Authors: Bara'a Ali Attea, Amina Dahim Aboud
Addresses: Department of Computer Science, University of Baghdad, Baghdad, Iraq. ' Department of Computer Science, University of Baghdad, Baghdad, Iraq
Abstract: We introduce an Evolutionary Algorithm (EA) for |colourising| greyscale images by evolving a colour |square patches| transfer process between a source, colour image and a target, greyscale image. The presented EA attempts to automatically choose colour patches from a source image and add their chromatic information to the achromatic values of the target image patches. The proposed EA uses first- and second-order statistics for luminance and texture matching. The genes that can represent best matching over all EA parents are recombined in a single multi-sexual recombination to form a single offspring. Then, mutation forms all other offspring. The results demonstrate the potential of our simple EA in colourising a variety of images, provided that the source and target images semantically belong to the same class.
Keywords: evolutionary algorithms; EA; colour transfer; multi-sexual recombination; greyscale images; luminance; texture matching.
International Journal of Advanced Media and Communication, 2007 Vol.1 No.3, pp.203 - 210
Published online: 03 Jun 2007 *Full-text access for editors Access for subscribers Purchase this article Comment on this article