In this paper, we deal with the generation of the two-dimensional color image by using interactive evolutionary computation. The tree structural functions to calculate the RGB values for respective pixels are generated by combining predefined functional and terminal symbols, and they are optimized by Genetic Programming. For the initial population, useless monotonous images (for example, all the pixels are black) are often generated because the symbols are combined at random. To solve the problem, we propose the new initialization method. The image entropy is used to measure the complexity of the generated images, and only promising images will be presented to users. Experimental results showed that our method is effective to generate the user's desired images.

2012 IEEE SMC Hiroshima Chapter若手研究会, 2012年7月14日, 広島市立大学