Item

Abstract

Natural scenes are fractal in space (ie they have 1/f B spatial frequency spectra) and time (1/f A temporal spectra), and can be compellingly mimicked by fractal textures. If dynamic fractal texture statistics are used to describe natural scenes, then data on discriminability of such textures are required.

The smallest detectable change was measured separately for 10 spatial (0.4 to 2.2) and 8 temporal exponents (static, and 0.2 to 1.4) with an adaptive staircase. Computational constraints limited each fractal to 64 frames (~ 2 s) of 64 × 64 pixel images. Spatial discriminations were easiest when the spatial exponent B was ~ 1.6 and were similar across all temporal exponents. Temporal discriminations were easiest when the temporal exponent A was ~ 0.8, and increased in difficulty as the spatial exponent increased.

This similarity in spatial discrimination thresholds for static and dynamic fractals suggests that the spatial and temporal dimensions are independent in dynamic fractals (at least for spatial judgments), as is often assumed. The dependence of temporal judgments on the coarseness of the texture (ie on the spatial exponent) is understandable, as a 1 mm change in position is more noticeable for a 1 mm object than for a 100 m object.