基于LP和DFB结构的离散Contourlet遥感影像融合算法及在QuickBird中的试验研究

The Discrete Contourlet RS Image Fusion and Its Pilot Study in QuickBird Image Based on LP and DFB Structures

针对目前最新发展的Contourlet变换能够比小波变换提供更丰富的方向和形状基,适合进行多尺度边缘增强的特点,利用Contourlet变换用于融合遥感全色和多光谱影像的算法,即利用LP(Laplacian Pyramid)捕获影像的低频分量,利用DFB(Directional Filter Bank)获得影像的高频分量,再对得到的低频近似系数和高频细节系数按照融合规则,采用算术平均和加权算子构造融合影像对应的对比度金字塔;最后,通过逆塔形变换重构融合影像。提出一种基于塔形方向滤波器组PDFB(Pyramidal Directional Filter Bank)的影像融合方法,算法一方面将Contourlet变换这一新的数学工具引入到影像融合中,另一方面对目前高分辨率影像数据源QuickBird进行了融合实验。此外,利用熵、扭曲度、偏差指数、相关系数、标准差等参量,对此融合方法的融合性能进行了评价与分析。实验结果表明:提出的融合算法能在保留多光谱影像光谱信息的同时增强了融合影像的空间细节表现能力和信息量,该算法是有效可行的。

Image fusion is a very important research field in remote sensing. Recently developed contourlet transform can offer a much richer set of direct ions and shapes,and thus it is more effective than wavelet in capturing smooth contours and geometric structures in images. A novel method of fusing panchromatic and multispectral remote sensing images is proposed based on contourlet transform. Using the laplacian pyramid（LP）analysis to capture the low frequency information of images,and directional filter bank（DFB）to obtain the high frequency information of images, and using an image fusion scheme based on arithmetical average and plus-power to construct contrast pyramid. Fusion images were obtained through inverse cont- ourlet transform. A laplacian pyramid and directional filter bank（PDFB）are given in this article. On one hand,with the proposed scheme,the contourlet transform as a new mathematical tool is introduced to remote sensing image fusion area. On the other hand, the algorithm is used QuickBird image experimentation. Besides, using information entropy, distorted tolerance, difference index, correlation coefficient, standard difference ere to evaluate and analyze the fusion results. The experimental results show that the proposed method can effectively retain spectral information. It can also improve spatial resolution and spatial information of multispectral images.