2002 年 58 巻 10 号 p. 387-394
This paper presents a method for objective evaluation of visual complexity for colored patterns in terms of two features obtained by image analyses. One of the features is entropy measure in the spatial frequency domain, and the other is fractal dimension of the colored pattern. The evaluation method was applied to colored mosaic patterns generated by using normal random numbers. On the other hand, 20 human observers estimated the visual complexity and beauty of the mosaic patterns on a CRT display. The result showed that the entropy measure was highly correlated to average value of complexity estimated by 20 observers, and the fractal dimension was also correlated to the visual complexity. Then, a linear multiple regression equation was introduced for quantitative evaluation of complexity, including the entropy measure and the fractal dimension as independent variables. Visual complexity calculated by the regression equation was in good agreement with that evaluated subjectively. Furthermore, the relationship between visual complexity and beauty was investigated. It was found that the colored patterns, which had medium complexity, were evaluated as beautiful. In other words, very complex or very simple patterns were not evaluated as beautiful.