Color Conceptualization Using Optimization

The rapid development of the digital network and multimedia leads to the largerand larger proportion of the digital image and video in medium. Image enhancementand processing methods emerge in endlessly. Color transfer is a common imageprocessing technique, which transfers the color of one image to another by deliveringthe color feature.A lot of classical methods on color transfer and colorization are reviewed andsummarized in this paper. The origin and principle of color conceptualization is alsointroduced. Color conceptualization is one kind of the color transfer methods and itgenerates prototypes of color distributions from an input library images, and labelseach distribution with a verbal term such as―warm‖and―cold‖. Then for each inputimage, it changes its mood by changing its hue model to a prototype. The existingimage color conceptualization method has discontinuity problem in some cases due tothe lack of spatial information. In this paper, to solve this problem, we present a novelmethod to force neighboring pixels with similar intensities to have similar color.Using this constraint, the color conceptualization is formalized as an optimizationproblem with a quadratic cost function. Moreover, we further expandtwo-dimensional (still image) color conceptualization to three-dimensional (video),and use the information of neighboring pixels in both space and time to improve theconsistency between neighboring frames. The performance of our proposed method isdemonstrated for a variety of images and video sequences.