IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 32（9）,2010
In order to tackle the limitation of conventional color-to-gray transformation, a new concept of visual cue is presented based on human visual perception priors. Then based on the Bayesian Theorem, a probabilistic graphical model is constructed to retarget the extracted visual cues to the synthesized grayscale image, which is deduced to be a MAP problem and solved by using partial difference equation (PDE). The experimental results demonstrate that the proposed approach preserves more visual details than the existing technologies. In addition, the proposed technique can be widely employed in image understanding, publishing/printing, and the barrier-free access to visual information of the colorblind people.
