Testing the efficacy of computer-generated cartoons

Reports experiments which examined how well full cartoons of famous faces could be recognised, compared with grey-level images, and assessed the separate contribution played by each of the `valledges´ and `threshold´ components. The results show that full cartoon drawings produced by the Pearson and Robinson (1985) operator are identified almost as accurately as grey-level images, but neither the `valledge´ nor the `threshold´ components alone were as well identified as the full cartoons. This finding is in line with previous work, and supports the suggestion that separate contributions to recognition may be made by components of `line´ and `mass´. It may be by the inclusion of valleys, edges and dark areas (cf. `line´ and `mass´) that the Pearson and Robinson operator is successful in creating natural, recognisable cartoons. Further experiments on the effect of negating different components of the cartoon are described