The latest development of no-reference digital image quality assessment
Yanfang Zheng, Feng Wang, Xuebao Li · ICIC express letters. Part B, Applications · 2015
Image quality is a characteristic of an image that measures the perceived image degradation. The study on quality measures is becoming a hot spot in the sense that they can be objectively determined in terms of deviations from the ideal models. There are several techniques and metrics (e.g., Full-Reference (FR) methods and No- Reference (NR) methods) that can be measured objectively and automatically evaluated by a computer program. In this study, we focus on the previous literature of the NR Image Quality Assessment (IQA) methods because the NR IQA fulfills the requirements of actual applications and is capable of playing an important role for real-time image processing system. We concentrate more on the latest methods and their applications in recent 3 years. After a cautious investigation, we realize that the existing methods have some limitations on the multi-scale, specific requirement-based and experience-based aspects. Therefore, the method based on human perceptual features is a direction of the NR IQA future research, and the new challenges arising from real applications will promote the development of future NR IQA methods. The study not only offers a valuable reference for relative studies such as image compression, but also gives a useful guidance for future NR IQA research work.