A model for the spatial characteristics of the human retina: Optimum intensity-dependent spread filters

In this dissertation, the structure of the intensity dependent spread summation (IDS) is analyzed extensively. It is shown that the major problems associated with the IDS filter is blurring and loss of edge localization at low intensities. The bounds on input image intensity variations which the IDS filter operates on properly (no blurring and loss of edge information) are derived. Then, algorithms to recover reflectance of a surface and image reconstruction of the IDS response are derived.; Next, an optimum intensity dependent spread (OIDS) filter is proposed as a model of the spatial information processings of the human retina. The volume of the filter is held constant and the optimum spread of the filter varies according to the solution to a functional minimization. The variational functional is formulated to minimize image noise subject to smoothing constraint. The resulting filter spread is to account for the degree of latteral excitation and optimum filtering mechanism in the retina. The OIDS filter is shown to resolve the blurring and loss of edge localization at low intensities associated with the intensity dependent spread summation (IDS) filters. This is shown to be due to the compression property inherent in the optimum scale function of the OIDS filters. ftn*Originally published in DAI Vol. 51, no. 11. Reprinted here with corrected bibliographic information.