【摘要】 针对目前自编码器方法没有考虑到图像所处场景中的光照因素，弱光条件下可见光图像严重退化、视觉感知较差等问题，提出了一种基于自编码器和双注意力机制的红外与可见光图像融合方法。首先在图像预处理阶段采用动态范围压缩增强方法，增强可见光图像整体对比度，其次在残差融合网络中加入通道注意力和空间注意力，形成双注意力机制，使得融合图像在维持可见光图像中背景信息的同时能够突出目标信息，最后为了使编码器和解码器能有一个更好的特征提取能力与特征重建能力，引入了一个两阶段训练策略。实验结果表明，本方法在多个定量评价指标上表现更优。融合图像的互信息提升了10%，结构相似性增加了8%，标准差提高了5%，这些指标表明，视觉效果明显优于现有方法。更多还原

【Abstract】 In order to solve the problems that the current autoencoder method does not take into account the lighting factors in the scene in which the image is located, and the visible light image is seriously degraded and the visual perception is poor under low light conditions, a fusion method of infrared and visible light image based on autoencoder and dual attention mechanism was proposed. Firstly, the dynamic range compression enhancement method is used in the image preprocessing stage to enhance the overall contrast of the visible light image, and secondly, channel attention and spatial attention are added to the residual fusion network to form a dual attention mechanism, so that the fused image can highlight the target information while maintaining the background information in the visible light image,and finally, in order to make the encoder and decoder have better feature extraction ability and feature reconstruction ability, a two-stage training strategy is introduced. Experimental results show that the proposed method performs better on multiple quantitative evaluation indexes. The mutual information of the fused images is increased by 10%, the structural similarity is increased by 8%, and the standard deviation is increased by 5%, which indicates that the visual effect is significantly better than the existing methods.更多还原