*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Perceptual Video Quality Assesment and Analysis using Adaptive Content Dynamics
Committee:
Dr. Ghassan AlRegib, ECE, Chair , Advisor
Dr. Anthony Yezzi, ECE
Dr. David Anderson, ECE
Dr. Elliot Moore, ECE
Dr. Berdinus Bras, ME
Abstract:
In this dissertation, the objective of the proposed research is to investigate perceptual quality assessment and analysis of videos subject to different types of distortion. We propose utilizing adaptive content dynamics to examine the impact of different error sources on the perceptual quality of the video. We design perceptual video quality estimators using novel handcrafted features inspired by the human visual properties. We explore new feature spaces and utilize them to capture varying video dynamics as experienced by our visual perception. Specifically, we introduce a new framework for perceptual video quality using pixel-level optical flow maps where we propose a motion processing procedure inspired by the hierarchical processing of motion in the visual cortex. Furthermore, we propose another perceptual video quality assessment approach by examining the varying properties of the temporospatial power spectrum. Using the power spectrum, we design a novel sensitivity measure to capture the impact of distortions on visual perception. This work includes a full-reference computationally efficient framework that captures both spatial and temporal characteristics in the frequency domain. We also examine the performance of various statistical moments and pooling strategies, at both spatial and temporal levels, with different visual feature maps. This aims at revealing the optimal pooling strategies most correlated with visual perception for every feature space with respect to different distortions.
