A Color Transfer Algorithm Based on Neighborhood Correlation and Optimization Techniques

The traditional Welsh´s grayscale image colorization method for color transferring is based on pixel´s luminance information only. The method generally cannot ensure the continuity of image transferring effect in space. This paper presents a color transfer algorithm based on neighborhood correlation and optimization techniques, which not only considers neighborhood´s luminance information but also makes full use of colored pixel´s color information. The algorithm automatically produces the initial colorized image via a color transfer method, removes the color information of pixel that is completely different from neighborhood pixels to form the regions of color seed and then use some optimization techniques to complete the colorization process by diffusing the distribution of color seed to the corresponding regions of image. Experimental results show that the new algorithm improves the drawback of the Welsh´s algorithm, and the color of the result image is more natural and delicate.