Year: 2025
Type: article
Abstract: - Abstract: This study introduces an efficient super-resolution (SR) method for enhancing low-resolution fingerprint images using an autoencoder combined with spline and bicubic interpolation techniques. The approach upscales images from 129×97 to 258×194 while preserving fine details and minimizing artifacts. Trained on the FVC 2004 dataset and tested on both FVC 2004 and FVC 2002, the model achieves strong performance with SSIM of 0.968, PSNR of 35.14, and MSE of 0.007. It also attains 100% identification accuracy using the SIFT algorithm, demonstrating its effectiveness for biometric and forensic applications. Key Words: Super-Resolution, Fingerprint, Autoencoder, Spline, Bicubic, PSNR, SSIM, SIFT, Biometrics
DOI: 10.55041/ijsrem49200
Author: Munipally Chakravarthy
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4410925800