Evaluating and Improving Image Quality of Tiled Displays
Year: 2015
Type: dissertation
Abstract: I hereby declare that I am the sole author of this thesis. This is a true copy of the thesis, including any required final revisions, as accepted by my examiners. I understand that my thesis may be made electronically available to the public. ii Tiled displays are created by grouping multiple displays together to form one very large display. These tiled displays are often the only suitable option for displaying very large images but suffer from a grid distortion caused by gaps between each sub-display’s active region. This grid distortion is fundamentally different from other, well-studied, image distortions (e.g., blur, noise, compression) and the impact of these grid distortions has thus far not been studied. This research addresses this lack of attention by investigating the grid distortion’s quality impact and creating perceptual algorithms to reduce this impact. We measure the quality impact of the grid distortion by creating two new image quality assessment (IQA) databases for tiled images. These databases provide significant insight into the unique characteristics of the grid distortion and provide a baseline against which to measure the performance of current IQA metrics. We use these databases to show that current metrics do not adequately reflect the quality impact of the grid distortions, and we create a new metric specifically for tiled images that statistically (with 95 % confidence) outperforms current metrics. We improve perceived tiled display image quality by creating new image-correction algorithms based on elements of the human visual system (HVS). These correction tech-niques modify the perceived quality of the displayed images without directly modifying the static grid distortion. These algorithms are shown, through the use of a third subjective user study, to clearly and consistently improve the perceived quality of tiled images. iii
Source: UWSpace (University of Waterloo)
Author: Steven McFadden
Institution: University of Waterloo
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W761225146
