基于双边滤波的可见光与红外图像自适应融合

针对传统融合方法不能充分提取可见光图像与红外图像各自特有的细节信息和难以确定融合权值的问题，提出一种基于双边滤波的可见光与红外图像自适应融合算法。该方法结合双边滤波与改进的双边滤波，将可见光与红外图像分解为多尺度局部细节、特有细节以及基础信息。在图像融合阶段，局部细节信息采用基于图像边缘能量的融合方法，最大程度保留了图像的细节；图像基础部分的融合，引入了基于局部能量值的正则化参数来自适应调整融合的权值，优化参数的选择；特有细节信息的融合采用了绝对值最大的融合规则以充分保留源图像特有的细节信息。仿真实验结果表明，通过主观判断所提方法融合后的图像视觉效果更好、对比度高，边缘细节的融合比其他算法更加好，在客观指标的评价中，所提融合方法的MI、EN、FMI、SD、Q^AB/F、Mean 6种指标相对于其他方法综合提升了22.6%、5.7%、0.7%、30.4%、14.2%、18.4%，在主观视觉上和客观评价指标中，所提的算法均具备更优的融合效果。

Visible and Infrared Image Adaptive Fusion Based on Bilateral Filters

Traditional fusion methods cannot fully extract the details of visible and infrared images, and it is hard to determine the fusion weights. An adaptive image fusion method based on bilateral filters is thus proposed. First, an improved bilateral filter is employed to decompose the source images into local details, unique details, and base parts. Second, the local details are merged based on the energy of image edges to retain details to the maximum. The base parts are fused using an adaptive regularization parameter based on local energy. Finally, the fused image is obtained. Experimental results indicate that the fused image obtained by the proposed fusion method has a better visual effect, higher contrast, and clearer edge details. In comparison with other four methods, the Mutual Information(MI), Feature Mutual information(FMI), Standard Deviation(SD), Mount of edge Information(Q), and Mean of the new approach have increased by 22.6%, 5.7%, 0.7%, 30.4%, 14.2%, and 18.4%, respectively. Therefore, the proposed image fusion method has a better performance than other methods in terms of both subjective and objective evaluation.