【作者】 徐国梁；

【导师】 谭庆平；

【作者基本信息】 国防科学技术大学 ， 计算机科学与技术， 2010， 博士

【摘要】 数字图像在输出前必须转换成与输出设备深度相同的图像。图像半色调化是一种将深度较高的图像转换成深度较低的图像,并且使两者在视觉上尽可能相似的技术。图像半色调化技术在涉及图像输出的很多领域中有广泛的用途,如印刷、户外广告制作、电视、电子游戏、动画、联机的图像显示和网络传输等领域,它是决定输出图像质量的关键技术。图像半色调化技术对研究输出图像的质量控制和优化、输出图像的颜色预测、针对不同输出设备的半色调化图像质量评价具有重要意义。在实际应用中,图像输出设备存在局限、图像输出设备的多元化、输出图像质量评价中主观因素给图像半色调化技术研究和应用带来了挑战。本文以彩色印刷为应用背景,以提高输出图像质量为目标,以图像半色调化及相关技术为研究内容,展开了从基于几何分割的半色调化抖动矩阵设计方法,到混合半色调化的色彩偏移问题,再到半色调化图像质量评价体系的系列深入研究。主要研究成果包括:(1)针对现有基于Delaunay三角化几何分割设计的混合半色调化抖动矩阵存在阴阳集束网点密度相差悬殊的缺陷,本文提出了通过优化几何分割减少该缺陷方法。该缺陷在实际应用中将妨碍集束网点密度的提高,而提高集束网点密度是提高输出图像质量的关键因素。本研究从欧拉公式推导出了两条关于几何分割的性质,其中关于全三角形几何分割的性质可以准确地解释造成现有研究成果中阴阳集束网点密度相差悬殊的原因,然后以全四边形几何分割的性质为依据,提出了优化几何分割的算法。实验中该算法将顶点数与子区域数的比例从优化前的0.5改进到了0.90以上。在该优化的几何分割的基础上设计混合半色调化抖动矩阵时,其阳集束网点和阴集束网点密度之比也相应地改进到了0.90以上。(2)提出了通过扩张初始集束网点中心点集构造阴阳集束网点密度相等的混合半色调化抖动矩阵的方法。它可以在满足印刷工艺限制的条件下进一步提高混合半色调化图像中的集束网点密度,从而提高印刷质量。为了验证扩张中心点集方法的合理性,实验中将扩张前后的中心点分布转换成半色调化图像以比较它们的均匀性。除此之外,还将全四边形化几何分割的四边形与优化的几何分割中四边形的质量因子做比较。实验结果表明:扩张前后中心点分布对应的半色调化图像的峰值信噪比(PSNR)接近,而且全四边形化几何分割中四边形质量因子与优化的几何分割中四边形质量因子相比也有所改进。(3)提出了一种用半色调化图像错位叠加仿真套印误差,并且用Yule-Nielsen Neugebauer颜色方程和彩色胶印颜色标准ISO12647计算色彩偏移量的方法。实验中将本论文研究设计的混合半色调化抖动矩阵的色彩偏移量与调幅半色调化的色彩偏移量比较后得到如下结论:用三个不同的随机数种子生成的混合半色调化抖动矩阵的色彩偏移量与调幅半色调化的色彩偏移量接近,色彩均匀性也接近。因此,用本论文方法设计的混合半色调化抖动矩阵,其色彩偏移量完全可以满足印刷工艺的要求。(4)提出了一个统一的针对不同图像输出设备的基于非理想打印机模型的半色调化图像质量评价体系。本论文研究用数学形态学的膨涨运算(Dilate)仿真印刷工艺过程的网点扩大现象,并将膨涨运算参数与胶印网点扩大曲线拟合,加上本文提出的小网点过滤算法,建立了面向胶印工艺过程的非理想打印机模型。结合人类视觉系统(HVS)模型计算人眼感知的图像,实验检验了基于非理想打印机模型的的半色调化图像质量评价方法。本论文研究的实验表明,用该评价方法比基于理想模型的评价方法得到的结果更接近胶印生产的实际。更多还原

【Abstract】 A digital image should be converted to the image of the same depth as the depth of image output device. To halftone an image is to convert a high depth image into a low depth image, so as to make the low depth image visually similar to the original one. Halftone technology can be applied extensively in many fields related to image output, such as printing, out-door advertisement billboard, television, video game, animation and on-line image display and transfer etc, where halftone is the key technology to determine quality of output image. Halftone technology is of great significance to the control and optimization of output image quality, output image color predicion and image quality evaluating methods applied to different output devices. .In the practice of halftoning image, human factors in output image quality evaluation, performance limitation and versatility of image output devices pose great challenges to the research and application of halftone technology. Aiming at improving output image quality, and taking color printing as application background and image halftone and its related technology as main topics, this thesis carries out a series of in-depth research, including the design method of hybrid dithering matrix based on geometry subdivision, color shift of hybrid halftone image overlay and halftone image quality evaluation system.The main contributions of this paper can be summarized as follows:(1) develops an optimized method for the geometry subdivision in order to reduce the great disparity between highlight and shadow clustered dot density of the hybrid dithering matrix based on the geometry subdivision. The unbalance of clustered dot density will limit the increase of the clustered dot density, while clustered density is crucial to high quality image print. Two theorems are deducted from Euler formula in the dissertation. Theorem of all-triangle geometry subdivision can be used to explain why big difference between highlight and shadow clustered dot density is made in the existing research. Based on theorem of all-bilateral geometry subdivision, an algorithm is developed to optimize the geometry subdivision. In the experiment of the optimization algorithm, the ratio between vertex and faces of the subdivision is improved to above 0.9 from 0.5. In the hybrid dithering matrix designed with the optimized subdivision, the ratio between highlight and shadow clustered density is also improved to above 0.9 from 0.5.(2) proposes a method to construct dithering matrix with equal clustered dot density by expanding the set of clustered dot central points and forming an all-bilateral subdivision with the clustered dot centers as the vertex. With dithering matrix of equal clustered dot density, higher clustered dot density and higher image print quality can be achieved. To justify the method of expanding the set of clustered dot central points, an experiment is designed to compare the evenness of the images of expanded central points and unexpanded one. In addition, a comparison is made between the quality factor of bilateral of subdivisions based on expanded central points and the original one. The result of the experiment shows: the PSNR of the images of expanded central points and original one is approximate equal.(3) presents a method to calculate color shift by simulating misalignment of color separations with halftone image overlay, Yule-Nielsen Neugebauer color equation and offset printing color standard ISO 12647. By comparing the color shift of amplitude modulated halftone and hybrid halftone developed in this research, the simulation experiment concludes that: color shift of the hybrid dithering matrix designed with three different random seeds can meet the requirement of printing procedure.(4) finally, establishes a unified halftone image quality evaluation approach based on non-ideal printer model by learning from the existing halftone image quality evaluation model based on the ideal printer model. By adopting mathematic morphology’s Dilate operation on halftone image and filtering small clustered dot algorithm, dot gain of offset printing is simulated and resulted dot gain curve fits the known standard one. By calculating human perceived images with HVS filter, an image quality evaluation test is conducted with approaches based on both non-ideal printer model and the ideal printer model. The result of the experiment indicates that the halftone image quality evaluation approach based on the non-ideal printer model fits practical offset printing better.更多还原