基于视觉显著图和相位一致性的红外与可见光图像融合算法

Infrared and visible image fusion algorithm based on visual significance map and phase consistency

为了获得更好的视觉融合效果,同时抑制边缘处的伪影和保留更多源图像特征,提出了基于视觉显著图(VSM)和相位一致性(PC)的红外和可见光图像融合算法。该方法利用信息损失更少的各向异性滤波(AnisGF)将源图像分解为多个高频信息细节层和一个低频信息基础层;利用VSM获取基础层的权重图,融合红外(IR)与可见光(Vis)基础层图像;利用最大-绝对值和相位一致性(PC)获取细节层的权重图,融合IR和Vis细节层图像,达到强化图像边缘信息并抑制噪声的目的,最后通过图像重构获得最终融合图像。实验结果表明,文章提出的方法与RFN-nest等五种方法相比,在清晰度、对比度和伪影等视觉效果均有改善。在AG(平均梯度)、EN(熵)、MI(互信息)、SF(空间频率)四种评价指标方面,均优于其他算法;在VIF(视觉信息保真度)方面,仅次于MSID-KBS方法。

In order to obtain better visual fusion effect,while suppressing artifacts at the edge and preserving more source image features,an infrared and visible image fusion algorithm based on visual salient map(VSM) and phase consistency(PC)is proposed.This method uses anisotropic filtering(AnisGF) with less information loss to decompose the source image into multiple high-frequency information detail layers and one low-frequency information base layer.Obtain the weight map of the base layer by using VSM,and fuse the infrared(IR) and visible(Vis) base layer images;The weighted map of the detail layer is obtained by maximum-absolute value and phase consistency(PC),and the IR and Vis detail layer images are fused to enhance the image edge information and suppress noise,and finally the final fused image is obtained through image reconstruction.Experimental results show that the proposed methods have improved the visual effects of clarity,contrast and artifacts compared with five methods,including RFN-nest.It is superior to other algorithms in terms of AG(mean gradient),EN(entropy),MI(mutual information),and SF(spatial frequency).In terms of VIF(visual information fidelity),it is second only to the MSID-KBS method.