Perceptual Quality Modeling And Application Of Immersive Images Considering Peripheral Vision Impact

Conventional images/videos are often rendered within the central vision area of the human visual system(HVS)with uniform quality.Recent virtual reality(VR)device with head mounted display(HMD)significantly extends the field of view(FoV)which includes both central and peripheral vision areas.Immersive image or video,usually presented with extended FoV and high resolutions,demands a large amount of network bandwidth for delivery with lossless content.Thus,in reality,the key problem is how to ensure the high-quality immersive media delivery with limited network bandwidth.To tackle this,we explore the unequal sensations of human visual system on immerisve images/videos when wearing the HMD,building analytical models to quantitatively measure such unequal sensation.These models would be utilized to place unequal quality scales across regions of rendered images/videos,leading to significant bandwidth reduction,but without the sacrifice of perceptual quality.The main contributions of this thesis are listed as follows:1.It's known that human eye exhibits the unequal image quality sensation among different areas due to the non-uniform distribution of photoreceptors on our retina.We propose to study the sensitivity of the human visual system with respect to the eccentric angle ? across different vision areas considering peripheral vision.Often times,image quality is controlled by the quantization stepsize q and spatial resolution s,separately.Therefore,we measure the visual threshold on q or s at different ? through extensive human subject tests.Furthermore,we develop analytical models that expressq-threshold or s-threshold as a function of ?.2.As the image quality is often jointly adjusted via both resolution and quantization step size,we further study the joint model of q-and s-threshold over the FoV and obtain the detailed threshold distribution.3.For the subjective assessment of immersive content with non-uniform quality settings,we derive the mapping between this image and its corresponding uniform quality settings as they give the similar visual experience.Then we combine the existing study and the mapping function to obtain the perceived quality model.These models can be applied to set different quality weights at different regions,so as to significantly reduce the transmission data size but without subjective quality degradation.We demonstrate that image rendering in a gigapixel imaging system can be speed up about 10× with the model-guided unequal quality configuration across various regions,in comparison to the the legacy scheme with uniform quality scales everywhere.