先验图像约束的全变差正则化CT图像重建算法

Prior image constrained total variation regularization CT image reconstruction algorithm

为解决基于先验图像的压缩感知CT图像重建算法中,当先验图像与待重建图像结构位置存在偏差时,无法得到良好结果的问题,提出一种先验图像约束的全变差正则化CT图像重建算法。选择与待重建图像类似的先验图像,计算先验图像中几种均匀介质的像素平均值作为先验信息;对投影数据进行ART算法重建,并加入非负约束,得到中间图像;将先验信息融入到基于全变差的最小化目标函数中,对中间图像进行优化;以上两步骤循环进行,直到满足收敛准则。采用Shepp-Logan体模进行仿真重建,仿真结果表明,该算法和ART-TV、PICCS算法相比,信噪比更高,误差更小,图像质量更优。

The CT image reconstruction algorithm based on prior image constrained compressed sensing（PICCS）fails to obtain good results when the motion between the prior image and to-be-reconstructed image is apparent.A prior image constrained total variation（TV）regularization reconstruction algorithm was proposed to solve this problem.Firstly,an image with good quality was selected as the prior image which was similar to the to-be-reconstructed image and the average pixel value of the prior image＇s each homogeneous region was calculated as the priori information.Secondly,the projection data were reconstructed using the ART algorithm to obtain the intermediate image.The positivity constraint was added to it.Thirdly,the intermediate image was optimized using TV regularization combining with priori information.The above two steps cycled until the convergence criterion was met.Results of Simulation experiments with the Shepp-Logan phantom demonstrate that the proposed algorithm is better than ART-TV and PICCS algorithm,it significantly improves signal to noise ratio and suppresses the noise,thus the image quality is improved.