Color Transfer Algorithm In Non-photorealistic Rendering

Non-photorealistic rendering is a new problem in computer graphics and image processing. It is an effective combination of computer science and art. It emphasizes on people's subjective feelings, designing to give people the joy of beauty. Color transfer is one of the key technology for non-photorealistic rendering. Effective color transfer algorithm will excellently enhance the artistic effect of non-photorealistic rendering. There is no unified criterion to evaluate the effect of color transfer. Therefore, studies on the color transfer algorithm and evaluation has a great signification for us.Based on the existing achievements, the main research contents of this paper consists of the following four parts:(1) Color transferring is carried out in the l?? space. The images are firstly transformed from RGB space into l?? space with orthogonal. We explain the orthogonal of the l?? space mathematically. Finally we introduce the color space and its conversion method.(2) We give deformations of the classical color transfer algorithm and the color transfer algorithm based on analog transformation, finally we obtain the color transfer states under five kinds of linear transformations and four kinds of nonlinear transformations, wide diversity of color transfer altering.(3) Study on color transfer algorithm based on fuzzy clustering, the experimental results show that this algorithm can realize the adaptive color transfer between images. It can solve the problem that the traditional color transfer algorithm can not transfer successfully between poor similarity color images.(4) In the full reference to the basis of existing achievements, we summarize four indexes which can be used as the evaluation criterion for the effect of color transfer: color transfer efficiency, effect of art perception, the intention of the expected by the graphic, the possibility of offering new products and service. After that, we propose an evaluation model for the image color transfer based on TOPSIS method and establish an actual issues model on this method. The experimental results show that the model can get a reasonable evaluation results.