Defining the notion of visual pattern for predicting visual target distinctness in a complex rural background

The goal of this paper is to analyze a new notion of visual pattern that captures the primary details for the task of quantifying the visual distinctness of targets in complex natural scenes. We show that given an objective function, which is defined as the mean of the fraction of correctly classified targets across a number of datasets and whose maximization is desired, the optimal notion of visual pattern in a complex natural background can be defined as congruence in a certain statistical structure at attentional points across a range of 2-D frequency bands. We draw this conclusion from several experiments in which the best definition of visual pattern is estimated based on the relation between the visual target distinctness measured by human observers and a computational distance that applies a simple decision rule to the differences between segregated visual patterns, (C) 2000 Society of Photo-Optical Instrumentation Engineers. [S0091 -3286(00)01802-X].