. The linear feature of palmprint isone of the identify clues for personal recognition, and palmprint extraction also is thenecessary steps for palmprint diagnosis expert system at present, therefore, extractingpalmprint with correct result could provide effective data support for palmprintdiagnosis expert system and identification. But these palmprint processing technologyis still in the stage of development, so the palmprint extraction research has certainscientific value and social significance.Because the factors of palmprint such as palmprint shapes, irregular distribution,low contrast, and the extracting method of palmprint is still lack of targeted design sofar, so it is the key and difficult point in palmprint processing technology. In order toobtain correct and detailed, natural palmprint features for research purposes, researchcontents of this papers are as follows:(1) The study of the extraction of palmprint principal lines. This paper present amethod of principal lines extraction which consists of three steps, the first step calledcoarse extraction, the second step is line positioning, and the third step is accurateextraction. This paper firstly proposed using Niblack’s algorithm for crude extraction,and its threshold parameter is analyzed. Based on its imperfection, the local squaressegmentation has been designed to reduce the produce of the pseudo lines; The sobeloperator with range template is adopted for principal lines positioning, after lots ofexperiments, this paper also put forward a method which used SUSAN operator forwide line positioning after the bottom-hat operator transformation, and design itsadaptive gray level threshold； Accurate extraction depend on the intersection ofoutputs of those two steps, the results obtained good effect and compared with other literature method.(2) The study of wrinkles extraction. To texture details such as wrinkles, thispaper use the LAW’s matching template to detect wrinkles, and the adaptivesegmentation rules are designed. In addition, the result cooperates with the principallines extraction to achieve binary image of palmprint.(3) The study of principal lines thinning problem. Firstly, three kinds ofprocessing results of ASFs in binary image and the influence of the number ofiterations was studied, and the result by second iteration was confirmed to get besteffect. Secondly, the applicability of zhang-suen operator and the mathematicalmorphology thining operator are compared. Finally, the tentative study on theprincipal lines fitting method, this paper put forward a idea that it according to themaximum connected domain and the least squares fitting method to implement thesmooth of principal lines, and analyzed the results.(4) The study of linear feature extraction. Considered of identification, Niblack’salgorithm is proposed with adaptive LAW filter template to detect linear feature ofpalmprint, it obtained stable extraction result, and the superiority of this method isproved by comparing with the methods from other literature,...