Image quality is defined as the degree of excellence the image conveys. It is assessed using visual (image) psychophysics and predictive modeling. Engineering type image quality models, employed in imaging system design and optimization, are based on the assumption that the perceived image quality is a function of both the imaging system performance and the human visual system performance. Such models employ imaging system performance parameters related to color, sharpness, resolution, noise, etc. as input parameters, along with models of the human visual system. System performance parameters are typically derived from simple test charts, captured under strict laboratory conditions. Similarly, vision models employed in image quality modeling are largely based on visual measurements obtained from simple test stimuli, such as sine-waves, Gabor functions, random noise, and uniform color patches. The modeled image quality is, as a result, system-dependent but scene content-independent. Computational models of image quality on the other hand base their predictions on the contents of captured and processed images. The modeled image quality, in this case, is scene content-dependent, but the system performance parameters are undifferentiated.
There are many unresolved challenges when it comes to the assessment and modeling of image and video quality. For a start, engineering-type quality models that account for the imaging system performance often fail to account for the characteristics in captured natural scenes. Yet advanced algorithms incorporated in commercial imaging systems are image content aware and become more or less “active” depending on the individual image structures, local tones, and colors. On the other hand, computational quality models derived directly from images of individual natural scenes often fail to relate in a meaningful manner to the imaging system parameters and are therefore unfit for informing imaging system design and optimization. Further, since most common human vision models employed in image quality modeling are based on simple stimulus viewing, they do not account for adaptation mechanisms that occur in free viewing of different scene contents. The validity and implementation methods of such models in image quality assessment are debatable. Finally, research on scene- and process- dependent image quality assessment is necessary in the evaluation of the robustness of non-human (machine) perception systems too. Such research is still in its infancy. The goal of this Research Topic is to act as a platform for addressing and debating the aforementioned unresolved challenges associated with image and video quality assessment, as well as relevant visual system modeling.
We are particularly interested in research related to imaging performance, image/video quality, and vision assessments that use images of complex (natural) scenes in replacement of traditional test charts and simple visual stimuli. We welcome Original Research articles, Reviews, and Brief Research Report articles.
Areas of interest to the Research Topic include (but are not limited to):
• Imaging system performance derived from natural scene images
• Image/video quality modeling accounting for imaging system performance and scene contents
• Visual system assessment and modeling using (images of) natural scenes
• Effects of scene contents on image quality assessment and in visual system modeling
• Image quality modeling for non-human vision systems
Image quality is defined as the degree of excellence the image conveys. It is assessed using visual (image) psychophysics and predictive modeling. Engineering type image quality models, employed in imaging system design and optimization, are based on the assumption that the perceived image quality is a function of both the imaging system performance and the human visual system performance. Such models employ imaging system performance parameters related to color, sharpness, resolution, noise, etc. as input parameters, along with models of the human visual system. System performance parameters are typically derived from simple test charts, captured under strict laboratory conditions. Similarly, vision models employed in image quality modeling are largely based on visual measurements obtained from simple test stimuli, such as sine-waves, Gabor functions, random noise, and uniform color patches. The modeled image quality is, as a result, system-dependent but scene content-independent. Computational models of image quality on the other hand base their predictions on the contents of captured and processed images. The modeled image quality, in this case, is scene content-dependent, but the system performance parameters are undifferentiated.
There are many unresolved challenges when it comes to the assessment and modeling of image and video quality. For a start, engineering-type quality models that account for the imaging system performance often fail to account for the characteristics in captured natural scenes. Yet advanced algorithms incorporated in commercial imaging systems are image content aware and become more or less “active” depending on the individual image structures, local tones, and colors. On the other hand, computational quality models derived directly from images of individual natural scenes often fail to relate in a meaningful manner to the imaging system parameters and are therefore unfit for informing imaging system design and optimization. Further, since most common human vision models employed in image quality modeling are based on simple stimulus viewing, they do not account for adaptation mechanisms that occur in free viewing of different scene contents. The validity and implementation methods of such models in image quality assessment are debatable. Finally, research on scene- and process- dependent image quality assessment is necessary in the evaluation of the robustness of non-human (machine) perception systems too. Such research is still in its infancy. The goal of this Research Topic is to act as a platform for addressing and debating the aforementioned unresolved challenges associated with image and video quality assessment, as well as relevant visual system modeling.
We are particularly interested in research related to imaging performance, image/video quality, and vision assessments that use images of complex (natural) scenes in replacement of traditional test charts and simple visual stimuli. We welcome Original Research articles, Reviews, and Brief Research Report articles.
Areas of interest to the Research Topic include (but are not limited to):
• Imaging system performance derived from natural scene images
• Image/video quality modeling accounting for imaging system performance and scene contents
• Visual system assessment and modeling using (images of) natural scenes
• Effects of scene contents on image quality assessment and in visual system modeling
• Image quality modeling for non-human vision systems
