|
| |
|
Project: Perceptual Video Assessment
People
Objective
Many approaches have been proposed for estimating the perceived quality of digitally com춑ressed and reconstructed video clips. Several metrics rely on human vision models
developed to predict the detection thresholds of basic signals. These human vision models tend to be complex and require the selection of a large number of parameters. As part of this project, we are conducting a number of psychophysical experiments. These
experiments are designed to answer several questions:
-
Do existing detection threshold models accurately predict the detection threshold of complex sig춏als in natural images? Can
the models be simplified or implemented more efficiently?
-
Given accurate detection thresholds, can the perceived quality of different video clips be accurately predicted?
-
How do the detection thresholds and the perceived quality depend on the appearance (blocky, blurry, fuzzy, etc.) of the artifacts introduced into a video?
How do they depend on the size and duration of the artifacts?
-
How important is the content of the video in determining the perceived quality of impaired video?
The experiments are structured to measure both the quality and some intermediate perceptual value for each test video clip.
For example, our initial experiments measured the detection threshold and the
quality of various video clips that were partially corrupted with severe MPEG-2
compression artifacts. The data were used to estimate the detection thresholds for these artifacts.
The detection thresholds have been used to evaluate the performance of detection models on real signals
(Question 1) and to help predict the perceived quality (Question 2) [1], [2].
The initial experiments were performed using severe MPEG-2 compression artifacts. These artifacts looked blocky and blurry. In additional
experiments, other types of artifacts will be used in experiments to help answer Question 3. The planned experiments include less severe MPEG-2 artifacts, artificial noise distributions, and MPEG-4 artifacts. Results from our initial experiments indicate that there are consistent shifts in detection threshold
and perceived quality as the appearance of MPEG-2 artifacts changes [2]. If these results can be extended to the other kinds of artifacts, then a single, simple model may be used to predict quality for all of these noise sources.
Other experiments will measure quantities such as the perceived strength of various artifacts and the importance of various regions in the
videos. Each of these measured quantities can be used to test both models developed to predict the quantity and models that predict quality using the quantity.
For example, using measured spatial importance values we can evaluate importance
map models that attempt to predict the most
significant regions in a video. We can also use the measured importance values to see if strong artifacts in important regions actually affect quality more than strong artifacts in unimportant regions (Question 4).
Publications
These materials are presented to ensure timely dissemination of scholarly and
technical work. Copyright and all rights therein are retained by authors or by
other copyright holders. All persons copying this information are expected to
adhere to the terms and constraints invoked by each authors copyright. In most
cases, these works may not be reposted without the explicit permission of the
copyright holder.
-
Michael S. Moore, John M. Foley, and Sanjit K. Mitra,
"A
comparison of the detectability and annoyance value of embedded MPEG-2
artifacts of different type, size and duration," Proceedings of the SPIE, Human Vision and Electronic Imaging
VI, San Jose, CA, vol. 4299, Jan. 2001. [abstract]
-
Michael S. Moore, John M. Foley, and Sanjit K.
Mitra, "Detectability and Annoyance Value of MPEG-2 Artifacts Inserted into Uncompressed Video
Sequences," Proceedings of the SPIE, Human Vision and Electronic Imaging V, San Jose, CA, vol. 3959, p. 99-110, January 2000.
[abstract]
|
|
