基于多阈值分割的红外与可见光图像融合  

Based on Multi-threshold Segmentation of Infrared and Visible Image Fusion

作　　者：王亮[1] 王敬东[1] 李鹏[1] 丁尤蓉[1] 

机构地区：[1]南京航空航天大学自动化学院,南京210016

出　　处：《光电子技术》2012年第4期278-283,共6页Optoelectronic Technology

摘　　要：对基于NSCT分解的可见光和红外图像的融合算法进行了优化技术研究。在场景复杂或有多个区域需要同时关注时,低频子图融合阶段采用简单的二值化区域分割方法并不能将不同区域的感兴趣目标有效地反映到融合结果中。所以本文采用基于最大熵的多阈值分割方法,并对每一个分割区域计算其区域能量比指导融合系数的求取;高频融合阶段采用最能保持图像细节的梯度最大规则。实验结果表明本文算法能够将多个区域中的感兴趣目标很好地反映到融合结果中,并保留了可见光图像丰富的细节信息和分辨率高特征,融合图像具有更好的视觉效果。This paper based on the NSCT the decomposition of the visible and infrared image fusion algorithm to optimize the technical research. In the complex scene or many regions needs to be focused on. In the step of low frequency process we adopt simple binary regional segmentation method but it can not effectively reflect different parts of the target which we are interested in to the fusion results. For this reason we use the way of multi-threshold segmentation based on the maximum entropy. And for each break up area calculating the area ratio of fusion energy coeffi- cient for guidance. High frequency fusion step we adopt the rules of maximum the gradient to keep image details. The experiment results show that this algorithm could successfully put targets in different images into one result image. And it keeps the optical image rich details information and high resolution characteristics. Fusion image has better visual effect.

关 键 词：图像融合 非下采样CONTOURLET变换 多阈值分割 区域能量比 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]