基于图像质量评价的显示器颜色复现技术研究

	基于图像质量评价的显示器颜色复现技术研究
	宋超
学位类型	博士
导师	王瑞光
	2014-07
学位授予单位	中国科学院大学
学位专业	电路与系统
摘要	随着社会工业水平和信息化程度的不断提高，彩色显示器逐渐进入生活和生产而成为人们获得信息的主要载体。图像颜色准确复现一直是广大消费者和各显示器厂商最关心的问题之一。然而现今流行的显示器如CRT、LCD、OLED、LED显示屏等，由于显示颜色的机理不尽相同，均存在不同程度的图像跨媒体颜色复现问题。本文对显示器颜色复现流程中的若干技术进行了深入研究，提出一种基于图像质量评价的显示器颜色复现技术。显示器的色度特征化和色域边界描述。针对LCD、CRT显示器和LED显示屏等显示设备的色度特性化，分别给出了特征化数学模型，然后利用PR655色度计对一台LCD显示器和一台LED显示屏箱体在暗室条件下分别进行特征化实验验证，并给出相应的特性化数据和曲线拟合结果。然后提出了一种显示器色域边界快速迭代计算方法，详细介绍了算法原理并结合LED显示屏分析了算法的计算精度和速度。使用色度特征化后的点间距为3mm的LED显示屏箱体对该色域边界描述算法进行了验证。实验结果表明：11次迭代运算后，拟合色域边界已非常光滑，每条映射线上的色域边界点真实值和计算值最大色差仅为0.23。与插值类计算方法相比，最大色差值和计算速度分别降低和提高了一个数量级。该色域边界描述方法具有广泛的通用性，可用于CRT显示器和LCD显示器等可用特征矩阵表征颜色三刺激值与三原色归一化亮度关系的显示设备。色域映射失真图像的质量评价。本文针对色域映射失真图像，在传统图像质量评价的基础上提出了一种新的图像质量评价方法。首先将源图像和映射图像转换到CIE LAB均匀颜色空间中。对图像的明度分量，提取其相位一致性特征和边缘强度特征，对图像彩度和色相分量，根据色域映射的特点，分别提取彩度和色相分量的彩度差和色相差作为图像的另外两个特征，形成最终的质量评价公式GMSIM。使用已有的3个色域映射图像集和TID2008图像库，对GMSIM算法进行参数拟合以及算法对两种图像库主观评价结果的预测准确性实验，其中对色域映射图像集采用命中率和显著性检验来表征算法性能。实验结果显示，GMSIM算法对色域映射失真图像的主观感知预测准确性比传统的图像质量评价方法提高了约4.5%，但仍有较大的进步空间，而在TID2008图像库上的预测性能也有所提高。该图像质量评价算法可以用于设计新的色域映射算法，也可在一定程度上代替复杂费时的心理物理学主观评价实验，来验证或对比色域映射算法的效果。基于图像质量评价的显示器色域映射。首先介绍了“基于图像特性选择算法”类色域映射算法的各种思路并引出本文色域映射算法思路。然后详细分析了各类色域映射算法的特点并从中选用了4种算法作为基本色域映射算法。提出了一种基于GMSIM图像质量评价算法和SLIC超像素图像分割与权重图融合策略的色域映射算法FUSION-I，在每个图像分块内以图像质量评价结果为选择标准，对已有多种不同的色域映射算法进行选择综合利用。针对FUSION-I算法容易产生色块的缺点，提出了对权重图进行引导滤波的色域映射算法FUSION-II。对本文提出的两种算法和其他基本色域映射算法进行主观评价实验，结果证明本文算法在图像颜色保真效果上要优于其他算法，但在计算速度上仍有待优化。
其他摘要	With the improvement of industrialization and informationization level of the society, color display panels gradually come into life and production, and become main carriers from which people get information. Correct image color reproduction has been one of the key problems that consumers and display panel manufacturers care about. Nowadays display panels like CRT, LCD, OLED and LED display panel et al are popular, and yet because of their different color display mechanisms, cross-media image color reproduction problems exist to varying degrees. In this paper, much attention was paid on techniques in the display color reproduction workflow, and a new color reproduction algorithm based on image quality assessment was proposed. Display panel color characterization and color gamut boundary description. For the color characterization of LCD、CRT and LED display panels, characterization mathematical models were provided separately, and characterization experiments were carried out for a LCD and LED display panel with a PR655 colorimeter, and corresponding characterization data and fitted curve were provided. After that, a fast and iterative gamut boundary calculation method for display panels was proposed. The method principle was introduced in detail, and the calculation accuracy and speed was analyzed with LED display panel as an example. A color-characterized LED display panel cabinet whose pixel pitch was 3mm was applied to verify the proposed gamut boundary description algorithm. Experiment result shows that the fitting boundary is very smooth after 11 iterations, and the maximum color difference between the true values and calculated values of gamut boundary points on every mapping line is only 0.23. Compared with the interpolation methods, the maximum color difference is decreased and calculation speed is increased by an order of magnitude respectively. The algorithm could be extensively applied to other display panels like CRT and LCD whose color tri-stimulus and normalized lightness couled be transformed by a characterization matrix。 Quality assessment method for gamut mapping distortion images was investigated. For gamut mapping distortion image, a new image quality assessment method was proposed based on traditional algorithms. Firstly, both original and reproduced image are transformed into CIE LAB color space. The phase congruency and edge strength characteristics of lightness component were extracted, and the chroma and hue difference characteristics were extracted for chroma and hue component, and final assessment formula named GMSIM was created based on these features. And then, being applied in 3 existed gamut mapping image sets and the TID2008 image library, the parameters of the GMSIM algorithm were fitted, and the prediction accuracy of the algorithm was analyzed. For gamut mapping image sets, hit rate was used to indicate the algorithm performance. Experiment result shows that, compared with traditional method, the proposed GMSIM algorithm prediction accuracy of subjective perceptions for gamut mapping distortion images was improved by about 4.5%, although there is still much room for improvement. The prediction performance for the TID2008 image library is improved to some extent. The proposed method could be used to design new gamut mapping algorithm, and to a certain degree, it could replace the complex and time consuming subject evaluation experiment to verify or compare the effect of different gamut mapping algorithms. Display gamut mapping based on image quality assessment. The ‘Choosing algorithms based on image characteristics’ type gamut mapping algorithms were introduced in detail firstly, and then the supposed gamut mapping idea was given. After that, the characteristics of each sort of gamut mapping algorithms were analyzed and 4 ones were chosen as the basic algorithms. A gamut mapping algorithm named FUSION-I was firstly provided based on GMSIM image quality assessment method and SLIC superpixel image segmentation applying the weight map fusion idea, and with image quality assessment results in each image segments as selection criteria, different sorts of gamut mapping algorithms were selected and comprehensively utilized. Color blocks could be generated using the FUSION-I method, and thus FUSION-II was provided based on the guided filtering of weight maps. Subjective evaluation experiments have been carried out for the provided 2 algorithms and other 4 basic gamut mapping ones, and experiment result shows that the proposed algorithms were better than others in keeping image color fidelity, and yet the process speed could be improved in the future.
语种	中文
文献类型	学位论文
条目标识符	http://ir.ciomp.ac.cn/handle/181722/41464
专题	中科院长春光机所知识产出
推荐引用方式 GB/T 7714	宋超. 基于图像质量评价的显示器颜色复现技术研究[D]. 中国科学院大学,2014.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
宋超.pdf（3852KB）			开放获取	CC BY-NC-SA	请求全文