肺部LDCT(Low-Dose Computed Tomography)图像中噪声及条状伪影等异常显著,顶部和底部图像尤为严重.为提高整个肺部LDCT图像的质量,本文提出一种基于结构联合字典的图像降噪方法.首先,利用肺部CT图像的灰度特点,将HRCT(High Resolution Computed Tomography)图像块分类并训练,获得4类字典,通过计算原子的信息熵和HOG(Histogram of Oriented Gradient)特征,得到相应的结构字典,进而构造出结构联合字典;然后,在对肺部LDCT图像进行非局部均值滤波的基础上,将结构联合字典作为全局字典,对图像进行稀疏表示及重构,获得降噪后的图像.为验证算法有效性,选用模拟和临床两类数据进行实验,并与KSVD、AS-LNLM、BF-MCA等3种算法对比.对比发现,本文算法在去除噪声和条状伪影以及保留细节方面效果较好,特别是对序列顶层和底层图像处理优势更加明显.该方法能够显著提升整个肺部LDCT图像的质量.%Noise and streak artifacts are serious in low-dose CT ( LDCT) images of lung,especially for the top and bottom layers. To improve the quality of LDCT sequential images,we proposed a method to remove noise and streak artifacts by using a joint structural dictionary. Firstly,according to the gray-scale features of lung CT image,we divided high resolu-tion CT (HRCT) images into four groups,and obtained four dictionaries for each group by dictionary training. Secondly, computed the information entropy and the histogram of oriented gradient (HOG) of each dictionary atom. And then obtained the structure dictionary of corresponding group by feature clustering. Finally,the joint structural dictionary can be obtained by combining these four structure dictionaries. On the basis of non-local mean filtering for LDCT images,the joint structural dictionary is used as a global dictionary to get the denoised image by sparse representation and reconstruction. In order to validate the effectiveness of the proposed method,simulated and clinical LDCT data are used. Compared with reported meth-ods (KSVD,AS-LNLM and BF-MCA),our method performed better in denoising,streak artifact-removing and details-pre-serving for the whole LDCT scans,especially for the top and bottom layers. The proposed algorithm can significantly im-prove the quality of lung LDCT images.
展开▼
