A Statistical Approach to Local Evaluation of a Single Texture Image
Petr Somol · 2007
The concept of texture implicitly suggests some local shift invariant statistical properties. Motivated by this idea we have shown in a series of papers that textures can be modelled by estimating the joint probability density of grey levels in a suitably chosen observation window. In this paper we apply the estimated density to evaluate the original texture image locally. At each position of the window the normal mixture density is computed for the circumscribed texture and the corresponding log-likelihood value is presented as a grey level at the central pixel of the window. In this way we can visualize the lowand highly probable parts of the original texture. The method is applicable e.g. to identify defects or abnormalities in grey-scale textures.