Partial fingerprint recognition

1. A method for partial fingerprint recognition performed using a computer or a processor, the method comprising the steps of:

extracting features including ridge orientations, valley images, minutiae, and pores from at least two fingerprint fragments;

aligning the fingerprint fragments;

matching the pores and minutiae on the fingerprint fragments after applying estimated alignment transformation;

calculating a final matching score based on a pore matching score and a minutiae matching score; and

identifying a person based on a result of the final matching score,

wherein the extraction of the pores further comprising:

dividing the fingerprint fragments into blocks;

identifying two spatial filters for each of the blocks based on ridge frequencies in the block;

using the spatial filters to generate block-wise image response;

thresholding the block-wise image response and estimate a difference, which generates initial pore extraction results; and

post-processing the initial pore extraction results to eliminate spurious pores.

2. The method according to claim 1 , wherein the minutiae matching score is a ratio between number of matched pore to total number of pore, and the pore matching score is a ratio between number of matched minutiae to total number of minutiae.

3. The method according to claim 1 , wherein the spatial filters are Gaussian filters.

4. The method according to claim 1 , wherein the aligning of fingerprint fragments further comprising:

detecting pore correspondences based on the ridge orientations and valley images of neighboring pores;

checking the pore correspondences based on global geometrical relationship between the pores;

estimating the alignment transformation between fingerprint fragments based on the detected pore correspondences.

5. The method according to claim 4 , wherein if the fingerprint fragments cannot be aligned, it is determined that the fingerprint fragments are from different fingers.

6. A method for partial fingerprint recognition performed using a computer or a processor, the method comprising the steps of:

extracting features including ridge orientations, valley images, minutiae, and pores from at least two fingerprint fragments;

aligning the fingerprint fragments;

matching the pores and minutiae on the fingerprint fragments after applying estimated alignment transformation;

calculating a final matching score based on a pore matching score and a minutiae matching score; and

identifying a person based on a result of the final matching, wherein the extraction of the pores further comprising:

dividing fingerprint fragments into blocks and terminating partition from localized ridge-orientation inconsistency information in the block;

defining a local model for each of the blocks which is used as a matched filter to filter the block;

thresholding filtered outputs to obtain initial pore extraction results; and

post-processing the initial pore extraction results to remove spurious pores.

7. The method according to claim 6 , wherein the minutiae matching score is a ratio between number of matched pore to total number of pore, and the pore matching score is a ratio between number of matched minutiae to total number of minutiae.

8. The method according to claim 6 , wherein the spatial filters are Gaussian filters.

9. The method according to claim 6 , wherein the aligning of fingerprint fragments further comprising:

detecting pore correspondences based on the ridge orientations and valley images of neighboring pores;

checking the pore correspondences based on global geometrical relationship between the pores;

estimating the alignment transformation between fingerprint fragments based on the detected pore correspondences.

10. The method according to claim 9 , wherein if the fingerprint fragments cannot be aligned, it is determined that the fingerprint fragments are from different fingers.