基于卷积神经网络的红外图像融合算法

红外图像与可见光图像融合的目的是为人类观察或其他计算机视觉任务生成信息更加丰富的图像。本文针对深度学习近年来在计算机视觉领域取得的巨大成功,提出一种基于卷积神经网络的红外与可见光图像融合算法。首先,使用引导滤波和高斯滤波器组成的尺度感知边缘保护滤波器对输入的源图像进行多尺度分解,基础层利用像素强度分布的加权平均融合规则进行融合,细节层借助卷积神经网络对空间细节进行提取融合。实验结果表明,本文算法可以较好的将特定尺度信息进行保存,并减小滤波对边缘细节带来的光晕影响,融合后图像噪声较少,细节呈现的更加自然,并且适合人类视觉感知。

Infrared image fusion algorithm based on convolutional neural network

The purpose of fusion of infrared and visible images is to generate more informative images for human observation or other computer vision tasks.In this paper,based on the great success of deep learning in the field of computer vision in recent years,an infrared and visible light image fusion algorithm based on convolutional neural network is proposed.Firstly,the input source image is decomposed at multiple scales by using a scale-aware edge protection filter composed of bootstrap filter and Gaussian filter.The base layer is fused by using the weighted average fusion rule of pixel intensity distribution,and the detail layer is extracted and fused by means of convolutional neural network.The experimental results show that the proposed algorithm can preserve the information of specific scales and reduce the influence of filtering on the edge details.After fusion,the image has less noise and the details present more naturally,which is suitable for human visual perception.