基于像素相关性的灰度图像上色算法  

A New Algorithm for Colorizing Grayscale Image Based on Pixel Correlation

作　　者：孔德慧[1] 肖小芳[1] 徐振华[1] 郭荆玮[1] 

机构地区：[1]北京工业大学,计算机学院,多媒体与智能软件技术北京市重点实验室,北京100124

出　　处：《北京工业大学学报》2009年第5期708-714,共7页Journal of Beijing University of Technology

基　　金：北京市自然科学基金资助项目(4061001).

摘　　要：考虑到一幅图像的空间邻域像素在亮度及色彩上存在着极大的相关性,设计了一种新的搜索方法对灰度图像进行彩色化.首先在灰度图像中按照空间位置恒定步长递增的方式选取部分像素,并采取全局最优搜索参考样本对其上色,剩余像素的上色是通过局部最优搜索其邻域像素的方式实现的.本算法在得到较好图像上色品质的前提下,极大地加速了灰度图像的彩色化过程,保证了上色后图像颜色在空间上的连续性.为了避免通道耦合,灰度图像和参考彩色图像间的颜色传递在各通道相关性极小的lαβ颜色空间进行.Considering tremendous relativity on the luminance and chromatic channels for pixels between image neighborhoods,we proposed a new search method for grayscale image colorization.First,the algorithm selected some pixels by constant step incremental way spatially,and then colorized them using global optimization search strategy that the best matching color sample was selected from the color image.Next,the algorithm finished the colorization procedure for the other uncolored pixels using local optimization search strategy from the neighborhoods pixels.The experiments demonstrate that the new search strategy improves the speed of colorizing obviously,keeps the continuity of the colored image in color space and does not damage the colorizing quality.In order to avoid channel-cross,the colorization procedure for grayscale image was done in laβcolor space,which was developed to minimize correlation among three coordinate axes in color space.

关 键 词：图像处理 上色 全局最优搜索 局部最优搜索 像素相关性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]