论文部分内容阅读
针对红外图像与SAR图像的灰度差异性大、两者融合图像不太符合人类视觉认知的问题,提出了一种基于联合稀疏表示的复Contourlet域红外图像与SAR图像融合方法。首先对红外图像与SAR图像分别进行复Contourlet分解。然后利用K-奇异值分解(K-Singular Value Decomposition,K-SVD)方法获得两幅源图像低频分量的过完备字典,并根据联合稀疏表示模型生成联合字典,通过正交匹配追踪(Orthogonal Matching Pursuit,OMP)方法求出源图像低频分量在联合字典下的稀疏表示系数,接着采用选择最大化策略对两个低频分量的稀疏表示系数进行选取,随后进行稀疏表示重构获得融合的低频分量;对高频分量结合视觉敏感度系数和能量匹配度两个活跃度准则进行融合,以捕获源图像丰富的细节信息。最后经复Contourlet逆变换获得融合图像。与3种经典融合方法及近年来提出的基于非下采样Contourlet变换(Non-Subsampled Contourlet Transform,NSCT)、基于稀疏表示的融合方法相比,该方法能够有效突出源图像的显著特征,最大程度地继承源图像的信息。
Aiming at the problem that the fused image between the two images is not very suitable for the human visual perception, a new method based on joint sparse representation is proposed for the fusion of complex image of Contourlet domain and SAR image. Firstly, the Contourlet decomposition of the infrared image and the SAR image are respectively performed. Then, an overcomplete dictionary of low-frequency components of two source images is obtained by K-Singular Value Decomposition (K-SVD) and a joint dictionary is generated according to a joint sparse representation model. Orthogonal Matching Pursuit , OMP), the sparse representation coefficients of the low-frequency component of the source image under the joint dictionary are obtained, then the sparse representation coefficients of the two low-frequency components are selected by using the maximization strategy, and then the sparse representation is used to reconstruct the low- The high-frequency component combines the two activity criteria of visual sensitivity coefficient and energy matching degree to fuse the rich details of the source image. At last, the fused image is obtained by inverse Contourlet transform. Compared with the three classical fusion methods and the non-subsampled contourlet transform (NSCT) proposed in recent years, the sparse representation based fusion method can effectively highlight the salient features of the source image. Inherit source image information.