基于SHAME 的图像色差模型的扩展及研究

VIP免费

3.0 刘畅 2024-11-07 22 4 1.64MB 68 页 15积分

侵权投诉

摘要

随着信息数字化程度的不断加深，跨媒体传播不断融入社会生活中，于是保

证大众获取高质量的信息显得尤为重要。这一过程中，跨媒体图像复制质量成为

图像颜色科学研究领域的热点，研究的目标都是确保图像原稿在不同媒介之间都

能真实还原，色彩准确再现。因此，如何评价不同媒介之间的图像质量也随之成

为一个关键问题。本文针对当前存在的问题，以如何评价软拷贝和硬拷贝之间图

像色差为出发点，以能够预测人眼观察者对软拷贝与硬拷贝之间图像色差为目标，

对以下几点进行了研究：

(1) 分析图像色差模型的研究框架，研究各个模型之间的区别及其各自优缺点；经

过实验比较后选择将 SHAME 图像色差模型作为基础构建适用于计算软拷贝

与硬拷贝之间图像色差的图像色差模型。

(2) 研究探讨了混合色适应变换的两个关键技术：不完全色适应和混合色适应，分

析同时观察软拷贝与硬拷贝时的实际情况从而选择合适的混合色适应变换作

为研究基础。

(3) 综合对以上两点算法模型的研究，提出了适用于计算软拷贝与硬拷贝之间图像

色差的算法模型设计方案，同时提出了整体的应用流程。

(4) 利用 Matlab 仿真模拟本文提出的算法模型，同时设计实验验证了该图像色差

模型在计算软拷贝与硬拷贝之间图像色差方面的有效性，以及整个应用流程的

可行性。

本文的创新点在于：

1. 首次提出结合色适应变换模型的优化图像色差模型，从而能够将图像色差模型

应用于计算软拷贝与硬拷贝之间图像色差过程中。

2. 首次提出了基于跨媒体图像复制过程的图像质量评价应用流程，适用于计算软

拷贝与硬拷贝之间的图像质量评价指数。

关键词：图像色差 SHAME 图像色差模型色适应变换软拷贝与

硬拷贝

ABSTRACT

With the continuously deepening of information digitization, cross-media

communication is emerging into social life. So it turns out to be very important that the

public can obtain information of high quality. During this process, we must make sure

that the original images can be reproduced correctly and the color can be exactly right

across different media. As a result, the quality of cross-media image reproduction has

been a research issue in the field of image color science. At the same time, how to

evaluate the image quality across different media has also become a key point as well.

Based on these existing problems, starting from how to evaluate the image color

difference between softcopy and hardcopy and targeting to predict the observing results

obtained by observers, this paper do research about these following points:

1 Analyze the research framework about image color difference, compare the

advantage and disadvantage among different models, choose the SHAME model as a

basis to do further extension which can be applied to predict the image difference

between softcopy and hardcopy;

2 Study and discuss the two key technology involved in the mixed chromatic

adaptation: incomplete chromatic adaptation and mixed chromatic adaptation. Pick out a

suitable mixed chromatic adaptation as the basis after analyzing the condition when

comparing the softcopy and hardcopy simultaneously;

3 Propose a design plan integrating the two above algorithms and model together,

which can be applied to calculate the image color difference between softcopy and

hardcopy. And a whole application framework is also been proposed.

4 Finish the simulation of the proposed model by MATLAB and design experiment

to verify the validity of the model and the feasibility of the application framework

The innovations of this paper are shown as follows:

1 Propose to optimize the image color difference model based on chromatic

adaptation model for the first time, which then can be used to predict the image color

difference between softcopy and hardcopy

2 Propose an application framework aiming to evaluate the image quality based on

cross-media reproduction for the first time. The framework is suitable for calculating

the image quality index between softcopy and hardcopy.

Key Words: Image color difference, SHAME model, Chromatic

adaptation, Softcopy and hardcopy

中文摘要

ABSTRACT

第一章绪论 ··············································································································· 1

1.1 研究背景、目的及意义 ··················································································· 1

1.2 国内外发展现状 ······························································································· 1

1.3 论文的研究内容及创新点 ··············································································· 3

第二章图像色差模型的分析研究 ············································································ 5

2.1 *

E色差模型 ································································································ 7

2.2 S-CIELAB 模型 ································································································ 8

2.3 色相角算法 ······································································································ 9

2.4 SHAME 模型 ·································································································· 11

2.4.1 SHAME 模型 ··························································································· 11

2.4.2 SHAME-II 模型 ······················································································· 12

2.5 各模型性能的分析与比较 ············································································· 12

2.6 本章小结 ········································································································ 14

第三章混合色适应变换的机制分析 ······································································ 15

3.1 混合色适应变换的关键技术分析································································· 16

