| 期刊名称： |
Pattern Recognition |
| 全部作者： |
Feng Liu, David Zhang |
| 出版年份： |
2014 |
| 卷 号： |
47 |
| 期 号： |
|
| 页 码： |
|
| 查看全本： |
|
The paperstudiesa3D fingerprint reconstructiontechniquebasedonmulti-viewtouchless fingerprint images. Thistechniqueoffersasolutionfor3D fingerprint imagegenerationandapplicationwhenonly multi-view 2Dimagesareavailable.However,thedifficulties andstressesof3D fingerprintreconstruc- tion aretheestablishmentoffeaturecorrespondencesbasedon2Dtouchless fingerprintimagesandthe estimation ofthe finger shapemodel.Inthispaper,severalpopularusedfeatures,suchasscaleinvariant feature transformation(SIFT)feature,ridgefeatureandminutiae,areemployedforcorrespondences establishment. Toextractthese fingerprintfeaturesaccurately,animproved fingerprint enhancement method hasbeenproposedbypolishingorientationandridgefrequencymapsaccordingtothe characteristicsof2Dtouchless fingerprint images.Therefore,correspondencescanbeestablishedby adopting hierarchical fingerprint matchingapproaches.Throughananalysisof4403Dpointcloud finger data (220 fingers, 2pictureseach)collectedbya3Dscanningtechnique,i.e.,thestructuredlight illumination (SLI)method,the finger shapemodelisestimated.Itisfoundthatthebinaryquadratic function ismoresuitableforthe fingershapemodelthantheothermixedmodeltestedinthispaper.In our experiments,thereconstructionaccuracyisillustratedbyconstructingacylinder.Furthermore, results obtainedfromdifferent fingerprint featurecorrespondencesareanalyzedandcomparedtoshow which featuresaremoresuitablefor3D fingerprint imagesgeneration.