基于区域和方向方差加权信息熵的图像融合

提出了一种基于非下采样Contourlet变换（nonsubsampled Contourlet transform, NSCT）的红外与可见光图像融合方法。首先对原红外图像进行图像分割，确定目标区域与背景区域，并将其映射到可见光图像中；然后对红外和可见光图像进行多尺度、多方向分解，分解后的低频部分在目标区域选择红外图像低频系数、在背景区域选择可见光图像低频系数，高频部分使用方向方差加权信息熵最大作为融合策略进行融合；最后对融合的系数进行重构得到融合图像。实验结果表明，本文算法在保留图像细节信息、增加信息量、方便目标检测方面都有显著地提高。

Image fusion based on region and directional variance weighted entropy

A fusion algorithm for infrared and visible images based on nonsubsampled Contourlet transform (NSCT) is proposed. Firstly, the infrared image is segmented to target region and background region, which is also mapped to the visible image. Secondly, the NSCT is performed on the infrared and visible image at different scales and different directions. In the target region, the low-frequency coefficients of the fusion image are selected with those of infrared image, and in the background region the low-frequency coefficients of the fusion image are selected with those of visible image. The high-frequency coefficients are fused with a rule of the maximum directional variance weighted entropy. Finally, the fused coefficients are reconstructed to obtain the fused image. The experimental results show that the proposed image fusion algorithm gets more image detail in the information, the amount of information increases significantly, and the target detection in the fused image becomes simpler.