基于RPCA模型的红外与可见光图像融合

Infrared and visible image fusion based on RPCA model

针对传统基于非下采样Contourlet变换图像融合方法易出现融合图像目标显著性较弱、图像失真以及纹理细节信息缺失的问题,本文提出一种基于RPCA分解模型的NSCT域红外与可见光图像融合方法。首先对红外与可见光图像进行RPCA分解,得到对应的稀疏矩阵;然后利用NSCT变换获取待融合红外与可见光图像的低频子带与高频方向子带,并分别采用基于稀疏矩阵和PCNN的方法求解待融合图像低频子带和高频方向子带的融合系数,再对融合后的图像低频和高频成分进行NSCT逆变换获取最终的图像融合结果;最后分别选取标准和真实场景测试图像集对本文方法和Contourlet、D-NSCT以及NW-P等代表方法进行实验测试对比。实验结果表明,本文方法图像融合结果的目标显著性明显增强,图像的边缘轮廓等细节信息更加丰富,有效地抑制了图像失真现象,相对于其它对比方法具有更高的融合精度和较好的鲁棒性。

For the problem of weak saliency,image distortion and texture loss caused by the traditional non-sampling Contourlet transform based image fusion method,a novel infrared and visible image fusion method based on RPCA decomposition model was proposed in this paper.Firstly,the infrared and visible images were decomposed to obtain the corresponding sparse matrixes by using the RPCA decomposition model.Secondly,the low frequency sub-bands and high frequency sub-bands of the infrared and visible image were extracted by using the non-sampling Contourlet transform.Thirdly,the sparse matrix based and PCNN based fusion strategies were applied to the low-frequency sub-bands and the high frequency direction sub-bands to compute the fusion coefficients.Thus,the fused image could be acquired by using non-sampling Contourlet transform based reconstitution of the low frequency sub-bands and high frequency sub-bands.Finally,the standard and real scene image sets were employed to evaluate the performances of the proposed method with other Contourlet based method,D-NSCT method and N-W-P method.The experimental results showed that the proposed method can improve the saliency efficiently,and overcome the problem of image distortion and texture loss,which indicated the proposed method has the higher accuracy and better robustness compared to the other methods.