摘要
研究了红外中波的两个细分波段(3.4~4.1μm和4.5~5.3μm)的成像特点。通过分析视网膜受域的结构特点,利用人眼中心-环绕对抗受域神经网络,设计一种新的神经网络模型,将两幅细分图像进行融合。实验结果表明,融合后的图像和细分图像相比,局部标准偏差和多向粗糙度都增加,太阳影响参数却下降了,融合了两幅细分图像中比较清晰的信息,减少了太阳饱和区,证明了该融合算法的有效性,说明了通过该融合方法可以获得比原红外中波细分图像成像效果更好的图像。
Image characterizations in subdivision band 3.4-4.1 μm and 4.5-5.3μm of mid-wave infrared (MWIR) are studied. Structure of the retina domain and the Center-Surround Shunting Neural Networks (CSSNN) in the human visual are used. In order to fuse these two subdivision images, I design a new network model. Comparing with the subdivision images, the fusion image has the pattern standard deviation and roughness concentration of multi direction increase, moreover the sun effect parameter decreases. The fusion result contains the relatively legible information of double sub-band MWIR images and reduces the sun saturation section, and the validity of algorithm is proved. The experimental results show that image of fusion is better than of original sub-band MWIR.
出处
《红外技术》
CSCD
北大核心
2009年第7期403-406,共4页
Infrared Technology
关键词
图像融合
中波红外
细分波段
中心-环绕对抗受域
Image fusion
mid-wave infrared (MWIR)
subdivisionband
Center-Surround Shunting Neural Networks(CSSNN)