基于视觉显著性的可见红外图像融合

多波段图像融合可以有效综合各个波段图像中包含的特征信息。针对可见光-红外图像，本文提出了一种结合红外图像视觉显著性提取的双波段图像融合方法，一方面旨在凸显红外图像的目标信息，同时又能尽可能的保留可见光图像的丰富细节信息。首先，在局部窗口内实现红外图像的显著性图提取，并通过窗口尺寸的变化形成多尺度的显著性图，并对这些显著性图进行最大值的优选叠加，以获取能反映整幅红外图像各个尺寸目标的显著性图；其次，通过结合显著性图与红外图实现显著性图的加权增强；最后，利用增强的红外显著性图进行双波段图像的融合。通过两组对比实验，数据表明该方法给出的融合图像视觉效果好，运算速度快，客观评价值优于对比的7种融合方法。

Dual-band image fusion for visible and infrared images based on image visual saliency

Dual-band image fusion is able to well synthesize the feature information from the different bands. To fuse visible and infrared images, in this paper, an infrared image visual saliency detection-based approach was proposed. This method aimed to highlight the target information from infrared image, meanwhile preserve abundant detail information from visible one as much as possible. Firstly, visual saliency map was extracted within a local window, and multiple window-based saliency maps could be obtained by changing the size of local window. And the final saliency map was generated by selecting maximum value, and this map could mirror all target information in infrared image. Secondly, the saliency map was enhanced by combining infrared image and the previous saliency map. Finally, the enhanced saliency map was used for dual-band image fusion. Comparing with other seven methods, the experimental results demonstrate that the proposed approach could rapidly produce fused image with good visual effect, and the objective evaluation is better.