Modelling of the stereoscopic HVS

We give an overview of fields of study, related with modelling of the human visual system and perception of visual information – HVS anatomy and physiology, HVS modelling, and visual quality metrics. We present the state‐of‐the‐art in HVS modelling, including single and multichannel models, and models of specific HVS properties – light adaptation, contrast sensitivity, spatiotemporal masking. We pay speciall attention to the visual mechanisms responsible for stereoscopic perception. We describe the algorithms behind the most popular visual quality metrics, and present a comparative study of 15 metrics over a database of stereoscopic video streams. Additionally, we present studies on metric performance evaluation, and available databases with video and image test sets. Finally, we present a model of the stereoscopic HVS, which is suitable for perceptual quality We propose an image processing channel which can be used for 3D video quality estimation, and selection of HVS models suitable for that channel. In this work, we aim at creating a model of the stereoscopic human visual system (HVS), which to be later used for perceptual quality estimation of 3D video streams. Such model is an outcome of a multidisciplinary study, which combines knowledge about HVS anatomy and physiology, psychophysical experiments, models of HVS features, and overview of existing visual quality metrics. We begin with an overview of the anatomy of the eye, optical nerve and areas of the brain cortex, which are dealing with vision. We study the continuous process of visual perception, and the specific properties of vision – light adaptation, contrast sensitivity, colour and motion perception. We pay special attention to the visual mechanisms which facilitate stereoscopic perception. We present an overview of the state-of-the-art in HVS modelling. Models of various HVS properties are discussed – such as contrast sensitivity function, perceptual colour spaces, temporal and spatial masking models. Additionally we study the different approaches in HVS modelling – single-channel and multi-channel models, psychophysical and neurobiological modelling. We give an overview and classification of visual quality metrics. We describe the algorithms behind the most popular ones. We present the parameters, used for quality metric performance evaluation, available databases of test images and the relevant VQEG studies on performance of video quality metrics. We present an implementation and comparative study of 15 metrics. Each metric is tested over the database of impaired 3D videos created earlier in our project. The quality score and computational performance of each metric is measured …