Feature-Based Matcher for Distorted Fingerprint Matching

Abstract

In one aspect, methods that use a novel representation, referred to as an octant feature vector (OFV), are used for matching distorted fingerprints. For instance, a feature-based matcher for distorted fingerprint matching may use a two-step local and global matching scheme to compare a set of feature vectors that are derived from minutiae of a reference fingerprint and a search fingerprint. The relative geometric relationships between the reference minutia and nearest minutiae may be derived and encoded into a feature vector based on orientation difference. The OFV is invariant to the rigid transformations and is insensitive to nonlinear distortions since the relative geometric relationships are independent from the rigid transformation.

What is claimed is: 
     
         1 . A method performed by one or more computers, the method comprising:
 identifying a geometrically aligned region between a search fingerprint image and a reference fingerprint image;   determining a set of globally aligned mated minutia pairs based on the geometrically aligned region between the search fingerprint image and the reference fingerprint image;   for each mated minutia pair included in the set of globally aligned mated minutia pairs:
 computing a first score representing a feature difference between (i) a search minutia and (ii) reference minutiae included in a set of reference minutiae that are identified as being closest neighboring minutiae to the search minutia in the geometrically aligned region, and 
 computing a second score representing a ratio of a number of paired mated minutiae in the geometrically aligned region included in the set of reference minutiae relative to a number of paired unmated minutiae in the geometrically aligned region included in the set of reference minutiae; 
   computing a match similarity score between the search fingerprint image and the reference fingerprint image based at least on the first score and the second score; and   providing the match similarity score for output.   
     
     
         2 . The method of  claim 1 , wherein identifying the geometrically aligned region between the search fingerprint image and the reference fingerprint image comprises:
 computing, for each search minutia included in the set of globally aligned mated minutia pairs, a local similarity score based at least on search octant feature vectors of the search minutia and reference octant feature vectors of the set of reference minutiae; and   determining a rotation angle between the search fingerprint image and the reference fingerprint image based on the local similarity scores.   
     
     
         3 . The method of  claim 2 , further comprising:
 determining, based on the local similarity scores, a set of local closest matched minutia pairs;   computing sets of rotation parameters based on the set of local closest matched minutia pairs.   
     
     
         4 . The method of  claim 3 , wherein computing the sets of rotation parameters comprises identifying angles corresponding to a set of top bins of an angle offset histogram of the set of local closest matched minutiae pairs. 
     
     
         5 . The method of  claim 1 , wherein each mated minutia pair included in the set of globally aligned mated minutia pairs comprises a search minutia and a corresponding reference minutiae included in the geometrically aligned region that is determined to have geometrically consistent features to the search minutia. 
     
     
         6 . The method of  claim 1 , wherein:
 the search fingerprint image comprises a distorted fingerprint; and   the match similarity score is used to determine a fingerprint match between the search fingerprint image and the reference fingerprint image.   
     
     
         7 . The method of  claim 1 , further comprising:
 comparing, for each mated minutia pair included in the set of globally aligned mated minutia pairs, features of a search minutia of a mated minutia pair and features of a corresponding reference minutia of the mated minutia pair;   based on comparing the features of the search minutia of the mated minutia pair and the features of the corresponding reference minutia of the mated minutia pair:
 identifying a set of rotation parameters for the geometrically aligned region, and 
 identifying a set of translation parameters for the geometrically aligned region; and 
 aligning the search minutiae to the reference minutiae included in the set of globally aligned mated minutia pairs based at least on the set of rotation parameters and the set of translation parameters. 
   
     
     
         8 . A system comprising:
 one or more computers; and   one or more storage devices storing instructions that are operable, when executed by one or more computers, to cause the one or more computers to perform operations comprising:
 identifying a geometrically aligned region between a search fingerprint image and a reference fingerprint image; 
 determining a set of globally aligned mated minutia pairs based on the geometrically aligned region between the search fingerprint image and the reference fingerprint image; 
 for each mated minutia pair included in the set of globally aligned mated minutia pairs:
 computing a first score representing a feature difference between (i) a search minutia and (ii) reference minutiae included in a set of reference minutiae that are identified as being closest neighboring minutiae to the search minutia in the geometrically aligned region, and 
 computing a second score representing a ratio of a number of paired mated minutiae in the geometrically aligned region included in the set of reference minutiae relative to a number of paired unmated minutiae in the geometrically aligned region included in the set of reference minutiae; 
 
 computing a match similarity score between the search fingerprint image and the reference fingerprint image based at least on the first score and the second score; and 
 providing the match similarity score for output. 
   
