ENTROPY VERSUS FRACTAL COMPLEXITY FOR COMPUTER-GENERATED COLOUR FRACTAL IMAGES (PO45, Pages 432-437)

The complexity notion is shared between scientific fields associated to vision or surface assessment, especially non-uniform surfaces. Nevertheless, there is no single definition allowing comparing experimentally the perception induced by observing a colour textured surface, the measure extracted from the existing metrological tools or from image processing approaches. Among the existing definitions, entropy is often used as a measure directly correlated to the surface complexity. The fractal dimension associated to the auto-similarity property is well adapted for mathematical developments in image analysis and corresponds for some natural structures to the inner organization on several analysis scales. In the context of CIE TC 8-14 dedicated to the definition and assessment of the spatio-chromatic complexity, we are interested by the theoretical and experimental comparison of different ways to assess the complexity of colour textured surfaces. In this work, we compare the entropy against the fractal dimension for computer-generated fractal surface images.