CORC  > 清华大学
基于图切分的交互式图像染色算法
贾云涛 ; 胡事民 ; JIA Yun-Tao ; HU Shi-Min
2010-06-09 ; 2010-06-09
关键词彩色化 图切分 全局能量最小化 交互式 colorization graph cut global energy minimization interactive TP391.41
其他题名Interactive Graph Cut Colorization
中文摘要将黑白图像颜色化是一个需要大量用户交互和时间的事情,传统的基于交互的做法主要有两种:先分割后着色和全局优化.前者速度快,但是往往因为分块太大而丧失细节;后者能保持颜色变化的连续性,但是求解速度慢.对此文中提出了一种图像颜色化的新方法:基于图切分技术的分割算法.图切分是一种基于全局能量优化的分割技术,因而可以保证大部分区域的颜色分布光滑,而只在灰度变化剧烈的边缘地区产生颜色跳变,并且算法具有很快的求解速度.在用户简单给定颜色种子基础上,基于相同灰度倾向于相同彩色的基本假设,首先计算图像每个像素周围的灰度分布并构造一个全局的能量函数.接着利用图切分(graph cut)的技术快速有效地求得最佳分割.随着用户的进一步交互,图切分可以很快地迭代计算.这样,通过简单的交互,用户可以很快地对一幅黑白图像彩色化,并获得自然的效果.; Traditional ways to handle monochrome image colorization which always requires considerable user interaction and a lot of time are Segmentation colorization and Colorization using optimization.The former works fast,but always lose the details because of the large segmentation;while the latter looks much more continuous but takes longer time.This paper proposeds a novel approach: Segmentation colorization based on Graph cut,which is a very fast segmentation technique of global energy optimization.So it can maintain smoothness almost everywhere except for the sharp discontinuity at the boundaries in the image.Firstly,with the few seeds of pixels set manually by the user,a global energy is set up according to the gray value distribution around each pixel,with the conception that similar gray intensity prefers same color.Secondly,using 'Graph Cut',the best segmentation is got fast and efficiently.As user specifies more colors,the energy minimization will be solved iteratively and much faster.So with few manual specifications,user can colorize a gray image in a very short time and get naturally looking results.; 国家自然科学基金(60225016,60321002)资助
语种中文 ; 中文
内容类型期刊论文
源URL[http://hdl.handle.net/123456789/55685]  
专题清华大学
推荐引用方式
GB/T 7714
贾云涛,胡事民,JIA Yun-Tao,等. 基于图切分的交互式图像染色算法[J],2010, 2010.
APA 贾云涛,胡事民,JIA Yun-Tao,&HU Shi-Min.(2010).基于图切分的交互式图像染色算法..
MLA 贾云涛,et al."基于图切分的交互式图像染色算法".(2010).
个性服务
查看访问统计
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。


©版权所有 ©2017 CSpace - Powered by CSpace