U-Finger: Multi-Scale Dilated Convolutional Network for Fingerprint Image Denoising and Inpainting

TL;DR

This paper introduces U-Finger, a multi-scale dilated convolutional network designed for fingerprint image denoising and inpainting, effectively handling various artifacts while preserving fine details.

Contribution

The paper proposes a novel multi-scale convolutional network with dilated convolutions and no padding, improving fingerprint image restoration performance.

Findings

Achieved second place in ECCV 2018 Chalearn LAP Inpainting Competition

MSE of 0.0231, PSNR of 16.9688 dB, SSIM of 0.8093 on test set

Demonstrated the effectiveness of dilated convolutions and multi-scale modules

Abstract

This paper studies the challenging problem of fingerprint image denoising and inpainting. To tackle the challenge of suppressing complicated artifacts (blur, brightness, contrast, elastic transformation, occlusion, scratch, resolution, rotation, and so on) while preserving fine textures, we develop a multi-scale convolutional network, termed U- Finger. Based on the domain expertise, we show that the usage of dilated convolutions as well as the removal of padding have important positive impacts on the final restoration performance, in addition to multi-scale cascaded feature modules. Our model achieves the overall ranking of No.2 in the ECCV 2018 Chalearn LAP Inpainting Competition Track 3 (Fingerprint Denoising and Inpainting). Among all participating teams, we obtain the MSE of 0.0231 (rank 2), PSNR 16.9688 dB (rank 2), and SSIM 0.8093 (rank 3) on the hold-out testing set.