CQAD: An Image Quality Assessment Dataset for CCTV

This paper presents the CCTV quality assessment dataset (CQAD), a novel image quality assessment (IQA) dataset specifically designed to reflect the complexities of real-world surveillance environments. Existing IQA datasets are largely based on natural images and fail to capture the diverse conditions and degradation characteristics commonly encountered in CCTV footage, such as varying illumination, fixed viewpoints, and environmental noise. To address this gap, CQAD comprises 120 reference images collected from actual surveillance cameras across a range of indoor and outdoor locations, captured under both day and night conditions. Each reference image was degraded using one of six common distortion types, and subjective quality ratings were obtained from 55 human participants using a mean opinion score (MOS) framework. Experimental analysis demonstrates that perceived image quality is affected not only by the type of distortion, but also by scene context, lighting conditions, and time of day. CQAD offers a valuable benchmark for the development of scene-aware IQA models, the evaluation of AI-based video analysis robustness, and the design of restoration techniques tailored to practical surveillance applications.