面向目标获取的图像质量表征与增强

CORC > 沈阳自动化研究所 > 中国科学院沈阳自动化研究所 > 光电信息技术研究室

题名	面向目标获取的图像质量表征与增强
作者	徐德江
学位类别	博士
答辩日期	2013-11-29
授予单位	中国科学院沈阳自动化研究所
导师	史泽林
关键词	人眼视觉目标获取性能图像质量表征图像信杂比图像增强
其他题名	Targeting-oriented Image Quality: Characterization and Enhancement
学位专业	模式识别与智能系统
中文摘要	如何从复杂的自然场景中快速、准确地获取目标，是光电成像探测技术研究的一个重点。长期以来，光电探测技术更多地关注在目标频谱特性上，利用目标与场景的辐射差异来检测和发现目标。视觉生理学和心理物理学的研究结果表明，图像结构特征的差异对目标获取性能存在重要的影响，这为目标获取提供了一个新的研究方向。本文重点围绕复杂场景图像的质量表征及图像增强方法这一科学问题展开了深入的研究。第一章介绍研究背景与意义，归纳总结了图像质量表征与增强方法的研究现状。第二章介绍了与本文有关的人眼视觉生理学、心理学以及心理物理学研究成果，分析了图像感知的结构自适应性及对比度掩盖效应等规律对人眼获取目标信息的影响机理。第三章提出了面向目标获取的新的图像质量表征量—图像信杂比SCR，给出了定义，建立了SCR与目标获取性能的关系模型。首先，在“图像杂波是一种人眼感知效应”的已有认识基础上，建立了人眼感知特性空间，并给出了感知特征计算模型，提出了目标与杂波相对差异的一种图像信杂比度量—FD因子；分析了大脑认知对人眼就目标—杂波的差异感知行为的影响机理，提出了“杂波是一种人眼感知与大脑认知的综合效应”的一种新认识，建立了认知敏感特征空间，给出了认知特征计算模型，结合大脑注意机制，进一步给出了一种新的图像信杂比表征模型DSIM。同时，建立了“目标—背景—人眼视觉—探测概率”目标获取性能模型，实验表明其预测性能优于美国Army Reseach Laboratory的CCM模型。最后，对信杂比表征模型进行了评估。第四章围绕如何提高图像信杂比DSIM，给出了一种图像对比度增强频域方法TOCE。深入研究了人眼感知获取图像信息能力下降的机理，结合人眼的多通道性、亮度掩盖效应、对比度掩盖效应以及图像结构的高度自适应性等视觉特性，提出了一种全局图像对比度增强频域算法，在凸显结构信息的同时提高了图像信杂比DSIM。本文是在国家973计划项目—目标探测与识别能力基础问题研究的支持下完成的，成果对复杂场景下目标获取理论与技术的发展起到了促进作用。
索取号	TP391.41/X74/2013
英文摘要	How to quickly and accurately acquire targets in complex natural scenes with severe clutter is a hard task in the research field of imaging detection. For a long time, the photoelectric detection technology tendentiously focuses on the target spectral expansion, distinguishing radiation difference to detect and identify a target from its background. While visual physiology and psychophysics studies suggest that image structural difference has a significant effect on targeting performance, which provides a new research direction in the field of target acquisition. This paper has an in-depth study on the scientific issues that focuses mainly on characterization and enhancement of a target image in complex natural scenes. Chapter I introduced the research background and significance, and summarized the research status in the field of image quality characterization and enhancement. Chapter II introduced the related research in the field of our visual physiology, psychology, and psychophysics, and analyzed the effect mechanism of the image structural adaptability of perception, contrast masking effect, et al. on affecting the target information acquisition. In chapter III gave a targeting-oriented image quality characterization: signal-to-clutter ratio (SCR)， gave its definition and established relationship models between SCR and target acquisition performances. First, based on the existing note that image clutter is a visual perceptual effect, we established human perceptual characteristics space, built a model for calculating perceptual features, and further put forward a kind of image signal-to-clutter ratio (SCR) metric that is FD factor for measuring the relative difference between a target and its background clutter. Analyzing the effect mechanism of brain cognition on the human perceptual activity to the difference between the target and clutter images, we proposed a new awareness that clutter is a combined effect of human perception and brain cognition, established brain cognitive characteristics space, built a model for calculating cognitive feature. Combining with brain attention mechanism, we further gave a new characterization model DSIM for image SCR. Meanwhile, we built "target-background-HVS-detection probability" target acquisition performance models. Experiments verified that their predicted performance were better than the CCM model which was given by the Army Research Laboratory of United States. Finally, we evaluated the stability of the proposed SCR models. Focusing on improving the rapidity and effectiveness of acquiring a target in complex natural scenes，in chapter IV we put forward a targeting-oriented image contrast enhancement in frequency domain (TOCE). Based on in-depth study on the mechanism of disability of the human vision for acquiring the image information, we gave a global image contrast enhancement method according to the human visual characteristics including multichannel, brightness masking effect, contrast masking effect, and highly image structural adaptability. The TOCE highlighted the image structural information while increasing the image signal-to-clutter ratio DSIM. The work has been supported by the national 973 project that is the basic research on the capability of target detection and identification. Its achievements play a catalytic role on the development of the target acquisition theory and technology in complex natural scenes.
语种	中文
产权排序	1
页码	118页
分类号	TP391.41
内容类型	学位论文
源URL	[http://ir.sia.ac.cn/handle/173321/14797]
专题	沈阳自动化研究所_光电信息技术研究室
推荐引用方式 GB/T 7714	徐德江. 面向目标获取的图像质量表征与增强[D]. 中国科学院沈阳自动化研究所. 2013.