A criterion for resolution of ambiguity of raster images

Although raster images are inherently ambiguous, humans exhibit uncanny abilities to come up with their likely interpretations. A methodology to simulate such behaviour on a computer is presented here. A numerical method for rejection of unlikely interpretations is given. A likely interpretation selected is found to depend on a particular raster pattern under study, but also on the architecture of the machine performing the study, as well as on the machine language used to elaborate possible interpretations. A number of suggestions for further research emerge, pertaining to the image description languages, machine vision computer architectures, graph interpretation systems, as well as comments on the feasibility of automatic, intelligent retouching of partially damaged images