Recent progression of artificial intelligence driven no-reference video quality assessment techniques in multimedia services

The importance of maintaining consistent visual quality has increased significantly due to the rapid advancement of video acquisition technologies, high-speed Internet connectivity, the widespread adoption of social media platforms, and the growing utilization of multimedia services across 4G and 5G networks. Video-based applications ranging from entertainment and online communication to surveillance and telemedicine continue to generate substantial network traffic and bandwidth consumption. Consequently, service providers are under significant pressure to meet the increasing expectations and demands of their clients as a result of this surge. Their primary goal is to deliver superior Quality of Service (QoS) and Quality of Experience (QoE) to all users. However, the high demand for video applications leads to significant bandwidth consumption, making it challenging to offer good QoS and QoE with limited network resources. In this context, video quality assessment (VQA) techniques, particularly No-Reference (NR) VQA methods, can help service providers measure and maintain video quality while optimizing network resources. This paper presents a comprehensive review of VQA techniques with a primary focus on NR-VQA methods. It begins by introducing various types of video distortions and their influence on perceptual quality, followed by an overview of the Human Visual System (HVS) and its role in visual perception. The review further classifies existing VQA methodologies and summarizes widely used VQA databases and evaluation metrics. In addition, a critical analysis of handcrafted feature-based, deep learning-based, and hybrid NR-VQA techniques is presented by highlighting their strengths, limitations, and application scenarios. Finally, the review synthesizes recent research progress, identifies existing challenges, and outlines promising future research directions for developing robust and perceptually aligned NR-VQA systems.