中文摘要：

为了更好地将可见光图像与噪声干扰严重的合成孔径雷达图像融合，提出了一种最大尺度硬阈值去噪的方法，在此基础上设计了一种融合规则，根据噪声和信号在NSCT（nonsubsampled Contourlet transform）域的分解系数特性，将NSCT分解的最大尺首先进行硬阈值去噪，其他高频尺度与最大尺度对应的像素点取值方式保持一致，在低频系数采用“简单绝对值取大”的融合规则，最后进行NSCT逆变换得到融合图像。实验结果表明，该方法能有效抑制斑点噪声，并能充分保留源图像重要特征。

英文摘要：

To obtain better fusion results of optical images and SAR images with severe speckle noise, this paper proposed a hard threshold denoising method in the finest scale level（DF-FSL） and a fusion algorithm. According to different properties of noise and signal＇s NSCT decomposition coefficients, frist the algorithm carried out hard threshold denoising in the finest scale level of high frequency coefficients. Then it fused other scale levels of high frequency coefficients in accordance with the finest scale level. Thirdly, it employed＇ choose absolute maximum＇ to select the low frequency coefficients of the fused image. Last it obtained the fusion images by inverse NSCT. The experiment resuhs show that the fusion rule not only restrains the speckle noise effectively, but also preserves salient information of original images well.