A unified framework for penalized statistical muon tomography reconstruction with edge preservation priors of l_p norm type

详细信息    查看全文

• 作者：Baihui Yu^a ; ^b ; Ziran Zhao^a ; ^b ; ^{zhaozr@mail.tsinghua.edu.cn" class="auth_mail" title="E-mail the corresponding author} ; Xuewu Wang^a ; ^b ; Dufan Wu^a ; ^b ; Zhi Zeng^a ; ^b ; Ming Zeng^a ; ^b ; Yi Wang^a ; ^b ; Jianping Cheng^a ; ^b ; ^{chengjp@mail.tsinghua.edu.cn" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Muon tomography ; MAP algorithm ; Edge preserving priors ; TUMUTY ; FTO model
• 刊名：Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment
• 出版年：2016
• 出版时间：11 January 2016
• 年：2016
• 卷：806
• 期：Complete
• 页码：199-205
• 全文大小：2818 K

文摘

The Tsinghua University MUon Tomography facilitY (TUMUTY) has been built up and it is utilized to reconstruct the special objects with complex structure. Since fine image is required, the conventional Maximum likelihood Scattering and Displacement (MLSD) algorithm is employed. However, due to the statistical characteristics of muon tomography and the data incompleteness, the reconstruction is always instable and accompanied with severe noise. In this paper, we proposed a Maximum a Posterior (MAP) algorithm for muon tomography regularization, where an edge-preserving prior on the scattering density image is introduced to the object function. The prior takes the l_p norm (p>0) of the image gradient magnitude, where p=1 and p=2 are the well-known total-variation (TV) and Gaussian prior respectively. The optimization transfer principle is utilized to minimize the object function in a unified framework. At each iteration the problem is transferred to solving a cubic equation through paraboloidal surrogating. To validate the method, the French Test Object (FTO) is imaged by both numerical simulation and TUMUTY. The proposed algorithm is used for the reconstruction where different norms are detailedly studied, including l₂, l₁, l_0.5, and an l_2–0.5 mixture norm. Compared with MLSD method, MAP achieves better image quality in both structure preservation and noise reduction. Furthermore, compared with the previous work where one dimensional image was acquired, we achieve the relatively clear three dimensional images of FTO, where the inner air hole and the tungsten shell is visible.