Study On Fingerprint Image Quality Evaluation Based On Multi-factors

With the overall level of fraud steadily rising, Automated Fingerprint IdentificationSystem (AFIS) has been widely used in identity authentication. However, the stabilityand reliability of AFIS are seriously affected by fingerprint image quality. As a result,reasonable assessment of fingerprint image quality is very useful and always plays animportant role in enrollment, for example, it can strictly control quantity of low-qualityfingerprint image and effectively improve performances of AFIS. Unfortunately, due tothe special texture and topology of fingerprint, obviously, quality assessment methodsof common digital image can’t adequately work on it. Besides, early quality assessmentmethods based on local information or global information usually have some limitationand can’t assess fingerprint image quality all sidedly. Therefore, a multi-factor basedmethod of fingerprint image quality assessment is proposed in this paper.In this paper, technology of fingerprint image collection and appearance oflow-quality fingerprint image were analyzed at first. Then, five impact factors offingerprint image quality were proposed: existence of Singular Points (SPs), localtexture clarity, global texture clarity, available size and dryness of image. Based onconvergence of the five impact factors, composite result of fingerprint image quality canbe obtained easily. Main contents are separated into several parts as follows:①Based on the analysis of collection technology and classification principle,fingerprint image quality was studied, and found out the causes of low-qualityfingerprint image and its impact factors, which laid the foundation for fingerprint imagequality assessment.②Because of studying on some SPs location method, a novel method of SPslocation was proposed. After preprocess, fingerprint image was easy to estimateorientation field and Homogeneous-Zones-Divide (HZD) in this method. Then, SPs canbe located by twice calculation of Poincare Index (PI). At last, the experiment showedthat this method was effective and robust.③Because of studying on some method of fingerprint image quality assessment,multi-factor based method was put forward as follows: detected SPs and made a wholepre-estimation, then, linked with other impact factors (such as texture clarity, availablesize and dryness of image and so on) to comprehensively assess fingerprint imagequality. The experiment showed that this method was able to classify fingerprint image quality into five grades and suitably apply into enrollment. Besides, impact factorsfusion was studied and a scientific fusion method was put forward.④To test the method of this paper, an integrated system of fingerprint imagequality assessment was designed. According to requires of enrollment, the system notonly showed an excellent performance, but also got perfection.Finally, the summarization and prospect of this study were proposed, and the nextplan was put forward.