摘要
红外成像技术可以全天候进行工作,受一些恶劣天气因素的干扰较少,因为其依靠目标的温度差和热辐射率来成像﹐缺点也很明显﹐成像结果细节模糊﹐目标与背景的对比度低,边缘平滑﹐视觉效果一般。而偏振图像由于其只保留某些特定方向入射光的特殊成像机制﹐从而可以获取目标与背景粗糙度﹑含水量、物质理化以及纹理特征等多维细节信息,但也正因为滤除了部分方向的光线,导致偏振图像整体亮度偏低。针对单偏振参量信息弱、对比度小而且红外强度图像细节模糊的问题,提出一种基于非下采样剪切波变换(NSST)的红外偏振图像融合算法,可以有效提高复杂背景中目标的辨识度。首先将偏振度图像与偏振角图像采用改进的区域方差方法进行融合得到初始图像。然后对初始图像与红外强度图像进行NSST分解处理,低频分量采用区域关联度与区域方差相结合的融合规则,高频分量的融合采用区域关联度与区域特性能量相结合的规则。实验结果表明,本算法主观视觉效果良好,客观评价指标也优于其他算法。
Infrared imaging technology can work around the clock,ignoring the interference of weather factors,because it relies on the temperature difference and thermal emissivity of the target to image.It also has obvious disadvantages,such as high brightness but fuzzy details,low contrast between the target and the background,smooth edges and general visual effect.Polarized images can obtain multi-dimensional details such as target and background roughness,water content,physicochemical properties and texture features because they only retain the special imaging mechanism of incident light in a specific direction.However,polarized images are generally low in brightness due to the filtering of light in most directions.Aiming at the problem about weak information and low contrast of single polarization parameter image,an infrared polarization image fusion algorithm based on non-subsampling shear wave transform(NSST) is proposed to effectively improve target recognition in complex background.First of all,the image of polarization degree and polarization angle are fused using an improved regional variance method to obtain the initial fusion image.Then perform NSST decomposition processing on the initial image and the infrared intensity image,the low frequency component adopts the fusion rule that combines the regional correlation degree and regional variance,while the fusion rules of the high frequency component resort the regional characteristic energy.Experimental results show that the subjective visual effect of this algorithm is excellent,and the objective evaluation index is also better than other algorithms.
作者
姜兆祯
韩裕生
谢瑞超
任帅军
JIANG Zhao-zhen;HAN Yu-sheng;XIE Rui-chao;REN Shuai-jun(Department of Information Engineering,Army Academy of Artillery and Air Defense Forces,Hefei 230031;Anhui Province Key laboratory of Polarized Imaging Detection Technology,Hefei 230031)
出处
《光电子.激光》
EI
CAS
CSCD
北大核心
2020年第11期1140-1148,共9页
Journal of Optoelectronics·Laser