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

Advances in digital technology have allowed us to embed significant processing power in everyday video consumption devices. At the same time, we have placed high demands on the video content itself by continuing to increase spatial resolution while trying to limit the allocated file size and bandwidth as much as possible. The result is typically a trade-off between perceptual quality and fulfillment of technological limitations. To bring this trade-off to its optimum, it is necessary to understand better how people perceive video quality. In this work, we particularly focus on understanding how the spatial location of compression artifacts impacts visual quality perception, and specifically in relation with visual attention. In particular we investigate how changing the quality of the region of interest of a video affects its overall perceived quality, and we quantify the importance of the visual quality of the region of interest to the overall quality judgment. A three stage experiment was conducted where viewers were shown videos with different quality levels in different parts of the scene. By asking them to score the overall quality we found that the quality of the region of interest has 10 times more impact than the quality of the rest of the scene. These results are in line with similar effects observed in still images, yet in videos the relevance of the visual quality of the region of interest is twice as high than in images. The latter finding is directly relevant for the design of more accurate objective quality metrics for videos, that are based on the estimation of local distortion visibility.