| 摘要 |
A sensor receives a print image from an authorized person (21) to form a template, and from a candidate (11) to form test data. Noise variance (12) is estimated from the test data as a function of position in the image, and used to weight the importance of comparison with the template at each position. Test data are multilevel, and are bandpassed and normalized (13) and expressed as local sinusoids for comparison. A ridge spacing and direction map (28) of the template is stored as vector wavenumber fields, which are later used to refine comparison. Global dilation (34) and also differential distortions (45) of the test image are estimated, and taken into account in the comparison. Comparison yields a test statistic (52) that is the ratio, or log of the ratio, of the likelihoods of obtaining the test image assuming that it respectively was, and was not, formed by an authorized user. The test statistic is compared with a threshold value, preselected for a desired level of certainty, to make the verification decision.
|
