No Reference Video Quality Assessment

Abstract

Video quality assessment is conducted blindly using extracted features from multimedia content, as there is no reference available. However, using NR-based methods to evaluate video quality can be difficult since the extracted features may provide limited information. As the usage of online video traffic continues to increase annually, videos acquired or transmitted by consumers may become distorted due to technological or environmental limitations. Therefore, accurately predicting video quality is crucial. The aim of research on Video Quality Assessment is to develop models that accurately predict the quality of videos, regardless of their content or the type and level of distortions present, in a manner consistent with human judgments. In the last decade, the quantity of recorded, shared, and viewed videos has grown significantly, with sharing of ”in-the-moment” experiences through video becoming increasingly popular on social media platforms such as Instagram and Facebook.