基于CNN与直方图规定化的红外与低照度可见光图像融合

针对红外与低照度可见光图像融合时，细微的纹理信息不能有效地保留的问题，提出了基于CNN与直方图规定化的红外与低照度可见光图像融合算法。首先，通过基于卷积神经网络的融合方法得到融合后的图像；其次，计算融合后图像的灰度直方图，通过直方图规定化将可见光图像的直方图映射到融合图像的直方图的区间上，以增强图像的纹理信息；最后，将直方图规定化的图像与红外图像通过卷积神经网络的方法进行融合，得到融合图像。实验结果表明，本文提出的算法在视觉效果和客观评价上均优于基于卷积神经网络的融合方法。

Infrared and Low Illumination Visible Light Image Fusion Based on CNN and Histogram Specification

Aiming at the problem that the subtle texture information cannot be effectively preserved in the fusion of infrared and low illumination visible light images, a fusion algorithm of infrared and low illumination visible light images based on CNN and histogram specification is proposed. Firstly, the fused image is obtained by the fusion method based on convolutional neural networks; then, the gray histogram of the fused image is calculated, and the histogram of the visible image is mapped to the interval of the histogram of the fused image through histogram specification to enhance the texture information of the image; finally, the histogram specified image and infrared image are fused by convolution neural network to obtain the fused image. Experimental results show that the proposed algorithm is superior to the fusion method based on convolutional neural networks in visual effect and objective evaluation.