基于NSST和区域生长的红外与可见光图像融合算法  

An infrared and visible image fusion method based on NSST and region growing algorithm

作　　者：罗杰 孔韦韦[2] 刘睿 

机构地区：[1]武警工程大学研究生管理大队,西安710086 [2]武警工程大学信息工程系,西安710086

出　　处：《中国科技论文》2016年第14期1673-1678,共6页China Sciencepaper

基　　金：国家自然科学基金资助项目(61309008)

摘　　要：针对当前红外与可见光图像融合中红外目标提取信息度不足和灰度对比度较低的问题,将多尺度多方向变换技术与目标提取技术相结合,提出了1种基于非下采样剪切波变换(non-subsampled shearlet transform,NSST)和区域生长的红外与可见光图像融合算法。首先对红外图像运用区域生长法提取出目标区域并投影到可见光图像中,然后对映射图像与红外图像做NSST分解得到各尺度各方向的子带图像。低通子带融合系数的选择采用基于区域能量的加权平均法,高频子带系数的选择采用基于改进型Sobel算子和区域平均梯度的融合规则。最后通过对融合系数做NSST逆变换得到期望的融合图像。实验结果表明,所提算法得到的融合图像红外目标突出,背景清晰,定位准确,平均灰度也更接近源可见光图像,具有良好的视觉效果。An image fusion method is presented for insufficient information content and low contrast ratio of fused images which combines the multi-scale and multi-directional technology with target extraction based on Non-Subsampled Shearlet Transform （NSST） and region growing algorithm. Firstly, the infrared image target region is extracted by growing algorithm and is mapped to the visible image. Then, the obtained fusion image is fused with the original infrared image by using NSST. The weighted average based on region energy is used in the fusion of low-frequencycoefficient, and the selection of fused high-frequency coefficient uses improved Sobel and averagegradient Finally, thefusion image was obtained after the NSST inverse transformation Experimental results indicate that fused image obtained bythe proposed methodnot only extracts the target characteristics more effectively and accurately, but also reserve the clear background. It receives a better fusion effect in the evaluation.

关 键 词：图像融合 红外图像 目标提取 区域生长 非下采样剪切波变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]