Histogram analysis and synthesis in processing series of images: Extending the concept of image similarity
В. А. Аринин · Pattern Recognition and Image Analysis · 2008
Specific elementwise operations over images based on analyzing and synthesizing histograms are described. An image is seen as part of an essentially two-dimensional continuous function given on a discrete grid and normalized to an interval of [0; 1]. Images with degenerate histograms, i.e., images that include big continuous sections or a small number of quantization levels, are not considered. A histogram is a distribution of the image function, which is discrete in our case, with its sum (or area) normalized to 1. Such normalization makes operations invariant with respect to pixel size of the image and allows, with no loss of generality, changing sizes of the sampling grid (resampling) and adding uniformly distributed noise for ill-conditioned histograms.