【作者】 李祥；

【导师】 刘凤玉；

【作者基本信息】 南京理工大学 ， 计算机应用技术， 2010， 博士

【摘要】 图像融合是指通过特定算法将工作在不同波长范围、具有不同成像机理的图像传感器对同一场景的多个图像信息融合成一个新的图像,从而克服单一传感器图像在几何、光谱和空间分辨率等方面存在的局限性和差异性,使融合图像可信度更高、模糊更少、可理解性更好,更适合人的视觉与计算机检测、分类、识别、理解等处理,从而扩大时空的传感范围、提高观测的准确性和鲁棒性。图像融合是多传感器数据融合的一个重要分支,是计算机视觉和图像理解领域中的一项新技术。本论文集中阐述了图像融合技术的研究背景与现状,对图像融合中的一些基础问题进行了深入分析,并结合微束分析处理领域,对基于内容特征的电子探针图像融合过程中的一些关键问题展开了研究,完成的主要工作和取得的研究成果如下：1、在深入分析Contourlet变换基础理论的基础上,结合Contourlet变换在图像处理方面的成功应用经验,对适应于电子探针图像的Contourlet变换融合方法和流程进行研究,并提出了基于提升小波-Contourlet变换的改进型区域融合方法,能较好的体现来自多源电子探针图像的信息量；2、结合图像稀疏表示与人眼对图像观察的视觉特性,在分析Bandelet变换基本特点与各向异性优点的基础上,提出将Bandelet变换应用于规则形状的电子探针图像融合处理,并通过优化融合规则的选择方法,进一步增强图像融合的信息量；3、在分析Directionlet变换的基础上,针对电子探针图像中的部分强边缘特征图像及多焦距影像,提出基于Directionlet变换的能获取强边缘特征的增强型图像融合方法,进一步丰富了电子探针不同焦距图像的复合处理需求；4、对融合应用进行了技术性与实用性探索。一方面,对基于融合图像的三维重构技术进行了研究,在获取融合图像的基础上,结合DEM成图原理和OpenGL技术优势,提出了基于融合图像的灰度三维重构的优化算法,构建了基于OpenGL的微表面三维影像场景；另一方面,完成基于ARM体系平台图像融合应用构建的基础性研究,从系统的软硬件框架设计、USB驱动的数据采集驱动开发到图像视频数据获取、数据压缩等,都进行了较系统研究。更多还原

【Abstract】 Image fusion refers to multiple images which acquired by image sensors with different mechanisms and wavelength range at the same scene into a new image by using specific algorithms,so that overcome the single sensor’s limitations and differences such as geometry, spectral and spatial resolution,as well as improve the credibility of fusion image and reduce ambiguity,which is more suitable for human visual and computer inspection, classification, identification, comprehension processing, thereby expand the scope of space-time sensor as well as improve the accuracy and robustness.Image fusion is an important branch of multisensor data fusion and a new tech of computer visual field.This paper focuses on the background and presents situation of image fusion technique,analyses some basic problems further,and study on some key problems based on content combined with microbeam analysis field.the main job and research achievements that has acquired are as follows:1.Based on analysing the theory of Contourlet transform,combined with the application experience of Contourlet transform in image process field,this paper studies on Contourlet fusion method and procedure that adapts to electron probe image.and presents an improved region method based on wavelet-Contourlet transform.which can better show the information from multisource images.2.Combined with image sparse representation and the visual feature of human eyes,based on analysing the feature of Bandelet and the advantage of anisotropy.by optimizing the fusion rule and from the subjective and objective angle.this paper presents an idea that applies Bandelet transform to electron probe image fusion.in order to improve the fusion information further.3.According to some images with strong edge and multi-focal length image in electron probe, this paper proposes an enhanced image fusion that can get strong edge based on Directionlet transform,so that enrich the process demand of electron probe image with the same type but different focal length.4. Explore the fusion application from technique and praticability.On the one hand, this paper studied on 3-D reconstruction technique,based on aquisition of fusion image.combined with DEM mapping principle and OpenGL technological advantages, this paper proposed a three-dimensional reconstruction of the optimization algorithm based on gray-scale image fusion,constructed a three-dimensional images of micro-surface scenes based on Open-GL;On the other hand, this paper has finished the research of image fusion application and have carried out a more systematic study on the basis of ARM platform from the system’s hardware and software framework design, USB-driven data acquisition to the image video data acquisition, data compression，更多还原