Color Correction Of Digital Image Based On Three-dimensional Look-up Table

Color is important information of images. Machine vision lacks of adaptability, intelligence and color constancy, so color data change on different devices. It is necessary to reproduce true color by adding color correction segment after machine vision. With the development of digitization and the application of image processing, color correction plays more and more important role, it has been one of the most important factors which influences the performance of system, such as printing and dyeing, television, industry controlling, computer aided design, image surveillance and so on.In practical application, the present color correction technology is difficult to achieve high precision, especially for image analysis and recognition system in which color information is very vital. Therefore, the further research of color correction is required to achieve higher precision color correction. The thesis researches color correction model of efficiency and precision, and its core algorithm is look-up table (LUT).This dissertation focuses on two important problems in color correction discussed above. One is the precision of color correction model; the other is how to select enough samples with reasonable distribution. The main research work can be stated as two parts as follows:1.Methods are proposed to improve the precision of LUT under the condition of small sampling:Under the condition of small sampling, it is difficulty for direct interpolation LUT to achieve higher precision. The multi-level LUT method is proposed, which subdivides cubes of LUT to realize higher precision.How to select LUT mapping function is discussed in the thesis. Firstly, the self-adaptation method is introduced into maximum entropy estimate, which selects the number of samples in view of concrete condition of samples in the neighborhood of the input. Secondly, according to photoelectric conversion properties, self-adaptive local nonlinear regression color correction is proposed based on self-adaptive local nonlinear regression. Its nonlinear mapping function well fits the complex mapping relation of color spaces.2.In view of the existed limits of sample selection, self-adaptive color sample selection method is proposed based on the color pre-correction with marker color block. In the method, samples are automatically selected by iterative self-organizing data analysis technology algorithm (ISODATA). Compared with the general samples (ColorChecker24), these samples are more representative of the characteristics of the image color gamut. In the thesis, a series of experiments are designed and carried out. Based on the analysis of these experiments results, the subjective evaluation and objective evaluation are given so as to lay a foundation for further research.The research work has promoted the technology of color correction. It is of great theoretical significance and practical value to the development of image processing.