3.1.1 不完全色适应(Incomplete Chromatic Adaptation) ································· 17

3.1.2 混合色适应(Mixed Chromatic Adaptation) ············································ 17

3.2 主流混合色适应变换模型的实现································································· 18

3.2.1 RLAB 色适应变换模型（1991） ··························································· 18

3.2.2 CAT97 色适应变换模型（1998） ························································· 19

3.2.3 S-LMS 色适应模型(2001) ······································································ 20

3.2.4 CAT2000 色适应变换模型（2001） ······················································ 21

3.3 分析与比较 ···································································································· 23

3.3.1 不完全色适应方法的比较 ······································································ 23

3.3.2 混合色适应方法的比较 ·········································································· 24

3.3.3 混合色适应模型的确定 ·········································································· 25

3.4 本章小结 ········································································································ 27

第四章基于色适应变换的图像色差模型 ······························································ 28

4.1 基于色适应变换的图像色差模型································································· 28

4.2 针对计算软拷贝与硬拷贝之间图像质量的应用流程 ································· 29

4.2.1 图像配准技术 ························································································· 30

4.2.2 扫描仪特性文件的制作 ········································································· 36

4.2.3 软拷贝图像处理方式 ············································································· 37

4.3 本章小结 ········································································································ 38

第五章图像色差模型的评价································································································· 39

5.1 实验方法 ········································································································ 39

5.1.1 配对比较(Pair Comparison) ··································································· 39

5.1.2 类别判断(Category judgment) ································································ 40

5.1.3 排序比较(Rank order) ············································································ 41

5.2 数据分析方法 ································································································ 42

5.2.1 Z 分数及置信区间 ··············································································· 42

5.2.2 系数比较 ································································································ 43

5.3 验证实验 ········································································································ 44

5.3.1 实验设备及环境、实验方法 ·································································· 44

5.3.2 测试图片 ································································································· 45

5.3.3 实验过程及步骤 ······················································································ 47

5.4 结果及分析 ···································································································· 48

5.5 本章小结 ········································································································ 52

第六章总结与展望 ································································································· 53

6.1 研究总结 ········································································································ 53

6.2 问题及展望 ···································································································· 53

附录 ·························································································································· 54

参考文献 ·················································································································· 61

在读期间公开发表的论文和承担科研项目及取得成果 ········································ 65

致谢 ·························································································································· 66

