中文摘要：

针对SAR、红外和可见光图像的灰度差异性大，融合图像感兴趣目标不突出的问题，提出一种基于NSST和稀疏表示的多源异类图像融合方法。首先将训练图像进行NSST变换，在低频系数上构建多尺度学习字典；对SAR、红外和可见光图像进行NSST变换，利用滑动窗口分解低频系数为图像块序列，对图像块序列零均值化后再稀疏分解，采用稀疏系数绝对值取大的融合规则；高频子带系数采用局部方向信息熵显著性因子取大的融合规则；最后对融合系数进行NSST逆变换得到最终的融合图像。

英文摘要：

This paper proposes a multi-source heterogeneous image fusion method based on NSST and sparse presentation to solve the problem that the interested targets are not prominent caused by great grey difference among SAR image, infrared image and visible image. Firstly, multi-scale study dictionary is built on the low frequency coefficients through carrying on NSST for the training image. SAR image, infrared image and visible image are transformed by NSST, and the low frequency coefficients are decomposed into image block sequence with sliding window method. Sparse decomposition is used for image block sequence after zero mean processing, and the fusion rule of the low frequency coefficients is that absolute value of sparse coefficient is larger. The fusion rule of high frequency subband coefficients is that significant factor of local orientation information entropy is larger. Finally final fusion image is obtained by NSST inverse transformation for fusion coefficients.