Research On Adaptive Color Constant Algorithm

In color management, to achieve consistent reproduction of image in different equipment environment, we must consider the effects of environment illumination, and overcome the image offset originated by environment light. But there is no algorithms can be considered as universal at present, it is needed to choose different algorithms according to different scenes.The thesis mainly studies the color constancy algorithm combining of physical algorithm and statistical algorithm. First, based on the physical model, it studys the method of analyzing the characteristics of low level image using dichromatic reflection model. Some improvements are made in view of its application limitation. Then it put forward a new specular calculation method based on chromaticity. Second, based on the statistical method, the paper states the method of using support vector regression to analyze the image color information. Then the weighted support vector regression method is proposed due to its weak capability of noise-resistance. Last, the two methods are combined. Then the adaptive color constant algorithm is put forward. The test results show that the proposed algorithm improves greatly in the precision and the adaptability compared to the traditional algorithm.