Fusion-based Super-Resolution for Facial Identification and Sketch-to-Image Translation

Advancements in technology have inadvertently
enabled easier access to crime-related data, resulting in an
increase of criminal activities. Law enforcement agencies
encounter substantial difficulties in identifying suspects and
preventing further crimes. To tackle these challenges, this
research introduces a proposed solution of fusion-based superresolution
approach combined with sketch-to-image translation
techniques. The method enhances low-quality facial images and
using ESRGAN and Real-ESRGAN. A fusion model integrates
multiple enhancement strategies, followed by a Siamese model
for face verification system. The experimental results
demonstrate improved recognition accuracy and image quality,
supporting efficient suspect identification. The proposed model
shows significant improvement in areas of criminal sketches.
This approach effectively strengthens facial recognition systems
in real-world surveillance and forensic scenarios.