基于ASD-POCS算法框架的自引导图像滤波重建算法  

Self-guided image filtering reconstruction algorithm based on the ASD-POCS algorithm framework

作　　者：王红霞 龚长城 WANG Hongxia;GONG Changcheng(School of Mathematics and Statistics,Chongqing Technology and Business University,Chongqing 400067,China;Chongqing Key Laboratory of Statistical Intelligence Computing and Monitoring,Chongqing 400067,China)

机构地区：[1]重庆工商大学数学与统计学院,重庆400067 [2]统计智能计算与监测重庆市重点实验室,重庆400067

出　　处：《中国体视学与图像分析》2024年第4期303-310,共8页Chinese Journal of Stereology and Image Analysis

基　　金：重庆市自然科学基金(cstc2021jcyj-msxmX0679,CSTB2022NSCQ-MSX0610);重庆市教委科技研究计划(KJQN202000808);重庆市研究生科研创新项目基金(CYS240568)。

摘　　要：计算机断层扫描(computed tomography,CT)技术广泛应用于医学成像和工业无损检测等领域,但是其应用往往受到辐射剂量高或扫描时间长等问题的限制。稀疏角和有限角采样策略能够有效改善上述问题。然而,在医疗诊断中,实现稀疏角采样需要频繁切换管电压或前准直器,这需要复杂的技术。有限角采样虽然易于实现,但由于采集到的投影数据相关性高,使得重建图像往往有明显的伪影。分段有限角(segmental limited-angle,SLA)采样避免了电压或前准直器的快速切换,降低了投影数据的相关性,有助于提升成像质量。然而,SLA投影数据继承了有限角投影数据的部分特点,重建图像仍然可能存在明显的伪影。本文在自适应梯度下降一投影到凸集(adaptive steepest descent-projection onto convex sets,ASD-POCS)算法框架下设计了求解自引导图像滤波(self-guided image filtering,SGIF)重建模型的算法。数值模拟实验表明,本文提出的算法有利于提高SLA CT图像的质量。Computed Tomography(CT)technology is widely applied in the fields of medical imaging and industrial non-destructive testing.However,its application is often limited by high radiation doses or long scanning times.Sparse-angle and limited-angle sampling strategies can effectively mitigate these issues.However,in medical diagnosis,implementing sparse-angle sampling requires frequent switching of tube voltage or pre-collimators,which involves complex technology.Although limited-angle sampling is easier to implement,the high correlation between the acquired projection data often leads to noticeable artifacts in reconstructed images.The segmental limited-angle(SLA)sampling technique circumvents the rapid switching of voltage or pre-collimators,reduces the correlation of the projection data,and helps improve image quality.Nevertheless,SLA projection data still inherits some characteristics of limited-angle projection data,and reconstructed images may still exhibit significant artifacts.In this paper,we propose an algorithm for solving a self-guided image filtering(SGIF)reconstruction model within the framework of the adaptive steepest descent-projection onto convex sets(ASD-POCS)algorithm.Numerical simulations demonstrate that the proposed algorithm effectively improves the quality of SLA CT images.

关 键 词：计算机断层成像 分段有限角 ASD-POCS算法 自引导图像滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]