A Research Of Color Space Conversion Model Based On RBFN

In modern printing workflow, the colorimetric characterization of devices in every link, even the same model is different. This would lead to color information difficult to achieve accurate cross-media transmission and reproduction, and seriously affect the accuracy of color printing. In order to solve this problem, color management came into being. Through the use of Device-independent color space, the color spaces of each device will connected with each other, and achieve accurate color transmission and reproduction.This paper mainly study color space conversion which are one of the key technologies in color management. We select the RBF neural network set up CMYK-Lab color space conversion model. First of all, this paper calibrates the output device in order to obtain good training data, and prints the test target after the status of equipment meet the experimental conditions, then measure and get sample data. Second, we study the structure of RBF neural networks and mathematical models, and initial create RBF neural network by using k-means clustering algorithm. At last, we determine the number of hidden layer nodes and the distribution coefficient by comparison test at the basis of training data. Now the complete CMYK-Lab color space conversion model based on RBFN is finished.Comparative experiments show that RBFN's learning speed and generalization ability are better than BPN. In this paper, chromatic aberration performance of RBFN which are trained by 300 sample data is equivalent to the modified neugebauer equation and BPN which are trained by 700 sample data. If further increase the number of training data, RBFN's chromatic aberration performance will be better than the neugebauer equation.