脊波变换在全色图像和多光谱图像融合中的应用分析-application analysis of ridgelet transform in panchromatic image and multispectral image fusion.docx

摘要图像融合是信息融合中以图像为研究对象的分支, 图像融合就是把多个不同模式的 图像传感器获得的同一场景的多幅图像，或同一传感器在不同时刻或不同方式下获得的 多幅图像合成一幅包含输入图像相关信息的图像的过程. 本文研究脊波变换在图像融合 中的应用。论文分析讨论了脊波变换的特点和实现过程，首先在图像的二维 Fourier 变换域上 进行 Radon 变换得到 Radon 变换域,然后在 Radon 域上的每个极径上进行一维小波变换。 在实现 Radon 变换时采用双线性插值法，虽然牺牲了一部分精确性，但是提高了算法效 率。论文基于 Slant Stack 变换,采用低频部分加权平均,高频部分取绝对值最大的融合 规则, 给出了一种基于脊波变换的多光谱图像和全色图像融合算法,并使用灰度方差,清 晰度,信息熵等评价参数对基于脊波变换的融合结果和基于小波的融合结果进行了比较, 结果表明前者的效果在图像的边缘细节信息的表示方面优于后者。关键词: 多光谱图像，全色图像，脊波变换，Radon 变换，Slant Stack 变换，图像融合ABSTRACTImage fusion is a branch of the information fusion. Image fusion’s research field is the fusion methods of images come from different sensors or images from the same sensor at different moment. Through taking advantage of the complementary information and the redundant information, the output image of image fusion can provide more comprehensive and more accurate description about the scene. The technique of the image fusion has been applied in the fields of military matter, remote sensing, robot, the processing of medical image and machine vision.This paper describes the theory of the Ridgelet transform and provides a way of the implementation of the Ridgelet transform. At first, through the radon transform on the 2-d Fourier domain, we get radon domain, second, using 1-d wavelet transform on every direction on the radon domain, we get the Ridgelet coefficients.BaseonSlantStacktransform,usingweightedaverageinthelow frequency part and choose-Max in the high frequency part as the fusion rules, we put the way of implementation of the Ridgelet transform into the fusionof Multispectral and Panchromatic Images. With the Gray variance, Information entropy and sharpness, the experiment show the performance of the Ridgelet transform in representing the high-dimension singularity is better than wavelet transform.Keyword: Multispectral image, Panchromatic image, Ridgelet transform, Radon transform, Slant Stack, image fusion目 录1 绪论.....................................................