FingerCode: A Filterbank for Fingerprint Representation and Matching

With the ident i ty fraud in o u r society reaching u n precedented proportions and wi th a n increasing emphas i s o n the emerging automat ic posit ive personal ident i f ication applications, biometrics-based identif ication, especially fingerprint-based identif ication, i s receiving a lot of a t ten t ion . There are t w o m a j o r shortcomings of t he traditional approaches t o f ingerpr in t representa t ion . For a significant f rac t ion of population, the . representations based o n explicit detection of complete ridge structures in the f ingerpr in t are d i f i cu l t t o extract automatically. T h e widely used minutiae-based representation does n o t uti l ize a significant component of the..riCh discriminatory in format ion available in the f ingerprints. T h e proposed filter-based algorithm uses a bank of Gabor f i l ters t o capture both the local and the global details in a f ingerpr in t as a compact 640byte fixed length FingerCode. T h e f ingerpr in t ma tch ing i s based o n the Eucl idean distance between the t w o corresponding F ingercodes and hence i s extremely fa s t . Our init ial results show identif ication accuracies comparable t o the best results of minutiae-based algorithms published in the open literature [ . I . Finally, w e show that t he matching per formance can be improved by combining the decisions of the ma tchers based o n complementary f ingerprint in format ion .