基于泊松-高斯混合型噪音强度估计的中子图像复原方法研究

Neutron imaging technique is a powerful tool for non-destructive testing of materials for industrial applications and research. However, it is physically inevitable that the images made with neutron imaging system are degraded severely due to some physical limitations of neutron imaging systems. Based on the experiences that the neutron imaging system with the neutron image restoration method was researched and developed successfully in our university, the project focuses on developing a novel and systematic neutron image restoration scheme based on practical Poisson-Gaussian noise estimation modeling. For dealing with the problem that there is no objective noise estimation way in practical restoration applications, noise estimation modeling will be presented as the key for neutron image denoising and restoration, by which the estimated result also can be referred in the design of neutron imaging system. On the other hand, the corresponding method also can be applied for other radiographic image restoration. By establishing a neutron imaging experimental platform, several degraded real images will be obtained and analyzed by achieving their sparse representations under frame basis. The corresponding restoration scheme can be arranged and provided by investigating some regularized deconvolution methods including Richardson-Lucy method. As a consequence, the key problems in this project will be solved by Poisson-Gaussian noise estimation modeling and neutron image restoration modeling, with which the degradation mechanism of neutron images can be revealed intrinsically. As a result, for the neutron imaging technique concerned, the project can provide some important results not only in theoretical field but also in practical applications.

中子成像技术是一类重要的无损检测技术，但由于中子自身所具有的物理性质以及中子成像系统在成像过程中的物理限制，中子图像都不可避免的遭到了严重降质。本项目在我们已有研制中子成像系统和实现中子图像复原的成功经验基础之上，拟开展中子图像里泊松-高斯混合型噪音强度的估计研究，提出一种新的中子图像复原解决方案。实现图像噪音强度的估计，能够为真实复原应用中的去噪处理提供客观评估依据，是实现中子图像复原方法建模的关键，同时实现噪音强度的评估也能够为中子成像系统的设计提供参考，并被其他成像复原研究所借鉴。通过建立专用的成像实验平台，基于图像的稀疏表示对降质图像进行分析，构建正则化方法开展复原研究，解决噪音强度估计建模、中子图像复原方法建模等关键技术问题。对揭示中子图像的降质规律，对进一步开展中子成像系统的研制以及中子图像处理研究，都具有重要的理论意义和应用价值。

中子成像技术是一类重要的无损检测技术，在实际应用中可以成为X光成像技术的有效补充。由于中子自身的物理性质以及成像系统在实现成像过程中的物理条件约束，中子图像在成像过程中会遭到了严重降质。在项目执行期间，课题组借助已有的中子数字成像研究基础，重点围绕本项目关注的两大问题，即遭受泊松-高斯混合型噪音降质的中子图像复原问题和中子数字图像中的噪音强度估计问题，展开了有序的系统性研究。在研究方法上，课题组坚决遵从了项目评审专家的指导意见，即重点关注中子成像的物理过程，避免形成一个脱离物理过程的纯数学表达。截止目前，课题组已经在如下三个问题研究方面取得了重要进展，这包括中子数字成像的物理过程建模问题、降质中子数学图像的复原方法以及具有实用性的成像噪音强度估计方法。到项目结题，课题组共发表论文十篇，其中被SCI检索的论文六篇，另有三篇十分重要的成果正在投稿或者整理中，相关内容在结题报告中都有阐述。在项目的支持和帮助下，课题组不仅在中子数字成像的若干问题研究中获得了成果，而且也加强了对外的联系，扩大了自己的视野，增厚自身的学术基础。