基于光照估计滤波的太赫兹图像融合研究

Terahertz image fusion based on illumination estimation filtering

碳纤维增强复合材料广泛应用于航空航天等军民领域,在制备和使用过程中往往存在诸多缺陷,太赫兹时域光谱成像技术为碳纤维增强复合材料无损检测提供了新方法。在使用太赫兹无损检测技术时,可以通过在时域或频域中选择不同的参数来获取图像,同样的缺陷需要时域和频域不同的参数才可以有效检测出来。采用基于光照估计滤波的算法将不同参数的太赫兹图像进行融合。首先,通过光照估计滤波分辨初始图像序列中像素的曝光情况。然后隶属函数调用这些初始估计值。最终输出的融合图像是通过隶属函数分配的权重在所有像素之间加权平均得到。实验结果表明,所提出的方法对图像边界信息和细节信息都处理得比较好,适用于太赫兹多波段图像融合,具有更好的视觉效果。

Carbon fiber reinforced composites are widely used in military and civil fields such as aerospace,and there are of⁃ten many defects in the processes of preparation and application.Terahertz time⁃domain spectral imaging technology provides a new method for nondestructive testing of carbon fiber reinforced composites.Images can be obtained by selecting different param⁃eters in time⁃domain or frequency⁃domain when terahertz nondestructive testing technology is applied.The same defect can be ef⁃fectively detected only when different parameters in time⁃domain and frequency⁃domain are required.Terahertz images with differ⁃ent parameters are fused using an algorithm based on light estimation filtering.Firstly,the exposure of pixels in the initial image se⁃quence is resolved by light estimation filtering.The membership function then calls these initial estimates.The final output fusion image is obtained by weighted averaging among all pixels through the weight assigned by the membership function.The experimen⁃tal results show that the proposed method can deal with the edge information and details of the image well,and is suitable for tera⁃hertz multi band image fusion with better visual effect.