Evaluation Method Of Finger Vein Image Quality

Biometric technology is a way to a personal identity authentication by using human physiological and behavioral characteristics.Confirmation of the identity of criminals is just one application of its identity identification function.Several facts,e.g.,continuously innovative biometric authentication technology,the technology matures and the continuous expansion of its application fields,as well as its market share continues to grow,are all the characterization of its broad development prospects.Among various technologies,the finger vein recognition technology is also constantly showing its own advantages in term of its identification ability and thus getting more and more recognition.However,in actual applications,the finger vein image in its acquisition process will be subject to a variety of different factors,such as unstable environmental lighting,different thickness of fingers of users,some irregular behaviors of users,etc.These factors will indirectly affect the recognition rate of finger vein recognition,also may affect the decision making level fusion results of finger vein and fingerprint.Therefore,the quality assessment of collected finger vein images is particularly important.Given the significant impact of the finger vein image quality,we compare the subjective and objective evaluation methods for finger vein images.We find that objective assessment methods are superior because they do not change the results of assessment according to human's subjective will.So we choose the methods of objective assessment.In the finger vein recognition process,in order to avoid poor quality pictures which may lower the recognition rate,firstly we select three aspects,namely the image defocus blur,the position deviation,effective area,to roughly assess the quality of the collected finger vein images.Then,based on finger vein characteristics and some common problems during collection process,we propose five single measure for evaluation indexes for finger vein characteristics: contrast,entropy,average gradient,clarity,and equivalent number.Finally,by evaluating each single measure of the quality of different finger vein images,it is confirmed that each single evaluation index is irreplaceable as it has a unique degree of response to the image.During the finger vein image acquisition process,not all of the images can achieve the desired effect.Therefore,this article uses those single measures one by one to assess the quality of acquired images,in order to select good quality images for subsequent the finger vein recognition and the decision fusion process of fingerprint and finger vein prints.By using non-contact acquisition mode,due to rotation and offset,the difference will be,so we do not always capture good quality images,but we can still get the complete finger vein information.In the template registration process,we give up non-compliant vein image.During the identification process,we discard those poor quality images,and for recognition,we process images with complete information and extract the features.We enhance the quality of finger vein images by combining contrast adaptive Histogram Equalization and the Median Filter,then correct the rotation and translation,and finally get the ROI regions.Through the comparison between the single measure assessment results of the raw finger vein images and the enhanced images,we determine the finger vein image processing method for each specific indicator,and prove the validity of every single measure indicator.For the finger vein image quality assessment,individual indicator does not fully reflect the overall quality of the finger vein image,but not all the indicators in the evaluation process are at the level of equal importance.Each evaluation shall be conducted for each indicator,which increases the complexity and instability of the evaluation system.Therefore,this article presents a weighting factor based on the determination of fault tree modeling method.The importance of each index is determined by using expert scoring.Fault Tree modeling approach for the first time is used to evaluate the image quality of the finger vein image,and to determine the weighting coefficient of each indicator.Finally,the performance evaluation system is assessed based on the accuracy and predictability of finger vein image quality evaluation.Finally,we design and implement the finger vein image quality evaluation system.The system comprises two modes: quality assessment single measure and quality assessment with two measures.Through the finger vein single and multiple measures of assessment measure,we verify the accuracy of the quality assessment algorithm based on the weighting coefficients.