Method and device for identifying fingerprints

A device for identifying fingerprints, the device comprising:

• sensor means for acquiring a test image comprising a number of test points each having a pixel value related to a three-dimensional structure of fingerprints, and defining a test surface;
• means for determining significant points among said number of test points;
• means, for each of the significant points, for determining respective pixel values of an array of test points surrounding the significant point, the array being a test region surrounding the significant point;
• and
• means for verifying local region similarities between the test regions surrounding said significant points and corresponding reference regions of a reference image comprising a number of reference points with pixel value so as to define a reference surface;
• wherein said means for verifying local region similarities comprise computing means for computing, for each of the significant points, an integral norm of the test region and the corresponding reference region by performing a double summation of magnitudes of differences between the pixel values of the test points of the test region and the pixel values of the corresponding reference region.

A method of identifying a fingerprint having a three-dimensional structure, comprising:

• acquiring a test image representing the fingerprint, the test image including a plurality of test points each characterized by a test point value representing a portion of the three-dimensional structure of the fingerprint;
• determining significant points among the plurality of test points;
• verifying a local region similarity between test regions surrounding the significant points and corresponding reference regions surrounding reference points of a reference image, wherein the verifying step includes, for each test region, computing an integral norm between the test region and a corresponding reference region of the reference image;
• for each test region, normalizing the integral norm for the test region to obtain a normalized integral norm for the test region;
• and
• for each test region, comparing the normalized integral norm to a threshold value.

A method of identifying a test image having test points with test point values, the method comprising:

• determining significant test points among the test points of the test image;
• for each significant test point determined, defining a test region surrounding the significant test point;
• for each test region, comparing the test region with a corresponding reference region surrounding a reference test point of a reference image and determining whether the test region matches the corresponding reference region, and thus, is a matching test region;
• for each matching test region, defining a plurality of intermediate segments connecting the matching test region with other matching test regions, each of the intermediate segments including a plurality of test points with test point values;
• for each of the plurality of intermediate segments defined, comparing the intermediate segment with a corresponding reference segment of the reference image by performing a summation of magnitudes of differences between the test point values of the intermediate segment and corresponding reference values of the corresponding reference segment;
• and
• determining whether the test image matches the reference image based on the comparison of the intermediate segments defined for the test image with the corresponding segments of the reference image.