基于色度空间梯度优化的图像着色算法  

Image Colorization Algorithm by Color Gradient Optimization

作　　者：陈昕[1] 聂栋栋[2] 

机构地区：[1]杭州电视台,浙江杭州310008 [2]燕山大学理学院,河北秦皇岛066004

出　　处：《电视技术》2011年第19期23-26,共4页Video Engineering

基　　金：河北省自然科学基金(A2011203053);秦皇岛市科技计划项目(201001A038)

摘　　要：介绍了一种简单的灰度图像着色算法,它通过引入优化约束方程,尽可能地使着色后的图像在相应的色度空间内,其色度值的全局变化最小。在实际应用中,用户只要预先指定某些像素点的颜色,利用该算法就可以达到较好的主观着色效果。峰值信噪比的分析结果显示,在仅仅指定0.1%像素的颜色时,利用该算法着色后图像的平均峰值信噪比已经在26 dB以上。而且与先前的着色算法比较,该算法在使用更低的采样率下,着色后图像有更高的峰值信噪比,也即该算法优化估计的结果与原始图像的误差更小。A simple colorization algorithm for grey- scale images is introduced in this paper, which proposes an optimization equation to minimize the global color variance in the corresponding color space. Using the proposed algorithm, users only need to indicate the desired colors of a few pixels in the grey - scale image, and then a desirable color image can be easily produced. By computing PSNR of the colorized image, it is shown that the average PSNR of the images colorized by the proposed algorithm is over 26 dB when using only 0.1% of whole pixels as references. Compared with the colorization algorithm before, the images colorized by the proposed algorithm have higher PSNR by using fewer sample pixels, so the optimization error is less.

关 键 词：图像着色 梯度 优化 伪彩色 色彩空间 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]