Cryptic animal coloration or camouflage is an adaptation that decreases the risk of detection. The study of the evolution of camouflage has strongly emphasized the minimization of visual information that predators receive from prey, by means of background matching. However, the evolutionary effects of information processing after its reception have been virtually ignored. I constructed a model that employs an artificial neural network and simulates the evolution of prey coloration in a visually complex and simple habitat. The model suggests: (1) the difficulty of a detection task is related to the visual complexity of the habitat; (2) it is easier to decrease the risk of detection by the means of camouflage in a visually complex habitat; (3) selection on camouflage can exploit limitations in predators information processing; and (4) there are shortcomings in using the degree of background matching as the measure of camouflage.