Enhancing Fingerprint Image Resolution Using Auto Encoder and Interpolation Techniques

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