Morphogenetic and evolutionary approach to similar image creation

Abstract

This study presents a method for creating various images similar to an input image. The method first generates a graph which node corresponds to a pixel of an input image and which edge is made between nodes if a given condition on similarity in brightness between the corresponding pixels is met. The generated graph structure is influenced by distribution of brightness in the input image. Then, the method executes an algorithm inspired by biological development which can form various structures by changing its parameters values on the generated graph to newly determine the brightness of each pixel corresponding to a node in the graph. Next, the method uses an evolutionary algorithm to optimize parameters of the algorithm inspired by biological development to make the newly created image close to the input image. Experimental results demonstrate that the presented method can create various images similar to an input image depending on how to form the graph and design a fitness function of evolutionary algorithm.