Interactive Colorization via Multi-Cues Manipulation

Colorization plays an important role in image processing, which aims to change the original image color according to specific image color characteristics. In this article, the authors present a regional color editing method based on multi-cues manipulation, including interactive segmentation, inpainting and gradient-preserving optimization. Firstly, the user draws strokes specifying the target region which needs to change color, through which the target image can be segmented by the K-means color clustering method. Then the example-based inpainting technology is applied to achieve natural transition along the boundary. After color propagation, they apply an optimization algorithm to preserve the gradient. The experiment results demonstrate that the proposed approach can not only achieve a visual satisfactory local color propagation results, but also preserve the texture details well.