Towards the optimization of visual inputs

Vision can be regarded as 'inverse optics', i.e., the process in which measured characteristics of an optical image of the environment are used to reconstruct the material properties of this environment. Depending on the type of measurement that is being performed, flexibility of the metric upon which the measurement results are represented can be exploited to optimize the discriminability of the items whose characteristics are measured. We will derive approximate analytical expressions for such optimal metrics and discuss the implications of this idea to (the optimization of) the quality of displayed images.