Title: Evolutionary algorithm for example-based painterly rendering

Authors: Bara'a Ali Attea, Laylan Mohammad Rashid, Wafaa Khazzal Shames

Addresses: Computer Science Department, College of Science, University of Baghdad, Baghdad, Al-Jadiria, Iraq. ' Computer Science Department, College of Science, University of Baghdad, Baghdad, Al-Jadiria, Iraq. ' Physics Science Department, College of Education Ibn Al-Haitham, University of Baghdad, Baghdad, Al-Waziriya, Iraq

Abstract: NPR may be seen as any attempt to create images to convey a scene without directly rendering a physical simulation. This paper describes an evolutionary algorithm (EA) for learning painting styles from example images. The evolutionary algorithm has two main stages: learning and painting. By learning the painting style from a source pair of training images (a source photograph and its artistic style), a style approximation can then be accomplished to another target image. The mechanism of evolutionary algorithms is used here to learn or capture different characters used to depict different regions of the source painting and use them to convert similar regions with similar texture statistics of the target image into artistic rendering. Two main perturbation operators are used in the proposed EA: a multi-sexual recombination and mutation operators. On overall, the algorithm, with its implementation simplicity and computation efficiency, can provide acceptable results with perceived artistic rendering.

Keywords: artistic rendering; multi-objective evolutionary algorithms; MOEA; non-photorealistic rendering; NPR; recombination operators; learning; painting styles; example images.

DOI: 10.1504/IJBIC.2010.032129

International Journal of Bio-Inspired Computation, 2010 Vol.2 No.2, pp.132 - 141

Available online: 10 Mar 2010 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article