Innovative Fingerprint Enhancement: Exploring Finger GAN’s Constrained Generation Scheme

Innovative Fingerprint Enhancement: Exploring Finger GAN's Constrained Generation Scheme

Sahana D Gowda, Professor, Department of Computer Science and Engineering, BGSIT, ACU, Mandya
Manoj V, Department of Computer Science and Engineering, BGSIT, 20CSE044,ACU,Mandya

Abstract— Inert unique finger impression improvement is a fundamental pre-handling step for idle finger impression distinguishing proof. Most inactive finger impression improvement strategies attempt to reestablish ruined dim edges/valleys. In this paper, we propose another strategy that for-mulates dormant unique mark upgrade as an obliged finger impression age issue inside a generative ill-disposed network (GAN) structure. We name the proposed network FingerGAN. It can implement its created unique finger impression (i.e, improved idle balance gerprint) unclear from the relating ground truth case as far as the finger impression skeleton map weighted by particulars areas and the direction field regularized by the FOMFE model. Since details is the essential component for unique mark acknowledgment and particulars can be recovered straightforwardly from the finger impression skeleton map, we offer an all encompassing structure that can perform dormant finger impression upgrade with regards to straightforwardly streamlining particulars data. This will assist with further developing inactive finger impression ID execution essentially. Trial results on two public inactive finger impression data sets show that our technique beats the condition of artistic expressions altogether.

Index Terms—Constrained fingerprint generation, deep convolutional generative adversarial network, latent fingerprint enhancement.