基于MATLAB的锥形束CT图像去噪研究

摘要

Objective:Cone-Beam CT is not only a new kind of CT imaging technology,but also the key equipment in the Image-Guided Radiation Therapy system.In view of the low contrast and the large scattering artifacts of Cone-Beam CT images,application and discussion of denoising method in the Cone-Beam CT image based on MATLAB is given for Cone-Beam CT images in this article.The main work is to find the best denoising method for Cone-Beam CT.Methods:We first applied different denoising method,such as neighborhood smoothing,median filter and wavelet denoising method.Then Contourlet transform was used for Cone-Beam CT denoising.We designed different Laplacian pyramid filter and the two-dimensional directional filter bank,to find the optimal filter assembly.Contourlet transform is a new two-dimensional image representation with properties such as multiresolution,localization,anisotropic,directionality and neighbor field sampling.We utilized the advantages of Contourlet transform in the processing of image geometric structure,to extract the image edge continuous feature,to distinguish between noise and edge,and thereby enhanced the image edge information and detail information,as well as suppressing noise.We compared the conventional denoising method,the wavelet denoising method and the Contourlet method with different filter assembly.Results:We made statistics and comparison with the qualities of clinical images of different body parts.The study shows that the wavelet threshold method and the Contourlet method have their different advantages.And the filter assembly of ‘pkva8' and ‘9-7' has the best performance in the Contourlet transform.Conclusions:The spatial smoothing method,the median filter and the traditional wavelet denoising method are no better than the wavelet threshold method and than the Contourlet denoising method in Cone-Beam CT denoising.Furthermore,it is proved that the Contourlet denoising method can effectively improve the CBCT image quality,especially in the chest images.%目的:锥形束CT既是一种全新的CT成像技术,也是图像引导下放射治疗系统的关键设备.针对锥形束CT图像的低对比度,散射伪影较大的缺陷,在MATLAB平台上对CBCT去噪方法进行研究和探讨,以寻找合适的锥形束CT去噪方法.方法:首先应用不同去噪方法,如邻域平滑,中值滤波,小波去噪方法等;再应用Contourlet变换进行锥形束CT去噪,设计不同的拉普拉斯塔式滤波器和二维方向滤波器组,寻找最优的滤波器组合；Contourlet变换是一种新的图像二维表示方法,具有多分辨率,局部定位,多方向性和近邻界采样和各向异性等性质.利用Contourlet变换在处理图像几何结构方面的优点,提取图像中边缘连续特征,来区别噪声和边缘,从而增强图像边缘和细节信息,同时抑制噪声.比较常规去噪,小波去噪,Contourlet去噪和不同滤波器组合去噪效果.结果:结合头部,胸部,盆腔各10组临床图像进行去噪效果统计和分析,表明小波阈值量化法和Contourlet法在锥形束CT图像去噪上各有优势,在Contourlet法中,滤波器组合“pkva8”和“9-7”的去噪效果最好.结论:Contourlet去噪方法和小波阈值量化法都比空间邻域平滑法,中值滤波法和普通小波去噪法有优势.而Contourlet去噪方法更能有效改进CBCT图像质量,特别是胸部图像质量的改善.