A Naive Complexity Measure for Color Texture Images

The notion of complexity is widely-known and used. Various definitions of complexity exist: the Hausdorff dimension, fractal dimension, Kolmogorov complexity, Krohn-Rhodes complexity, Lyapunov exponents or the entropy, some of them with several definitions. All these measures were defined strictly for mathematical objects, but they may apply to real signals like texture images in particular. We are interested in new definitions of complexity or how the existing ones can be extended to color and spectral texture images. In this paper, we propose the definition of a naive complexity measure for color texture images as three times the number of colors divided by the image resolution. We show that such a simple definition may have interesting properties, by comparing its performance to the color entropy previously defined. We show and discuss our experimental results, then draw the conclusions.