基于图像增强和滚动引导滤波的红外与可见光图像融合

Infrared and Visible Image Fusion Based on Image Enhancement and Rolling Guidance Filtering

针对红外与可见光图像融合易发生热目标亮度损失、可见光图像细节信息丢失的问题,提出一种基于图像增强和滚动引导滤波的多尺度融合算法。首先,提出一种自适应图像增强方法,提高可见光图像的整体亮度,并保持细节处的对比度。然后,根据特征的不同将源图像分解为三层,采用基于引导滤波的显著性提取方法得到亮度层;利用滚动引导滤波良好的尺度感知和边缘保持特性,并结合高斯滤波得到基础层和细节层。最后,对亮度层采用像素值取大的融合规则,提出一种新的最小二乘优化方案对基础层进行融合,使用修正拉普拉斯能量和作为清晰度的度量对细节层进行融合。实验结果表明,与其他融合方法相比,所提方法在主观评价和客观评价上都有较好的表现。

A multiscale fusion algorithm based on image enhancement and rolling guidance filtering is proposed to solve the problems of thermal target brightness loss and visible image detail information loss caused by infrared and visible image fusion.First,an adaptive image enhancement method is proposed to improve the overall brightness of the visible image and maintain the contrast of the details.Second,according to the different features,the source image is divided into three layers,and the luminance layer is obtained by using the significant extraction method based on guidance filtering.The favorable scale perception and edge preservation characteristics of rolling guidance filtering are used,and the basic layer and detail layer are obtained by combining Gaussian filtering.Finally,the fusion rule of large pixel value is used for the luminance layer,a least-squares optimization scheme is proposed for the basic layer,and the sum of the modified Laplace energy is used as a measure of sharpness for the detail layer.The experimental results show that,compared with other fusion methods,the proposed method has better performances in both subjective and objective evaluations.