COLOR TRANSFER BASED ON HIERARCHICAL IMAGE-REGION MATCHING WITH INTERACTIVE EVOLUTIONARY COMPUTATION

There have been some methods about applying coloration of painterly art to a photographic image. Most of the proposed method use some absolute criteria, e.g., by a certain mathematical formulas, for example the nearest distance in CIE L*a*b* color space, to convert a color to another uniquely. However, we want to develop a method that can convert color depending on the feelings and sensitivities of users themselves, like a painter makes a picture in the real world. Our algorithm can produce a color-transformed image resulting from two input images, which are a target photograph and a reference painting. Our method based. on interactive evolutionary computation (IEC) and can produce a variety of color-transformed images. The user interactively selects some candidates from those varieties to get the final result which looks similar to reference paintings. In this paper we exploit image-region matching for color transfer. Our experiments illustrate some color-transferred images and searching efficiency for our algorithm.