     
     
         9 . The system of  claim 8 , wherein identifying the geometrically aligned region between the search fingerprint image and the reference fingerprint image comprises:
 computing, for each search minutia included in the set of globally aligned mated minutia pairs, a local similarity score based at least on search octant feature vectors of the search minutia and reference octant feature vectors of the set of reference minutiae; and   determining a rotation angle between the search fingerprint image and the reference fingerprint image based on the local similarity scores.   
     
     
         10 . The system of  claim 9 , wherein the operations further comprise:
 determining, based on the local similarity scores, a set of local closest matched minutia pairs; and   computing sets of rotation parameters based on the set of local closest matched minutia pairs.   
     
     
         11 . The system of  claim 10 , wherein computing the sets of rotation parameters comprises identifying angles corresponding to a set of top bins of an angle offset histogram of the set of local closest matched minutiae pairs. 
     
     
         12 . The system of  claim 8 , wherein each mated minutia pair included in the set of globally aligned mated minutia pairs comprises a search minutia and a corresponding reference minutiae included in the geometrically aligned region that is determined to have geometrically consistent features to the search minutia. 
     
     
         13 . The system of  claim 8 , wherein:
 the search fingerprint image comprises a distorted fingerprint; and   the match similarity score is used to determine a fingerprint match between the search fingerprint image and the reference fingerprint image.   
     
     
         14 . The system of  claim 8 , wherein the operations further comprise:
 comparing, for each mated minutia pair included in the set of globally aligned mated minutia pairs, features of a search minutia of a mated minutia pair and features of a corresponding reference minutia of the mated minutia pair;   based on comparing the features of the search minutia of the mated minutia pair and the features of the corresponding reference minutia of the mated minutia pair:
 identifying a set of rotation parameters for the geometrically aligned region, and 
 identifying a set of translation parameters for the geometrically aligned region; and 
   aligning the search minutiae to the reference minutiae included in the set of globally aligned mated minutia pairs based at least on the set of rotation parameters and the set of translation parameters.   
     
     
         15 . A non-transitory machine-readable data storage device that stores instructions that, when executed by a data processing apparatus, cause the data processing apparatus to perform operations comprising:
 identifying a geometrically aligned region between a search fingerprint image and a reference fingerprint image;   determining a set of globally aligned mated minutia pairs based on the geometrically aligned region between the search fingerprint image and the reference fingerprint image;   for each mated minutia pair included in the set of globally aligned mated minutia pairs:
 computing a first score representing a feature difference between (i) a search minutia and (ii) reference minutiae included in a set of reference minutiae that are identified as being closest neighboring minutiae to the search minutia in the geometrically aligned region, and 
 computing a second score representing a ratio of a number of paired mated minutiae in the geometrically aligned region included in the set of reference minutiae relative to a number of paired unmated minutiae in the geometrically aligned region included in the set of reference minutiae; 
   computing a match similarity score between the search fingerprint image and the reference fingerprint image based at least on the first score and the second score; and   providing the match similarity score for output.   
     
     
         16 . The device of  claim 15 , wherein identifying the geometrically aligned region between the search fingerprint image and the reference fingerprint image comprises:
 computing, for each search minutia included in the set of globally aligned mated minutia pairs, a local similarity score based at least on search octant feature vectors of the search minutia and reference octant feature vectors of the set of reference minutiae; and   determining a rotation angle between the search fingerprint image and the reference fingerprint image based on the local similarity scores.   
     
     
         17 . The device of  claim 16 , wherein the operations further comprise:
 determining, based on the local similarity scores, a set of local closest matched minutia pairs;   computing sets of rotation parameters based on the set of local closest matched minutia pairs.   
     
     
         18 . The device of  claim 17 , wherein computing the sets of rotation parameters comprises identifying angles corresponding to a set of top bins of an angle offset histogram of the set of local closest matched minutiae pairs. 
     
     
         19 . The device of  claim 15 , wherein each mated minutia pair included in the set of globally aligned mated minutia pairs comprises a search minutia and a corresponding reference minutiae included in the geometrically aligned region that is determined to have geometrically consistent features to the search minutia. 
     
     
         20 . The device of  claim 15 , wherein:
 the search fingerprint image comprises a distorted fingerprint; and   the match similarity score is used to determine a fingerprint match between the search fingerprint image and the reference fingerprint image.