摘要
鉴于NL-means算法在图像去噪领域的卓越表现,提出一种基于旋转不变NL-means迭代优化的POCS CT图像重建算法。第一步,进行POCS重建,使图像满足投影一致性和非负性,第二步,对重建图像进行旋转不变NL-means优化,平滑均匀区域,保护边缘信息。以上两步交替进行,直到满足停止迭代准则。在稀疏角度和有限角度两种情况下,对该方法进行shepp-logan体模重建实验,通过对重建图像的视觉和指标评估,所提方法能更有效地抑制图像噪声,信噪比更高,重建结果更接近理想图像。
To reconstruct good CT image from few projection views, based on the fact that nonlocal means （NL-means） filter behaves well in noise removing, an iterative POLES image reconstruction using rotationally invariant NL-means method is proposed. In each iteration reconstruction, POCS method is firstly used to reconstruct an image, and then the process of ensuring the non- negativity of the reconstructed image is performed. The second step is to perform rotationally invariant NL-means method to the above image, maintaining the edge information well and smoothing the homogenic region. Reconstruction experiment of the Shepp-Logan phantom is presented to test the feasibility of the algorithm in few projection views. Results show that the proposed method could reconstruct CT image with a higher signal to noise ratio, maintaining the image edge information and closer to the true image.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第7期2487-2491,共5页
Computer Engineering and Design
基金
山西省科技攻关基金项目(20110322004)
中北大学青年科学基金项目(2011)
