【作者】 方华；

【导师】 陆系群；

【作者基本信息】 浙江大学 ， 计算机应用技术， 2006， 硕士

【摘要】 图像半色调技术是指用少量的色彩将一幅连续色调图像(如灰度图像和彩色图像)量化为一幅二值图像或是只有少数几种色彩的彩色图像,并且量化后图像在一定距离的视觉效果和原始图像相似的技术。半色调技术在低分辨率的设备上有着广泛的应用背景,如激光或油墨打印机、传真机和廉价液晶显示器等。许多移动显示设备上也用到了半色调技术,如蜂窝式手机和PDA。 半色调图像作为未来几十年里传播信息的重要手段之一,被广泛地应用在报纸、杂志、身份证、货币、支票和机密文件等工业上。由于Intemet网络是一个开放式的网络架构,数字信息传输的安全技术十分重要,如数字水印技术。数字水印技术能够将具有特定意义的标记(水印),隐藏在数字图像、声音、文档、图书、视频等数字产品中,用以证明创作者对其作品的所有权,同时通过对水印的检测和分析来保证数字信息的完整可靠性。 合理安排像素点,用有限的色域范围来表现图像连续色调效果是半色调技术的关键。本文介绍了几种常用灰度图像的半色调方法,如有序抖动(ordered dithering),直接折半查找法(direct binary search),标准误差传递法(error diffusion)以及自适应error diffusion方法,还介绍了彩色图像的error diffusion(误差传递)半色调系统,并从算法的计算复杂度、视觉效果等方面对这些方法的性能作了对比分析。文章还介绍了几种评价半色调算法性能的标准。 本文介绍了目前有代表性的半色调图像的数字水印技术,并对这些算法的抗攻击能力和负荷量作了定量的分析,对其优缺点作了一一点评。通过分析可知半色调图像由于受到嵌入水印的影响,其与原始连续色调图像之间的视觉效果差距较之未加水印前变大。而且现有的数字水印技术一般只对较温和的攻击较为敏感,对有些恶意攻击则鲁棒性较差。寻找对尽可能多的攻击算法有免疫性的半色调图像的数字水印方法将持续成为研究的主要方向之一。 论文提出了一种新的测评error diffusion半色调算法视觉特性的方法。使用一组符合人眼视觉系统(Human Visual System,HVS)工作原理的整型滤波器阵列(integer-coefficient filter bank)对半色调图像在水平和垂直方向上进行低通和高通滤波,将图像分解为16个子带,然后使用高阶分布特征统计量(Skewness和Kurtosis)对16个子带图像的能量分布特性进行统计分析,原始图像和误差图像之间相关性也被用来定量计算由于半色调技术所造成的失真程度。实验使用几种常见的error diffusion半色调算子(Floyd-Steinberg、Jarvis和Stucki)对一幅常值灰度图像进行半色调,实验结果表明Jarvis算子生成的半色调图像的能量分布特征最接近于正态分布。相关性的实验结果也和前人的分析工作结果相一致。 论文最后对图像半色调技术和水印的研究作了总结,对应用前景和发展方向进行了讨论。更多还原

【Abstract】 Digital halftoning is to quantize a continuous-tone image to a binary image or a color image with limited colors that the output halftoned image looks like the original multitone image when viewed from a distance. Halftone images routinely appear in low-resolution facilities, such as laser and inkjet printers, facsimile machines and low-cost liquid crystal displays. It can also be applied in mobile equipments, such as cellphones and PDAs.Halftone image is commonly applied in the industries of newspaper & magazine publication, identification, currency, check or confidential documents. It would be one of the most important means of exchanging information in the foreseeable future. Since the internet is an open network infrastructure, it is indispensable to employ effective protection mechanisms, for example the watermarking techniques, to protect halftone images from being attacked when transmitted along the network. Digital watermarking is a process to embed sign information (watermark) into a digital signal, such as digital image, speech, text document, book or video, to protect the ownership of the digital information, to prevent the illegal use without permission, and to authenticate the integrity of the content by examining the existence of the watermark.The key problem of digital halftoning is how to arrange the pixels of the output halftoned image with a limited scope of a gamut. Several halftone techniques for graylevel images are introduced in this paper, such as ordered dithering, direct binary search, standard error diffusion and adaptive error diffusion, as well as a color error diffusion halftoning system. The performances of the presented algorithms are evaluated by comparing the computational complexity and the visual effect. Furthermore, this paper presents some new performance criterions for halftone algorithms.We also study some representative watermarking technologies for halftone images and quantitatively analyze the robustness and payload of these algorithms. From experimental results, we find that, comparing with a halftoned image without a watermark, the visual perception of the output halftoned image with a watermark is very different from the original continuous-tone image. And these watermarking methods could only resist some mild attacks but be vulnerable to other vicious attacks. It would be one of the unremitting research works to seek watermarking methods for halftoned images that could resist as many attack algorithms as possible.This paper proposes a new evaluation scheme to assess the halftone qualities of different error diffusion halftoning algorithms. An integer-coefficient filter bank is applied to every error diffusion halftone image, and the halftone image is decomposed into 16 subbands by using low-pass filters and high-pass filters along the horizon and vertical orientations respectively. Based on these subband images, high-order distribution statistics are defined to evaluate the qualities of different error diffusion filters. The correlations between the subbands of the residual image and the original image are calculated to measure the distortion caused by the error filter in error diffusion halftoning. In order to evaluate the qualities of different error diffusion filters, we use a constant-tone patch to be halftoned by three commonly used error diffusion filters (Floyd-Steinberg, Jarvis and Stucki) and the energy distribution of the patch generated by Jarvis error filter is most similar to the normal distribution among these three error filters. And our experimental results of correlation are consistent with the previous research works.Finally this paper discusses the future research works and the potential application of digital halftoning and the watermarking technologies.更多还原