基于改进非局部先验的Bayesian低剂量CT投影平滑算法

Improved nonlocal prior-based Bayesian sinogram smoothing algorithm for low-dose CT

针对低剂量CT(LDCT)图像质量退化的问题,提出了一种改进的非局部先验模型,并将基于该模型的Bayesian统计算法应用于LDCT投影降噪中.首先将方向性测度引入到传统的非局部先验模型中,构建一种改进的先验模型;同时结合基于加权欧氏距离的距离测度,提高权重系数计算的准确性;然后运用基于该先验模型的Bayesian统计算法对LDCT投影进行平滑降噪;最后依据降噪后投影,利用滤波反投影(FBP)方法进行重建,得到改善的LDCT图像.实验结果表明,与典型的传统LDCT重建算法相比,该算法在抑制噪声、去除伪影的同时,较好地保留了重建图像细节信息.

Aiming at the problem that low-dose computed tomography （LDCT）image quality is de-graded,an improved nonlocal prior model is proposed and the prior-based Bayesian statistical algo-rithm is also applied to sinogram denoising for LDCT.First,the orientation measure is introduced into the traditional nonlocal prior model to construct a novel prior model.The accuracy of calculating the weight parameters is increased by incorporating the distance measure based on the weighed Eu-clidean distance into the improved prior model.Then,the Bayesian statistical algorithm is used in si-nogram smoothing for LDCT.Finally,the improved reconstructed image for LDCT is obtained by the filtered back-projection （FBP）from the smoothed projection data.Experimental results show that compared with traditional reconstruction algorithms for LDCT,the proposed algorithm is more effective in suppressing noise and eliminating streak artifacts while maintaining more reconstruction image details.