Restoration and Enhancement Learning Techniques Supporting Gun Image Recognition

• Hoai Nam Vu,
• Bao Q. Bui,
• Khanh G. Nguyen,
• Dat A. Tran,

Vol. 19, No. 10, October 31, 2025
10.3837/tiis.2025.10.001, Download Paper (Free):

• Deep Learning
• Image recognition
• Image restoration
• gun recognition
• multi-frame synthesis
• surveillance enhancement

Abstract

Gun image recognition in surveillance camera footage has emerged as a pivotal challenge in the field of computer vision, particularly in security and law enforcement applications. Traditional approaches often rely on processing distinct neural networks tailored to handle specific types of image noise encountered in complex environments, which can be both time-consuming and computationally expensive. This necessitates the development of unified image enhancement techniques capable of effectively restoring and improving image quality from single images or image sequences to facilitate accurate gun recognition. In this study, we introduce R2IQ, a novel and efficient image quality enhancement network designed to address these challenges. Our method is inspired by multi-image learning strategies and the synthesis of features across camera frames, enabling the reconstruction and enhancement of image quality. We conduct extensive experiments to validate the efficacy of our approach. The results demonstrate that R2IQ outperforms existing state-of-the-art methods in image quality enhancement and achieves competitive performance on the NTIRE2024 and T-Gun datasets. Furthermore, we introduce T-Gun, a novel dataset for gun recognition in simulated armed robbery scenarios, to validate our approach.

Statistics

Cite this article