Word descriptors of image quality based on local dispersion-versus-location distributions

We consider an image descriptor that is derived from the scatter plot of range.vs.midrange values, taken uniformly across the image. The points in the scatter plot are classified as belonging to one of four main regions: high-contrast/medium-luminance, low-contrast/high-luminance, medium-contrast/medium-luminance and low-contrast/low-luminance; finally, a word descriptor is obtained by quantizing, normalizing and ordering the percentages of points in each of the four regions.