条纹
插值(计算机图形学)
算法
投影(关系代数)
计算机科学
混叠
螺旋(铁路)
影子(心理学)
图像(数学)
人工智能
计算机视觉
数学
滤波器(信号处理)
光学
物理
数学分析
心理学
心理治疗师
摘要
Recently, Katsevich proposed an exact filtered backprojection (FBP) algorithm to solve the long object problem. Although the Katsevich algorithm is theoretically error-free, it may produce image artifacts if its numerical implementation is not well designed. To evaluate the Katsevich algorithm, we implement it in planar detector geometry. Then, we study four types of image artifacts associated with the numerical implementation of this algorithm, which are “texture”, “streak”, “shadow” and “winkle” artifacts. The “texture” artifacts appear if the endpoints of the PI-Line are not treated appropriately. The “streak” artifacts are caused by an inadequate number of filtering lines Q, which can be reduced by setting 0.3N<Q<0.7N. In the step of Hilbert filtering, the aliasing effect results in “shadow” artifacts if the window is not truncated properly. The “wrinkle” artifacts are due to both the interpolation and difference operators, which may be minimized using more sophisticated interpolation methods, increasing projection data and using a more accurate derivative formula. Our results should be valuable to apply the Katsevich algorithm optimally for practical spiral cone-beam CT.
科研通智能强力驱动
Strongly Powered by AbleSci AI