|
Work item:
|
P.OQAI
|
|
Subject/title:
|
Objective quality assessment of AI-enhanced video content
|
|
Status:
|
Under study
|
|
Approval process:
|
AAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2026-12 (Medium priority)
|
|
Liaison:
|
-
|
|
Supporting members:
|
Telefon AB - LM Ericsson; Ilmenau University of Technology; Blekinge Institute of Technology(BTH)
|
|
Summary:
|
In recent years, significant developments have been made in both resolution upscaling and framerate up-sampling techniques that can be used to enhance the overall video quality. These techniques are mainly based on deep learning approaches such as convolutional neural networks (CNNs), generative adversarial networks (GANs), etc. These approaches may introduce artifacts that are different from the ones produced when using traditional signal-processing-based approaches. Consequently, the existing video quality assessment models may not be able to take into account such artifacts in the overall quality estimation process. Hence, as part of this work item, the focus is on the creation of databases that consist of impairments generated by such deep learning approaches applied to the already compressed video sequences with human annotated quality judgments and using them as ground truth for the subsequent model development process. The subjective test plan for these databases will be discussed and decided on by the project participants. Further details regarding this Work Item will be described in Terms-of-Reference (ToR) and Requirement Specification documents to be created during the project.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2025-01-23 16:49:06
|
|
Last update:
2025-01-23 18:43:07
|
