Resolution Loss by Pointwise Degradations
Pointwise image degradations are caused by several sources; added specular or indirect reflections, additive scattering, radiance attenuation in haze and flash photography. It is found through a linear-systems analysis that in these cases there is a cutoff frequency induced by noise despite having no optical blur in the imaging model. This low signal-to-noise ratio (SNR) results in resolution loss similar to image blur. The intention of this research is to perform image recovery on images which suffer from this phenomenon, specifically in hazy images.
While image recovery under noise is widely studied, there is little emphasis on performance as a function of object size. This research results in a tool to assess the ability to recover an object or feature in an image depending on its size, distance from the camera and radiance difference from its nearby background per attenuation coefficient of the medium. This analysis also suggested the potential of polarization in visibility enhancement and allowed for a quantization of its viability.