条纹
插值(计算机图形学)
计算机科学
人工智能
计算机视觉
图像缩放
图像(数学)
理想(伦理)
图像质量
图像处理
算法
模式识别(心理学)
光学
物理
认识论
哲学
作者
Wang Dian-hua,Yuanjin Li,Yudong Zhang,Tao Wang
标识
DOI:10.1142/s0218213021400108
摘要
In the course of generating the CT images, the streak metal artifacts emerge from the reconstructed images, often degraded the quality of the images and blur the fringe information around the metal implant. Although a number of attempts had been reported, among them, our proposed interpolation-based method is the simplest and most efficient approaches. In this paper, three interpolation approaches are compared with subjective and objective criterion based on both simulation and clinical cases. Our results have shown an improvement from the original images. As for the comparison with NRMSD and MAD. For the execution time, the L-MAR possesses the shortest time with S-MAR time being the slowest among the interpolation-based methods. For NRMSD and MAD, the digits from small to large are P-MAR, S-MAR, L-MAR and original. This shows that among interpolation-based methods the image corrected by P-MAR approach is the closest to the ideal image, followed by S-MAR correction, L-MAR correction, and the gap between the original image and the ideal image is the largest.
科研通智能强力驱动
Strongly Powered by AbleSci AI