一种改进的图像色彩迁移方法

Improved Method of Image Color Transfer

传统色彩迁移方法的迁移效果依赖于主观评价,且应用局限性大。为此,提出一种改进的图像色彩迁移方法。通过亮度拟合、分类筛选等操作从多幅源图像中选出与目标图像最匹配的参考色彩区域,针对彩色图像,采用隶属度概率进行平滑处理,利用参考色彩区域的颜色分量均值和方差,计算迁移合成值。针对灰度图像,对颜色分量进行归一化处理,划分灰度等级,引入Sigmoid加权实现迁移合成。实验结果表明,改进方法在视觉效果和客观评价方面都有较好的表现。

The color migration effect of traditional methods depends on the subjective evaluation, and has a big application limitations. This paper proposes an improved method of image color transfer. It chooses the best matched color of the reference color area from the multiple source images by the brightness fitting, category selection. For the color images, the various regions of the target image are smoothed by using the probability of membership function, and the synthetic value of color transfer is computed by using the mean and variance of the reference color area; for the grayscale images, the components of reference and source color image are normalized, the gray level is divided, and the membership is weighted by the Sigmoid function, to achieve the color migration of gray images. Experimental results show that the improved method can generate convin~：ing results with well performance in terms of objective evaluation or visual effects.