基于NSCT和混合粒子群算法的红外与可见光图像融合

Infrared and Visual Image Fusion Based on NSCT and Hybrid Particle Swarm Optimization

针对红外与可见光图像融合的特点,提出一种基于非下采样Contourlet变换(NSCT)和混合粒子群算法的红外与可见光图像融合算法。通过NSCT变换对红外图像和可见光源图像进行分解,对低频子图像,采用一种基于区域平均值改进的加权平均法进行邻域融合,对高频子图像的最高层采用区域标准差选大法进行融合;对高频子图像的其他层采用以混合粒子群优化算法选取阈值,基于平均梯度选择的邻域算法进行融合。最后进行NSCT逆变换得到融合图像。实验结果表明该方法可以获得融合效果更佳的融合图像。

Aiming at the characteristics of the infrared and visual image, an algorithm for image fusion based on Non-Subsampled Contourlet Transform （NSCT） and hybrid particle swarm optimization is proposed. NSCT is used for decomposition of the infrared and visual source images. To the low-frequency sub-images, an improved weighted average method based on regional averages is adopted for neighborhood fusion. To the top layer of the high-frequency sub-images, the fusion method of ＂choosing the biggest＂ for the regional standard deviation is adopted; to the other layers of the high-frequency sub-images, the hybrid particle swarm optimization is used for selecting the threshold value, and the neighborhood algorithm based on the average gradient selection is adopted for fusion. Finally, NSCT inverse transform is utilized to obtain the fused image. The experimental results show that the method can obtain the ideal fusion image and more detail information.