Human perception of halftoned images

Abstract

An empirical study of human perception of halftoned images was conducted to determine which of five different halftoning algorithms generated the best images. The subjects viewed each of the 20 stimuli (halftoned images) at two distances, and although all images were more preferred at the far distance, the rating of the pictures was dependent upon the algorithm used in generation. Images containing a high level of detail were rated highest when halftoned by the neural network and the simulated annealing algorithms of [4], whereas pictures that had little detail and many smooth surfaces were rated highest under the Floyd-Steinberg model [3].