Research On Efficient Image Decolorization With A Multimodal Contrast-preserving Measure

Image decolorization is widely used in a variety of computer vision applications as a basic algorithm and has practical research significance.Image decolorization aims to convert a three-channel color image to a single-channel grayscale image.The process of image decolorization suffers from the loss of visual information inevitably.The key issue in the decolorization process is how to maximize the contrast information of the input color image for the output grayscale image.The grayscale can be obtained quickly and easily by extracting the lightness channel from the input color image.However,it cannot distinguish the contrast between the iso-luminant pixels containing different colors.In recent years,many researchers aim to study algorithms of mapping a color image to a grayscale one based on the contrast information.Based on the local and global color contrast,and the multimodal Gaussian distribution function,this paper proposes a new image decolorization algorithm with a contrast-preserving multimodal energy function.This proposed algorithm can effectively relax the loss of color contrast of weak order color pairs during the mapping process.Our decolorization algorithm mainly includes three key steps.Firstly,the dominant non-local color pair set is constructed by making full use of the advantage of the linear bounding volume hierarchy while the local color pair set is produced by removing duplicate instances of spatially-neighboring color pairs.Secondly,by using the linear parametric function and the contrast-preserving multimodal energy function,a final energy optimization function is established for the mapping of color image and grayscale image.Finally,based on the discrete searching scheme,a group of linear parameters can be obtained which have the smallest energy value from all parameter combinations.All pixel values of the output grayscale image is calculated using these linear parameters.Quantitative experimental comparisons show that the proposed algorithm is able to achieve a high score.As a consequence,the proposed algorithm can effectively preserve the color contrast of the original color image and thus improve the effect of image decolorization.