【作者】 李斌；

【导师】 沈民奋；

【作者基本信息】 汕头大学 ， 信号与信息处理， 2009， 硕士

【摘要】 图像配准是将不同时间、不同传感器或不同视角拍摄的同一场景的两幅或多幅图像的匹配过程,以获得更为全面、更为准确的图像描述供观察或进一步处理。目前图像配准技术已经在许多领域中得到了广泛的应用,在图像融合、图像跟踪和卫星遥感系统几个应用领域中的作用尤为突出。经过多年的研究,关于图像配准技术的研究已经取得许多成果,但在实际应用中,仍存在许多亟待解决的问题。图像配准方法应该达到高效高精度的要求。近年来,国内外的众多研究工作都在致力于各种基于特征配准方法的研究。基于小波变换的数字图像处理方法引起了人们的高度重视,为了达到快速高精度配准的目的,将小波变换、角点特征检测、马氏距离测度有效的结合起来并应用到图像配准技术中将是一种可行的方法。本文首先阐述了研究背景,提出了目前在基于特征的图像配准技术中存在的问题,表明本文的研究目的和内容;接着引入小波理论,分析小波理论在角点检测中应用的可能性和有效性;然后将介绍目前常用角点检测技术、评价方法、角点检测的指导思想和发展趋势,在此基础上提出一种基于SUSAN技术的自适应阈值角点提取方法,并将上述介绍的小波理论分析方法和自适应SUSAN角点检测技术相结合,通过实验给出基于SUSAN的多尺度角点检测方法的实验结果和并分析其优点;最后通过马氏距离相似性测度来作为图像间的匹配准则,采用全局遍历搜索的方法实现图像配准。图像配准技术正朝着高速高效高精度的方向发展。基于角点特征的图像配准技术,能有效的提高图像的配准速度,而精确的角点检测算子则为配准的精确性提供了保障。本文的研究结果表明,将小波变换、角点特征检测、马氏距离测度有效的结合起来并应用到图像配准技术中可以有效的提升图像的配准速度和精度。更多还原

【Abstract】 Image registration is the process to match multiple images about the same scene which is shot from different time, different sensors or different perspectives . So that we can obtain more comprehensive and more accurate description of the image for observation or further treatment. Now, image registration technology has been widely used in many areas, such as image fusion, image tracking and satellite remote sensing systems. After years of research, image registration techniques has been achieved many results, but in practical applications, there are still many problems. Image registration method should be efficient to meet the requirements of high precision. In recent years, a large number of research work at home and abroad are working on registration method which is based on the feature. Based on wavelet transform in digital image processing method has aroused great attention. In order to achieve rapid high-precision registration purpose, the wavelet transform, corner detection, Mahalanobis distance measure combined with an effective and applied to image registration techniques would be a feasible solution.In this paper, we firstly introduce the research background, proposed the exist problems in the image registration based on feature, and show the aims and content about this paper. Next, we introduced wavelet theory, analyzed the wavelet theory in the application of corner detection and the possibility of validity. And then introduced the guiding ideology and the development trend currently about corner detection technology , the evaluation method of corner detection. On this basis, we proposed the adaptive threshold corner extraction method based on the SUSAN corner detector. Then combined with the wavelet theoretical analysis methods and adaptive SUSAN corner detection technology, given the multi-scale corner detection method based on the SUSAN through the experiments and analysis of its merits from the experimental results. Lastly, take the Mahalanobis distance as image matching similarity criteria, use the global search method to implement image registration.Image Registration is trending to the efficient and high-speed and high precision. The image registration techniques based on the feature of the corner, can effectively improve the speed of image registration, and accurate corner detector has provided a guarantee for the accuracy of registration. These research results show that the wavelet transform, corner detection, Mahalanobis distance measure can effectively enhance the speed and accuracy of image registration.更多还原