基于全变分和梯度域卷积稀疏编码的稀疏角度CT重建算法  

Sparse Angle CT Reconstruction Algorithm Based on Total Variation and Convolutional Sparse Coding in Gradient Domain

作　　者：李雨 史娜[1] 孔慧华[1,2] 雷肖雪 Li Yu;Shi Na;Kong Huihua;Lei Xiaoxue(College of Science,North University of China,Taiyuan,Shanxi 030051,China;Shanxi Key Laboratory of Signal Capturing&Processing,North University of China,Taiyuan,Shanxi 030051,China)

机构地区：[1]中北大学理学院,山西太原030051 [2]中北大学信息探测与处理山西省重点实验室,山西太原030051

出　　处：《激光与光电子学进展》2021年第12期323-332,共10页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(61971381,61871351);山西省自然科学基金(201701D221121)。

摘　　要：对于不完全的扫描数据,传统算法无法保证医学电子计算机断层扫描(CT)重建图像满足诊断要求。根据压缩感知理论,可以从不完全的扫描数据中重建出具有稀疏表示的医学CT图像,这可为诊断提供可靠的信息。从重建的角度出发,提出了一种基于全变分和梯度域卷积稀疏编码的图像重建算法。梯度域卷积稀疏编码是对特征图施加梯度约束,采用梯度正则化约束来抑制离群点,从而解决了因滤波器不准确而造成的结构丢失或新伪影的问题。所提算法直接对整个图像进行操作,以获取局部邻域之间的相关性,并利用梯度图像的全局相关性来产生更好的边缘和清晰的梯度图像特征,它能有效地捕捉到图像的局部特征。此外,通过引进全变分作为正则项,可进一步恢复图像的微小结构和细节并有效地抑制噪声。实验的定性和定量结果表明,与其他算法相比,所提算法在去除伪影的同时保留了更多的细节,具有更高的重建质量,这验证了该方法的有效性。For incomplete scanning data,the traditional algorithm cannot guarantee that the medical computed tomography(CT)reconstruction image meets diagnostic requirements.According to the compressed sensing theory,the medical CT image with sparse representation can be reconstructed from the incomplete scanning data,providing reliable information for diagnosis.From the perspective of reconstruction,this paper proposes an image reconstruction algorithm based on total variation and convolutional sparse coding in gradient domain.Gradient domain convolutional sparse coding is to apply gradient constraint to feature images,and gradient regularization constraint is used to suppress outliers,which solves the problems of structure loss and new artifacts caused by the inaccurate filter.The proposed algorithm directly processes the whole image to obtain the correlation of local neighborhoods,and uses the global correlation of gradient images to generate better edge and clear gradient image features,which can effectively capture the local features of the image.In addition,by introducing total variation as the regularization term,the micro structures and details of the image can be further restored and the noise can be effectively suppressed. Qualitative and quantitative experimental results show that,compared with other algorithms,the proposed algorithm retains more details and has higher reconstruction quality,which verifies the effectiveness of the method.

关 键 词：图像处理 计算机断层成像 稀疏角度 全变分 卷积稀疏编码 梯度图像 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]