基于非下采样Contourlet变换的极化SAR图像融合与去噪

Fusion and speckle reduction of multi-polarization SAR images based on nonsubsampled Contourlet transform

针对非下采样Contourlet变换(NSCT)在处理噪声影像中具有的优势,以及同极化SAR图像(HH、VV)之间的相关性与互补性,本文实验了一种基于非下采样Contourlet变换的极化图像融合方法。该方法首先对每个极化图像进行多尺度、多方向分解,然后对不同分解子带系数分别采用有利于斑点噪声去除和信息增强的融合规则进行融合,最终通过NSCT反变换得到融合图像。通过信息熵、相关系数以及等效视数等指标的评价,验证了该方法可以有效地实现信息增强,同时该方法也在一定程度上降低了斑点噪声的负面影响。

Nonsubsampled Contourlet Transform （NSCT） is a newly improved multiresolution geometry analysis technique based on Contourlet Transform. NSCT can represent images more effectively because of its flexible multi-resolution, multi-direction and shift invarianee, and has been approved to be very suitable for noisy images, such as SAR images. A method of fusing and reducing the speckles of muhi-polarization SAR images based on NSCT was proposed in the paper. The images were firstly decomposed by NSCT, and then different fusion rules of decomposed coefficients were chosen with the consideration of denoising and information enhancing. Finally, the fused NSCT coefficients were reconstructed to obtain fusion result. In this paper, HH and VV co-polarization images were fused using the proposed method, and the entropy, correlation coefficient and Equivalent Number of Looks （ENL） were calculated to evaluate the results. Experiments showed that the proposed algorithm could obtain a good fused image with the improved spatial information, at the same time the speckle noises could be reduced to some extent.