Attention modeling for video quality assessment: Balancing global quality and local quality

Abstract

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.  You may not further distribute the material or use it for any profit-making activity or commercial gain  You may freely distribute the URL identifying the publication in the public portal If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Quantifying the importance of preserving video quality in visually important regions at the expense of background content

“…6 where participants gave the video a score (MOS) different from the one we calculated (ES) by simply averaging the quality of all video regions. The latter observation explains why objective quality assessment algorithms can benefit from utilizing saliency information [36,37]. Moreover, from our results, we would expect that the latter is even the case for quality prediction of videos degraded with diffused coding artifacts, contrary to what has been suggested in [9,21,38].…”

Video Quality Pooling Adaptive to Perceptual Distortion Severity

“…However, predicting human fixations remains a difficult and unsolved problem. More recently, the authors of [23] defined an attention map encapsulating various factors such as color, orientation, motion, etc. when combining local quality scores.…”

Efficient video quality assessment based on spacetime texture representation

“…We have compared our method with a set of VQA algorithms studied in a recent survey paper [9] and two recent visual attention-guided VQA methods ("VA-You" [15] and "VA-Ma" [5]) in Table 2. In comparison, our method demonstrates remarkable quality-prediction performance.…”

“…While most attention-guided VQA methods employ attention models purely based on spatial cues (e.g., color, intensity and spatial orientation), some recent VQA methods take into consideration the motion-driven attention, in which the motion estimates are usually based on the optical-flow representation (e.g. [15]) or simply described by the adjacent frame difference (e.g., [5]). In this study, we reuse the local motion descriptors computed in Section 2.1 for visual attention modeling.…”