AI-POWERED IMAGE ENHANCEMENT IN FORENSIC APPLICATIONS: CHALLENGES AND OPPORTUNITIES

Recent progress in machine learning (ML) and computer vision has markedly enhanced the performance of automated facial recognition systems. Nonetheless, forensic applications frequently face issues with poor-quality and low-resolution images that pose challenges even to the most advanced facial recognition technologies. This research explores the potential of neural-based image enhancement and restoration techniques to recover degraded images while maintaining original correspondences for legal use. We evaluate super-resolution and deconvolution methods across two comprehensive and varied facial datasets. Our study employs twelve distinct GAN- and diffusion-based enhancement techniques. Our results offer insights and recommendations for the effective application of image enhancement in forensic facial recognition, highlighting the advantages and potential limitations of these advanced technologies.