第一章绪论

第一章绪论

1.1 研究背景、目的及意义

随着数字化时代的不断深入，印刷业的发展模式在不断变化，影响着整个印

刷业的发展趋势。印刷品质量的要求也随着印刷产业的发展不断提升，因此对印

刷品质量进行准确的评价及控制变得至关重要且必不可少。传统意义上的印刷产

品主要都是以半色调图像形式呈现，利用人眼的视觉特性，最终人们看到的仍然

是层次清晰、色彩变化连续的图像。但是随着跨媒体传播技术的蓬勃兴起，产品

的输出终端种类增加，显示器、手持终端设备都成为输出设备，因此保证在不同

输出终端显示相同质量的图文信息成为研究热点[1]。

因为缺少相应的物理量化评价指标，目前对跨媒体复制图像进行质量评价的

方法主要是通过人眼在对印刷图像进行评价。当同时显示两幅图像时，对于大多

数人来说很容易去判断哪幅图像具有更高的质量，但是这里所谓的更高质量与观

察者的个人喜好有很大关系，同时也会受到观察者当时心理状态的影响，主观依

赖性很强。但如果能够从客观角度评价图像质量，用客观评价结果代替主观评价

或者主客观角度相结合，结果就更合理更有说服力。所以对图像质量的客观评价

研究显得至关重要。在评价图像质量时，图像颜色是首先要考虑的因素。如果印

品与原稿之间的图像色差超过了可接受的范围，那么该印刷品的质量肯定是不合

格，不会被客户所接受。对于均匀色块的印刷品，可以使用分光光度计等测量设

备测量色块的色度值，再通过计算即可得到均匀色块之间的色差。但是因为目前

测量设备的局限性以及图像的复杂性，这种方法对于图像印刷品不可实现。从理

论上来说，图像色差不同于一般均匀色块色差是因为图像的色差不仅仅有观察条

件即外界的影响因素，如光源、背景色等，其本身的像素之间就会产生很多相互

影响，从而对整个图像色差产生影响[2,3]。所以对观察者来说，图像色差既可以认

为是明显的标准，也可以认为是模糊的概念。选择出更高质量的图像很简单很明

显，但解释“为什么做出这个选择”就不是那么容易了。同时通过观察者来评价图像

质量需要消耗很多时间及精力，而且因为无法合理量化“个人喜好”这个因素，所以

创建一个能够预测图像感知色差的数学模型显得十分必要，尤其是能够满足跨媒

体图像色彩再现的要求[4]。

1.2 国内外发展现状

跨媒体图像复制过程中不同输出终端的白点(white point)是不同的，如最常见

的显示器显示的软拷贝(softcopy)与打印稿硬拷贝(hardcopy)，硬拷贝的观察环境通

常是 D50 标准光源，但是显示器显示的软拷贝不仅受到显示器光源(如D65)的影

响，还会受到周围环境光源(如D50)的影响。所以为了构建能够准确预测跨媒体输

基于 SHAME 的图像色差模型的扩展及研究

出图像质量的算法模型，选择何种图像色差度量指标及色适应转换过程都是十分

重要的。

目前主要有两种建立图像色差模型的方法，即损坏（impairment）方法和质量

（quality）方法[5]。损坏方法是指比较一幅待评价图像与参考图像或理想图像之间

颜色差异程度的度量方法。而质量方法是指利用数学方法直接构建一幅图像的质

量评价模型，不需要借助参考图像。质量方法也可以被看作是：将一幅图像直接

和一些基本的理想心理表征进行比较而构建的图像质量模型[5,6]。

而在实际进行图像色差评价时，一般都会将再现图像与原稿图像作对比而后

得到评价结果，也就是上述的损坏法。传统计算图像色差的典型方法就是将

CIELAB，CIE94 和CIEDE2000 等色差公式应用到图像中，求出逐个像素的色差再

计算出平均值作为图像色差。但是这种方法是建立在同种色块的小色差基础上的，

预测图像色差时其本质上还是将一幅图像中的每个像素作为一个完全独立的色刺

激，没有考虑周围像素对其的影响，所以不适合用在预测图像色差。

为了解决这一问题，1998 年斯坦佛大学的 XueMei Zhang 提出了在进行色差计

算之前，先对图像进行滤波处理，去掉图像中人眼不能识别的部分，使得计算结

果更加贴合人眼的视觉效果，即 S-CIELAB 模型[7]。这一突破性的创新思维对后来

的图像色差模型研究起到了很深刻的影响。但是 S-CIELAB 模型的缺点是模拟人

眼视觉的模型精度不是很高，于是，很多科研人员又在 S-CIELAB 模型的基础上

做了优化及改善，比如 Fairchild 将S-CIELAB 模型中的空间域图像滤波改为频率

域图像滤波[8]，这样能够更精确地控制 CSF(Contrast Sensitive Function 对比度敏感

函数)的形状，从而保证图像色差的预测结果达到更高的精度。

2002 年，英国利兹大学的罗明等人提出了色相角算法来计算图像色差[9]，该

算法以色相为基准对各像素进行权重分配，因此比用色差公式直接计算图像色差

有了进一步的提高。但是因为该算法忽略了人眼对比度敏感的影响，所以预测精

度达不到特别好的效果。尽管如此，色相角算法的核心思想仍然意义重大，是后

续相关研究的基础。

Marius Pederson 和J Y Hardberge 在验证色相角算法的效果后，在其基础上将

人眼视觉滤波应用到该算法中，提出了 SHAME 图像色差模型[10]。SHAME 模型优

化了色相角算法在计算图像色差时没有解决的问题，提高了半色调图像色差的评

价精度。

从查阅文献的情况来看，国内对于图像色差方面的研究还比较少，云南师范

大学的石俊生团队以及北京理工大学颜色科学与工程国家专业实验室对这方面稍

有研究[6,8,11]，但是他们的研究都是从对图像色差模型的应用角度出发，并未对相

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

15 积分 4人已下载

立即下载 VIP免费下载

摘要：

摘要随着信息数字化程度的不断加深，跨媒体传播不断融入社会生活中，于是保证大众获取高质量的信息显得尤为重要。这一过程中，跨媒体图像复制质量成为图像颜色科学研究领域的热点，研究的目标都是确保图像原稿在不同媒介之间都能真实还原，色彩准确再现。因此，如何评价不同媒介之间的图像质量也随之成为一个关键问题。本文针对当前存在的问题，以如何评价软拷贝和硬拷贝之间图像色差为出发点，以能够预测人眼观察者对软拷贝与硬拷贝之间图像色差为目标，对以下几点进行了研究：(1)分析图像色差模型的研究框架，研究各个模型之间的区别及其各自优缺点；经过实验比较后选择将SHAME图像色差模型作为基础构建适用于计算软拷贝与硬拷贝之间图...

展开>> 收起<<

基于SHAME 的图像色差模型的扩展及研究.pdf

共68页,预览7页

还剩页未读，继续阅读

基于SHAME 的图像色差模型的扩展及研究

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

推荐作者

热门标签

举报选择: