Color transfer via local binary patterns mapping

Abstract

Color transfer is a process of carrying over image colors from one image to another. Since images have diverse texture, color, content and other features, key challenge for color transfer is to find a correct mapping between image and target image. In this paper, a new color transfer method based on feature points extraction and local binary patterns(LBP) mapping is proposed. We construct a framework for feature points extraction and mapping. In the framework, we use SIFT descriptor for feature points selection and we design a LBP mapping method for feature points matching between source and target image. Final feature points in source image are obtained using this mapping. Greyscale source image and final feature points are used as inputs of a scribble-based colorization method to get the final color transfer result. Experimental results show the benefits of our method. © 2010 IEEE.