Painterly renderings using a synthesis of styles based on visual perception

Abstract

In this paper, we propose a painterly image generation system which generates images based on input 3-dimensional models, and which uses a systematization of painting styles in modern paintings based on visual perception. We present methods for parametric transformation of each painting style by controlling parameters based on vision in our proposed system. In this research, we systematized painting styles used in modern paintings, such as Impressionism, Cubism, and abstract paintings, by introducing parameters that include form vision, space vision and color vision, all of which are associated with visual information processing. Based on this, our proposed image generation system determines the painting styles to be applied and their levels of abstraction according to the amount and balance of visual parameters, and then performs parametric transformation from an input 3D model. Finally the system generates a painterly image in which the respective features related to form vision, space vision and color vision have been emphasized, diminished or omitted.