The use of fractal image statistics in the estimation of lateral spatial extent

It has been argued that the characteristics of many commonly occurring surface textures are such that the resulting luminance distributions have the statistical properties of fractals, over a wide range of spatial scales. We show that, when fractal luminance distributions are spatially filtered, the spatial density of zero-crossings obtained is inversely proportional to the scale of filtering, and is not strongly dependent on the fractal dimension of the pattern used. We propose that this predictable property of natural images could provide a basis for the estimation of lateral spatial extent by counting zero-crossings within an interval at a variety of spatial scales, and averaging over spatial scale. We carried out experiments to compare the relative apparent lateral extents of fractal patterns and patterns of equally spaced bars, as a function of the number of bars. The results are in good agreement with theory.