Abstract
Video quality assessment (VQA) aims to evaluate the video quality consistently with the human perception. In most of existing VQA metrics, additive noises and losses of primary visual information (PVI) are decoupled and evaluated separately for quality assessment. However, PVI losses always include different types of distortions such that PVI distortions are not evaluated well enough. In this paper, a novel full-reference video quality metric is developed by decoupling PVI distortions into two classes: compression distortions and transmission distortions. First, video denoising method is adopted to decompose an input video into two portions, the portion of additive noises and the PVI portion. Then, maximal distortion regions searching (MDRS) algorithm is designed to decompose PVI losses into transmission distortions and compression distortions. Finally, the three distortions are evaluated separately and combined to compute the overall quality score. Experimental results on LIVE database show the effectiveness of the proposed VQA metric.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call