Research On Technologies About Fingerprint Image Enhancement And Feature Extraction

Fingerprint has the nature of uniqueness,stableness and hereditary.Biological research shows that many kinds of diseases are related with specific fingerprint features.Therefore,researcher have been interested in extracting features from fingerprint images and trying to find the relations between the features and diseases.Currently,features,potentially disease related,are extracted by doctors manually.Such methods are limited by low reliability,poor accuracy and hard to repeat.Moreover,the analysis results subjectively depends on personnel expertise Obviously,automated features extraction by using computer will definitely provide a objective and accurate platform for the research on the relation between the features and diseases.In order to accurately extract features from fingerprint images,this paper focuses on some key technologies in fingerprint image processing,including the establishment of fingerprint directional field,fingerprint image enhancement and singular point detection,and puts forward some methods trying to solve the shortcomings of traditional methods.The contribution of this paper are as follows:1.The accuracy of the directional field of fingerprint is significant to the fingerprint recognition system.Aiming at the dilemma that the traditional algorithm is difficult to effectively balance the calculation accuracy and noise suppression,this paper presents a directional field establishment method based on multi-scale composite windows and a hierarchical smoothing algorithm for directional field.For the multi-scale based method,a series of directional fields are firstly established under multiple scales of composite windows.Subsequently,the square gradient consistency is calculated to determine the weight of each scale in the series.Finally,the direction fields are integrated according to the weights to get a rough directional field,named as Orient.In order to overcome the drawbacks of traditional smoothing algorithms for directional field,which will cause the singularity to shift,a hierarchical smoothing algorithm for directional field is proposed.The proposed algorithm contains two steps.The first step is to quantize the directional field into two directional modes,and eliminate the isolated directional pattern to obtain the filter mask,Maskl.According to the filter masking,Mask1,different filtering strategies are adopted to Orient to obtain the filtering result Orient 1,The second step is quantized Orient 1 into three different directions,and eliminate the isolated block to get mask,Mask2,and then according to Mask2,Orient 1 is filtered by using different filtering strategies.And finally an accurate and smooth directional field Orient2 is obtained.The experimental results show that the proposed algorithm is effective and feasible.2.Most of the fingerprint images obtained by the sensor show poor quality,and the enhancement of fingerprint image is essential for the reliability of the fine feature extraction.In this paper,a fingerprint image enhancement algorithm based on Short Time Fourier Tranformation(STFT)analysis is studied.Aiming at the problem that the existing enhancement algorithm based on STFT analysis shows low accuracy in calculating the frequency and orientation images and thus incur unsmooth ridge structure,we propose an fingerprint image enhancement algorithm based on curvature and STFT Analysis.Firstly,the fingerprint image is divided into small blocks overlapping with each other and the blocks are transformed by using STFT analysis.Subsequently,the threshold segmentation and morphological opening and closing operations are performed to get the effective block spectrum.Then the effective region of the orientation and frequency of the block is calculated to reduce the influence of noise.Based on the orientation coherence of the orientation and frequency estimation of the spectrum,the ridge curvature is calculated.Finally,different filtering methods are applied based on the ridge curvature,which a Butterworth notch bandpass filter is applied on the low curvature region with high orientation consistency,and a Gaussian bandpass filter is applied on the high curvature region with low consistency.Experimental results show that the proposed algorithm is more simple in the filter design compared with the existing method and can effectively remove noise while not causing false ridge.3.Fingerprint singularity,as an important global feature,is of significance in various applications.In order to overcome the difficulty of threshold selection of singularity detection algorithm based on complex filter,susceptible to noise of singularity detection algorithm based on Poincare Index(PI)and weigh the accuracy and robustness of single-scale singularity detection algorithm,we propose a multi-scale singularity detection algorithm based on information fusion.Firstly,the directional field at multiple scales are calculated and the direction field of each scale is filtered by using the complex filter.Subsequently the singularity whose filtering response is greater than a certain threshold and satisfies the PI judgment condition is identified as the candidate singularity.Then a coarse scale is gradually shifted to the fine scale to correct and filter the candidate singular points.Finally,the accurate singular point detection results are reached.By observing the experimental results,the validity of the multi-scale singularity detection algorithm is verified.All of the experiments in the paper is performed on the platform of Visual Studio 2010 +OpenCv.The experimental samples used includes contact fingerprint images and non-contact fingerprint images.