图像颜色迁移技术的研究
Research on Color Transfer between Images
【中文摘要】 评价图像颜色迁移的研究成果主要体现在:(1)处理效果,处理后的结果图像颜色是否自然和谐、画面逼真,这是衡量颜色迁移最重要的指标;(2)处理效率,人工参与量的多少和计算机处理算法的复杂性是影响处理效率的两个方面。现有算法处理的图像景物区域交界处连接不够自然,需要进一步优化;虽然有的算法在某些技术上实现了自动化,但颜色迁移的过程中仍需要适当的用户交互,完全自动的颜色迁移算法有待于研究。相对于人工参与量,算法复杂度的降低能显著加速处理进程。本文研究工作正是围绕这两方面展开。本文提出了基于区域交界处的域值分割颜色迁移算法,利用边缘检测方法提取景物边缘交界处区域,进行域值分割,将区域交界处的像素严格分割到所属的区域中,使得属于同一区域的像素迁移同类颜色,最后将已迁移颜色的目标图像均值方差归一化。用户交互虽然在一定程度上能够提高颜色迁移的准确性,但其执行效率低,而且具有一定的主观性,自动选取样本的颜色迁移算法有效的解决了这些问题。本文将网格划分减法聚类应用到颜色迁移中,完全实现算法的自动化,聚类方法按照亮度值的大小依次聚类,从各个聚类域中提取出数据点密度较大的像素点集组成样本块,首先对每个样本块间进行颜色迁移,再完成样本块以外的其他像素的颜色迁移。本文算法既适用与彩色图像间颜色迁移,也适用于彩色图像到灰度图像的颜色迁移。最后,本文设计了图像颜色迁移软件系统,该系统分为用户交互颜色迁移界面和自动颜色迁移界面。系统易于实现,用户使用简单
【英文摘要】 The results of image Color transfer is mainly reflected in : (1) Effect. Whether the color is natural and the picture is vivid is two most important measurements; (2) Efficiency. Both artificial participation and algorithm complexity are two aspects affecting efficiency. Using existing algorithms to process image, the junction region is unnatural, which requires better algorithm. Although some algorithms can bring out partly automation, in processing, user interaction is still inevitable. Comparing with artificial participation, algorithm complexity can obviously accelerate the processing speed. In the paper, research begins with these two areas.The algorithm in this paper is based on territory value segmentation of the junction region. The algorithm makes use of edge detection to distill the region of the edge, and conducts domain segmentation, then comminutes the pixels of regional at the junction each region strictly, transfer the similar color to the same region. Finally, the target image mean-variance is normalizedAlthough user interaction is, to some extent, able to improve the accuracy of color transferring, the efficiency of their implementation is too low, and has certain subjectivity. Automatic color transfer effectively solves this issue. Carve up Roseau and subtractive clustering is applied in this paper. Completely automatic algorithm implementation is realized. In clustering, pixels are selected according to their brightness, and those with larger data density are cluster to sample block. First of colors in the same block are transferred first, others are transferred later. This algorithm can be used to transfer color not only between two colorful images, but also between a grey image and a colorful image.Finally, this paper designs an image color transfer software system. The system is divided into user interaction color transfer interface and automatic color transfer interface. And this system is a very user-friendly one.
【中文关键词】 颜色迁移; 图像聚类; 域值分割; 自动分割; 颜色迁移软件系统
【英文关键词】 Color Transferring; Image Clustering; Domain Segmentation; Automatic Segmentation; Color Transfer Software System
【毕业论文目录】
摘要 4-5
Abstract 5
1 绪论 10-17
1.1 研究背景 10-11
1.2 研究的目的和意义 11-12
1.3 颜色迁移技术的研究现状 12-15
1.3.1 全局图像颜色迁移的研究现状 13
1.3.2 用户交互颜色迁移的研究现状 13-14
1.3.3 自动颜色迁移的研究现状 14-15
1.4 课题研究的主要内容 15-17
2 图像颜色迁移概述 17-30
2.1 图像颜色迁移的概念 17-18
2.2 颜色空间 18-21
2.2.1 RGB 颜色空间 18-19
2.2.2 lαβ颜色空间 19-21
2.3 几中典型的颜色迁移算法 21-28
2.3.1 全局颜色迁移算法 21-22
2.3.2 用户交互颜色迁移算法 22-24
2.3.3 自动颜色迁移算法 24-28
2.4 各种颜色迁移算法的比较 28-29
2.5 本章小结 29-30
3 基于区域交界处的域值分割颜色迁移算法 30-39
3.1 区域交界处的域值分割颜色迁移算法概述 30
3.2 图像样本块上色 30-31
3.3 全局彩色化 31-35
3.3.1 目标图像初始分割 31-33
3.3.2 区域边缘交界处域值分割 33-34
3.3.3 全局颜色迁移 34-35
3.4 实验分析 35-38
3.5 本章小结 38-39
4 自动选取样本的颜色迁移算法 39-53
4.1 图像聚类分割 39-43
4.1.1 聚类分析 39-40
4.1.2 基于网格划分减法聚类 40-42
4.1.3 模糊C 均值聚类 42-43
4.2 自动选取样本块 43
4.3 颜色迁移 43-46
4.3.1 彩色图像到灰度图像的颜色迁移 44
4.3.2 彩色图像间颜色迁移 44-45
4.3.3 本文自动选取样本块颜色迁移算法过程 45-46
4.4 实验分析 46-51
4.4.1 本文聚类结果分析 46-47
4.4.2 颜色迁移结果分析 47-51
4.5 本章小结 51-53
5 图像颜色迁移系统实现 53-63
5.1 设计目标和开发平台 53
5.2 图像颜色迁移的总体设计 53-63
5.2.1 用户交互颜色迁移界面 55-61
5.2.2 自动颜色迁移界面 61-63
6 结束语 63-66
6.1 全文工作总结 63-64
6.2 今后研究工作展望 64-66
参考文献 66-71
致谢 72
