Automatized computer system for evaluation of rust using modified single-scale retinex

The problem for assessing the degree of rusting is considered. The rust percentage on the coatings can be objectively computed using digital image processing methods. As illumination can significantly influence on the result of segmentation the basic model of single scale retinex is analyzed for luminosity equalization. It is shown that this model is based on the usage of local contrast inversion. The parametric model of luminosity equalization is built. It is based on logarithmic transformation of inversion of weighted contrast. After this preliminary rust image processing the global thresholding is applied. Percentage of segmented rusted area is used for designating the degree of rusting.