PFI-TT: Automatic Diagnosis of Video Quality Problems in Networked Camera Systems

University Of Cincinnati Main Campus, Cincinnati OH

Investigators

Abstract

The broader impact/commercial potential of this Partnerships for Innovation - Technology Translation (PFI-TT) project will enable a networked camera system to produce videos with better quality. This project will design and demonstrate a software prototype that automatically detects various video quality degradation scenarios, provides feedback to human operators on the cause of quality degradation, and recommends strategies to improve the quality of received videos. The proposed prototype will save manual labor used for inspecting video quality problems in networked camera systems. The solution can benefit various video use cases in public safety, such as indoor and outdoor monitoring, traffic surveillance, access control, and emergency operations. The results from this project could also impact the design of a wide range of intelligent systems with visual sensing in applications like smart health monitoring, industrial process control, smart and connected vehicles, and virtual/augmented reality systems. The students participating in this project will receive training in technology commercialization and entrepreneurship. The project will also include educational and outreach activities to broaden the participation of under-represented minority students, helping to build the future workforce in the field of computing. The proposed project aims to develop and demonstrate a video quality assessment and adjustment software prototype for networked camera systems. Multiple factors cause the degradation of video quality in a networked camera system, such as noise or motion blur during video capturing, degraded resolution, too much compression, and bad network conditions. This project will develop a systematic video quality solution for complex networked camera systems with sensing, processing, and communication components. The proposed research will focus on: (1) investigating the system settings and the data generated from a practical enterprise-level surveillance system, (2) leveraging the characteristics of video quality evaluated by both human users and automatic video analytics tools, (3) predicting video quality through light-weight image analysis and video bitstream analysis, and (4) predicting the types of video distortion and applying corresponding quality restoration algorithms. The proposed technologies will be integrated in a software prototype that could be used by the operators of enterprise-level camera systems. Iterative development and evaluation tasks will be carried out to ensure that the proposed prototype works well for systems with different types of cameras, network scales and connections, and processing units. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.