Computer Engineering and Applications ›› 2018, Vol. 54 ›› Issue (22): 180-185.DOI: 10.3778/j.issn.1002-8331.1801-0329

Previous Articles     Next Articles

MAP projection domain denoising based on anisotropic weighted prior model

LI Andi1,2, LIU Yi1,2, ZHANG Quan1,2, GUI Zhiguo1,2   

  1. 1.Shanxi Provincial Key Laboratory for Biomedical Imaging and Big Data, North University of China, Taiyuan 030051, China
    2.School of Information and Communication Engineering, North University of China, Taiyuan 030051, China
  • Online:2018-11-15 Published:2018-11-13

各向异性加权先验模型MAP投影域降噪

李安迪1,2,刘  祎1,2,张  权1,2,桂志国1,2   

  1. 1.中北大学 山西省生物医学成像与影像大数据重点实验室,太原 030051
    2.中北大学 信息与通信工程学院,太原 030051

Abstract: Low-Dose Computed Tomography(LDCT) reduces the radiation of X rays to human body. However, the artifacts and noises caused by the radiation dose decrease in the reconstructed images are very disturbing for clinical medical diagnosis. A new smoothing algorithm, called Maximum A Posteriori(MAP) projection domain denoising algorithm based on improved anisotropic weighted prior model, is proposed to solve this problem. Considering that the intuitionistic fuzzy entropy can distinguish the smooth area and the edge area, and combined with the traditional anisotropic diffusion coefficient, the algorithm constructs a new adaptive diffusion coefficient, and utilizes the local variance to realize its adjustment. Finally, the improved diffusion coefficient is fused to the Huber-based MAP, which realizes different intensity denoising for different regions in projection data. The simulations of the low-dose CT image reconstruction for pelvis model, Shepp-Logan model and thorax phantom model are used to test the effectiveness of the proposed algorithm, and compared with other three algorithms:Filter Back Projection(FBP), Penalized Reweighted Least-Squares(PRWLS) and anisotropic weighted prior sinogram smoothing algorithm. The experimental results show that the artifacts in the reconstructed images are significantly reduced, while the edges and details of the images are well preserved. The signal-to-noise ratio of the three models is 20.5020 dB, 23.2948 dB and 21.0184 dB respectively. The required time is 49.50 s, 49.60 s and 8.59 s respectively.

Key words: low-dose CT, intuitionistic fuzzy entropy, anisotropic diffusion, Huber prior, Maximum A Posteriori(MAP)

摘要: 低剂量计算机断层扫描技术(Low-Dose Computed Tomography,LDCT)降低了X射线对人体的辐射,但射线剂量降低造成重建图像中存在严重的伪影和噪声,对临床医学诊断有很大干扰。针对此问题,提出一种改进的各向异性加权先验模型的最大后验(Maximum A Posteriori,MAP)投影域降噪算法。该算法考虑到直觉模糊熵能够有效区分平滑区域和边缘细节区域,将其与传统的各向异性扩散系数相结合,构造了一种新的扩散系数,并采用局部方差实现其自适应调节;最后将该扩散系数融合于基于Huber先验的MAP优化估计算法框架中,实现对投影数据不同区域进行不同强度的降噪处理。该算法分别采用数字骨盆模型、Shepp-Logan头模型和数字胸腔模型三种体模进行验证,并与滤波反投影重建算法(Filter Back Projection,FBP)、惩罚重加权最小二乘法(Penalized Reweighted Least-Squares,PRWLS)、各向异性加权先验正弦图平滑算法进行对比。实验结果表明,利用所提算法重建出的图像中伪影明显减少,同时较好地保持了图像的边缘和细节信息。三种体模的信噪比分别为20.502 0 dB、23.294 8 dB、21.018 4 dB,所需时间分别为49.50 s、49.60 s、8.59 s。

关键词: 低剂量CT, 直觉模糊熵, 各向异性扩散, Huber先验, 最大后验(MAP)