Presentation

NoR-VDPNet++: Efficient Training and Architecture for Deep No-reference Image Quality Metrics
SessionTalks
Contributors
Francesco Banterle
Alessandro Artusi
Alejandro Moreo
Fabio Carrara
Event Type
Talk
Interest Areas
Gaming & Interactive
Production & Animation
Research & Education
Registration Levels
Ultimate Supporter
Ultimate Attendee
Exhibitor Ultimate
Time
Location
DescriptionWe present a no-reference metric for the assessment of the image quality of SDR and HDR images in real-time by distilling HDR-VDP2.2 into a convolutional neural network. In this work, we focus on how to improve the training of this metric and how to optimize its architecture for quality.
Contributors
Francesco Banterle
ISTI-CNR
Alessandro Artusi
CYENES Ltd,
Nicosia
Alejandro Moreo
ISTI-CNR
Fabio Carrara
ISTI-CNR