基于目标提取与拉普拉斯变换的红外和可见光图像融合算法

Image Fusion Algorithm of Infrared and Visible Images Based on Target Extraction and Laplace Transformation

为了在可见光图像中尽可能的突出红外目标,提高红外与可见光融合图像的质量,提出了一种基于目标提取的红外与可见光图像融合算法。对红外图像分别进行边缘提取和阈值分割,将提取的两幅二值图像进行融合,获得目标图像。然后将红外图像、红外目标图像和进行色度、饱和度、亮度(HSV)空间转换后的可见光的亮度分量进行拉普拉斯变换,高频系数采用基于区域互信息、匹配度以及区域能量的融合规则,低频系数的融合则在红外目标图像低频系数的指导下,结合基于区域的融合规则进行。最终融合图像重构是通过拉普拉斯变换及反变换实现的。实验结果表明,该融合算法能较好地突出红外目标信息,提供丰富的背景细节,在图像的清晰度和人眼的视觉效果方面优于其他传统算法。

For highlighting the infrared targets in the visible image and advancing the quality of infrared and visible fusion images, an image fusion algorithm of infrared and visible images is presented by target extraction. The two binary images are fused to obtain the target image by edge extraction and threshold segmentation on the infrared images. The infrared image, the infrared target image and the value component of visible image turned to hue, saturation, value （HSV） color space are decomposed multi-resolution by Laplace transformation. The high- frequency decomposition coefficients are fused by rules based on calculating the mutual information, matching degree and energy of corresponding region, and the low-frequency coefficients are fused by region information rules eomhining with the rules based on regional fusing. Lastly the fusion image reconfiguration is realized through Laplace transformation and inverse transformation. Experimental results show that the image fusion algorithm presented highlights the targets of the infrared image as much as possible, and injects details information of the visible image. Fusion image definition and visual effects are better than the traditional algorithms.