针对具有多种颜色区域图像以及灰度图像的颜色迁移合成问题，提出了一种基于高斯混合模型的区域颜色迁移算法，以用于将一幅源图像的颜色迁移到一幅目标图像中。该算法首先在lαβ颜色空间中，采用高斯混合模型对源图像的颜色样本空间进行聚类拟合来得到具有源图像颜色信息的参考色彩模型；然后利用该参考色彩模型对目标图像的颜色样本集进行分类决策，以确定目标图像与源图像各区域的色彩匹配关系；最后，提出了一组新的基于区域的颜色迁移公式，用于完成对目标图像的颜色迁移。实验结果表明，该算法不仅能实现两幅彩色图像之间的颜色迁移．而且能实现对灰度图像的上色处理，并均取得了较好的迁移效果。

This paper addresses the problem of regional color transfer from source image to object image based on Gaussian mixture model(GMM). Instead of choosing RGB color space, images are processed in lαβcolor space, which minimizes correlation between the channels. Authors firstpick up color information of the source image to construct a reference color model by adopting GMM, then classify the color samples in object image according to the model and determine the color match between object image and source image. At last, new color transfer formulas are proposed for regional color transfer. Experimental results demonstrate that the proposed algorithm works very well not only on transferring color between two color images but also on colorizing gray image by color transfer from a color image.