Texture perception in humans and computers: models and psychophysical experiments

Abstract

In this paper, research studies aimed at the development of computational models, specifically for an important preattentive cue, texture, are discussed. The primary basis for the development of computational texture models is to match them with how a human perceives textured patterns. This matching between human and quantitative models is accomplished with the help of an extensive series of psychophysical experiments. Observers view textured patterns, which are generated with a precise mathematical model, and human responses are used to develop a ranking of distinctness of texture patterns. We illustrate the use of these computational and psychophysical experimental tools for the analysis and characterization of real-world high resolution imagery acquired in thermal and visible bands. These results are applicable to image clutter characterization, camouflage assessment, and automatic target detection and recognition tasks.