Information Theoretic Based Image Quality Evaluation

A framework is proposed for predicting image quality based on ensemble evaluation of the Shannon mutual information between the perceived image and the original pristine object. The mutual information is evaluated statistically using simulations of the image formation chain including radiometric conditions, camera motion and turbulence induced blurring, physics of the imaging device, image formation and enhancement processing, display, and a human visual system model. Preliminary evaluations demonstrate strong agreement with the General Image Quality Equation over the range of its validity and to visual assessment experiments.