Image analysis for advertisement purposes: A computational model of visual perception

Abstract

The present contribution describes a new procedure to evaluate the optical conspicuousness of advertisement boards in a natural environment. The object of this procedure is the computer-aided positioning of advertisements, as well as optimizing the layout of individual boards or posters. For this purpose, the scenario to be examined within a modelling process is first geometrically described. The geometric models thus created contain additional attributes, such as color and luminance at calibrated illumination. A ray casting process then creates an image of the board to be evaluated and of its environment from an observer position to be analyzed. To evaluate the optical conspicuousness a two-pass-model of the human visual system is presented, modelling the retina by way of isotropic Laplace-filters and simple cells of the visual cortex by anisotropic edge extractors and Fourier transform. The image analysis giving a quantitative measure for optical conspicuousness is carried out on the image presentations appearing in the individual model phases. The method of the procedure described here is documented with examples and compared with classical methods of eye-tracking in psychology of advertising.