针对低照度应用场景,提出一种基于卷积神经网络的可见光与近红外融合算法,采用端到端网络实现了图像融合,所得融合图像能够兼顾近红外图像的信噪比与可见光图像的色彩。采集了真实场景下精确配准的近红外-可见光图像对作为训练集样本,提升了网络对实际数据的融合效果。通过对训练样本进行融合预处理,提升了网络对近红外图像中细节纹理信息的提取能力。各项测试表明,本文算法在主观感受和客观评价上均优于现有算法。

Herein, for low-illumination application scenes, an end-to-end convolutional neural network (CNN) is proposed for the fusion of near-infrared (NIR) and visible images. The fused image can combine the signal-to-noise ratio of an NIR image and the color of visible images. To verify the capability of CNN for practical fusion tasks, a real dataset with accurate registration was collected. Moreover, the training set was preprocessed via information fusion, thereby enabling the network to extract additional information from NIR images. Experimental results reveal that the proposed method is superior to existing fusion methods in terms of visual quality and quantitative measurements.