基于直方图映射和分层的图像迁移算法研究  

Colorization of Gray Images Based on Histogram Mapping and Image Layers

作　　者：李宏[1] 向遥[1] 张卫[2] 胡可成[1] 

机构地区：[1]中南大学信息科学与工程学院,湖南长沙410083 [2]湖南大学建筑学院,湖南长沙410082

出　　处：《小型微型计算机系统》2007年第6期1110-1114,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60433020)资助

摘　　要：本文提出了一种基于直方图映射和分层迁移的灰度图像上色HisT算法.对于给定的颜色图像和需要迁移的灰度形状图像,先从RGB色彩空间转换到一个正交空间lαβ,消除空间各分量之间的强相关性.然后分别统计两幅图像的灰度直方图,对于灰度图像中的象素点,保留其灰度值,而利用颜色图像中与之对应或相邻灰度等级的颜色信息为灰度图像进行粗糙上色,得到一幅过渡图像.参照灰度直方图,将颜色图像和过渡图像进行分层处理,并利用统计学方法对图像各层和整体统计量进行分析,提取颜色图像的颜色信息和过渡图像的形状信息,最终实现颜色图像到灰度形状图像的上色过程.实验证明,该算法能快速准确地实现灰度图像的上色,具有广泛的应用前景.We propose a fast algorithm named HisT for colorizing gray images by transferring colors from a source, color image to a destination, gray image. Instead of choosing RGB color space, we process images in 1αβ color space, which minimizes correlation between channels. As gray image varies only along luminance, colors based on the luminance histograms of the two images are transferred as follows. In the first step, we match luminance histogram scales between images, and transfer the mean of chromatic information from the source to the target, where we get the transition image. Secondly, we split the color image and transition image into layers, and impose improved color transfer equations onto each pixel in the transition image, which is used for color transfer between corresponding layers from color image to transition image. The experimental results prove that the HisT algorithm proposed herein is a fast algorithm with little human interactivity.

关 键 词：颜色迁移 图像处理 灰度直方图 图像分层 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]