基于非下采样轮廓波变换的中子/X射线图像融合算法研究

Neutron and X-ray Radiation Images Fusion Method Study Based on NSCT

由于中子与X射线的成像特性存在很好的互补性,将这两种成像模式下的DR(Digital Radiography)图像进行融合,可以丰富图像细节的信息量。本研究提出了基于简化脉冲耦合神经网络(PCNN)的非下采样轮廓波变换(NSCT)域中子/X射线图像融合算法,所提出的方法通过NSCT算法对源图像进行多尺度、多方向分解,有效地保留了各层信息,使得融合图像展现出良好的清晰度和对比度。同时,为了更好地保留源图像的信息,本研究对图像的不同频带分别采用取平均、绝对值取大和结合PCNN系数的融合规则进行融合。通过选取结构较为复杂的石英表、机械表以及结构相对较为简单的U盘进行中子与X射线图像采集,并利用多种客观评价指标定量评价基于小波变换融合、拉普拉斯变换融合、NSCT融合、NSCT-PCNN融合、Zhu的方法以及本文提出的方法生成中子与X射线融合图像的质量。实验结果证明,本文所提出的融合算法能够显著提高图像融合效果,获得清晰的目标信息,提高了图像对比度,同时保留丰富的细节信息,在客观评价标准和视觉效果上均优于其他典型方法。

Since the characteristics of neutron imaging and X-ray imaging are highly complementary,the fusion of digital radiography(DR)images of these two imaging modes can significantly enrich the information of image details.In this paper,a fusion algorithm of neutron image and X-ray image in nonsubsampled Contourlet transform(NSCT)domain based on simplified pulse coupled neural network(PCNN)is proposed.The proposed method decomposes the source image in multi-scale and multi-directional by NSCT,which effectively retains the information of each layer,so that the fusion image show good clarity and contrast.At the same time,in order to better retain the information of the source image,the different frequency bands of the image are fused by taking the average,taking the largest absolute value and the fusion rules combined with PCNN coefficients.We collect neutron and X-ray images for quartz watch and mechanical watch with complex structures and USB flash disk with simple structures.Then wavelet transform fusion,Laplace transform fusion,NSCT fusion,NSCT-PCNN fusion,Zhu’s method and the proposed method are used to fuse the neutron and X-ray images.The quality of fusion images of different methods is quantitatively evaluated by various objective evaluation indexes.The experimental results show that the fusion algorithm proposed in this paper can significantly improve the image fusion effect,obtain clear target information,improve the image contrast,while retaining rich detailed information,and is superior to other typical methods in objective evaluation criteria and visual effect.