BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260727T023745EDT-93260MVvsi@132.216.98.100
DTSTAMP:20260727T063745Z
DESCRIPTION:Abstract\n\nImage Quality Assessment (IQA) plays a critical rol
 e in optimizing visual experiences by approximating human perception of im
 age quality. However\, existing IQA methods often fail to generalize acros
 s variations in illumination\, display properties\, and content\, which ar
 e all too common in real-world scenarios involving modern display technolo
 gies. In this thesis\, we address these challenges through a series of con
 tributions spanning perceptual studies and applications of deep learning f
 or IQA. First\, we conduct a subjective experiment to quantify the influen
 ce of ambient illumination on human perception of image quality. To comple
 ment these findings\, we introduce a framework that extends the applicabil
 ity of existing IQA methods to a wider range of illumination and display p
 arameters\, effectively modeling viewing conditions from complete darkness
  to bright daylight. Next\, we explore the application of vision transform
 ers (ViTs) for IQA\, analyzing the feature representations of various pre-
 trained ViTs to identify architectures better suited for IQA and to gain i
 nsights into how these models encode image quality distortions. Building o
 n this analysis\, we introduce Vision Transformer for Attention Modulated 
 Image Quality (VTAMIQ)\, a novel full-reference IQA model that leverages V
 iTs to capture global dependencies in images and achieves state-of-the-art
  performance on standard IQA datasets. Finally\, while most existing IQA m
 ethods and datasets are tailored for Standard Dynamic Range (SDR) imaging\
 , we address the challenges of training deep IQA models on High Dynamic Ra
 nge (HDR) data by integrating specialized fine-tuning and domain adaptatio
 n techniques. Models trained with our approach surpass previous baselines\
 , converge significantly faster\, and reliably generalize to HDR inputs. A
 ltogether\, our findings offer valuable insights into how viewing conditio
 ns influence human perception of image quality and support the development
  of more robust and generalizable IQA models\, enhancing their adaptabilit
 y and performance in real-world applications.\n
DTSTART:20250703T140000Z
DTEND:20250703T160000Z
LOCATION:Room 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H
 3A 0E9\, 3480 rue University
SUMMARY:PhD defence of Andrei Chubarau – Vision Transformers for Image Qual
 ity Assessment on High Dynamic Range Displays
URL:https://www.mcgill.ca/ece/channels/event/phd-defence-andrei-chubarau-vi
 sion-transformers-image-quality-assessment-high-dynamic-range-displays-365
 928
END:VEVENT
END:VCALENDAR