No -reference and reduced reference video quality metrics: New contributions

In the past few years, a big effort in the scientific community has been devoted to the development of better digital video quality metrics that correlate well with the human perception of quality. Although many video quality models have been proposed, little work has been done on studying and characterizing the annoyance and visibility of individual artifacts found in digital video applications. In addition, most of the proposed metrics are very complex and require the original video for estimating the quality, making their use in real-time transmission applications very difficult.;In this dissertation, we present an investigation of digital video quality assessment for real-time applications using no-reference and reduced reference video quality metrics following a multidimensional approach that combines individual artifact metrics to produce an overall annoyance model. Since this approach requires a good knowledge of the types of artifacts present in digital videos, we study four of the most relevant artifacts using synthetic artifacts that look like "real" artifacts, yet are simpler, purer, and easier to describe. We develop a system for synthetically generating these artifacts that allows us to study the individual artifacts by themselves or combined and, therefore, provides a powerful tool for research in video quality or video processing. Annoyance models are created by combining the individual artifact perceptual strengths using a Minkowski metric and a linear model.;We propose a set of artifact metrics for estimating blockiness, blurriness, ringing, and noisiness. The metrics are improvements over techniques currently available in the literature and designed with the requirement of being no-reference metrics simple enough to be used in real-time applications. A reduced reference approach is also proposed that does not require any changes to the algorithms. We also obtain a model for overall annoyance based on a combination of the best artifact physical measurement (artifact metrics) using both a Minkowski metric and a linear model.;As an alternative way of implementing a no-reference metric, we propose an objective quality metric based on data hiding. The design of the system includes a psychophysical experiment to evaluate the visibility and annoyance of the artifacts caused by the embedding algorithm. The system also includes a step for estimating the 'best' mark strength based on its visibility and the data hiding capacity of the host video.