Homogenous Color Transfer Using Texture Retrieval and Matching.

Chang Xing,Hai Ye,Tao Yu,Zhong Zhou

DOI: https://doi.org/10.1007/978-3-319-48896-7_16

2016-01-01

Abstract:Color transfer is a simple but effective technique of realistic rendering. Most methods of color transfer select the source image manually, which makes an inconsistent transformation in semantic areas. We propose a novel approach of homogenous color transfer by using texture retrieval and matching. Several images are found out from a database using the texture features of a target image, then the image with the highest texture similarity is set as the source image. With a simple interaction of brush stroke in the target image, the texture features of the covered pixels are used to extract a homogenous region in the source image and match between such regions. Afterwards, an adaptive color transfer scheme is applied in the matched regions. Owing to the texture retrieval and matching, this method produces a consistent visual effect results. We demonstrate experiments in image colorization, style conversion and exposure adjustment to verify the characteristics.