Research On Some Color Transfering Problems

Altering an image's color is one of the most common tasks in image processing, and Reinhard's color transfer supports such a good way. It describes a method for a more general form of color correction that borrows one image's color characteristics from another.Based on Reinhard's classical methods,this dissertation will give a much deeper research on the methods of altering an image's color such as Wide Diversity of Color Altering, and a new notation called color analogy.All of the algorithm are operated in 1αβcolor space which minimizes correlation between channels for many natural scenes. The chapter Wide Diversity of color altering gives linear and nonlinear transformations for altering an image's color. We can get target image data's maximum, minimum, mean value and standard deviation, and we can get a lot of transformations with these variables. All these transformations are based on Reinhard's classical formulas. As a result, it gives birth to a lot of methods for an image's color altering.Hertsmann's image anlogies describes a new framework for processing images by example, a process of learning style. The chapter Color Analogy and its Application,given three images A, B and C, we can achieve color correction by choosing appropriate source images A and B, and apply its color learning process to the third image C. Our goal is to do so with a simple algorithm, which operated in 1αβcolor space. Experiments show that the colorizing results are perceptual satisfaction. After this, we give an ameliorating algorithm for color analogy. Considering about the correlation between the third image and the first image,we add a correlating coefficent when computing the needed mean value. We get this coefficent through the ratio by the third image and the first image's standard deviation. Consequently, this method will be more scientific and the effect will be much more natural and true.