动态射线数字图像序列降噪算法及快速实现

兵工学报 ›› 2011, Vol. 32 ›› Issue (11): 1331-1336.

动态射线数字图像序列降噪算法及快速实现

杨民¹，孟凡勇²，梁丽红³，魏东波¹

(1.北京航空航天大学机械工程及自动化学院，北京 100191；2.中国科学院过程工程研究所多相复杂系统国家重点实验室，北京 100190；3.中国特种设备检测研究院，北京 100013)

收稿日期:2010-03-12 修回日期:2010-03-12 上线日期:2014-05-04
通讯作者: 杨民 E-mail:minyang.ndt@263.net
作者简介:杨民（1975—），男，副教授，博士
基金资助:
国家自然科学基金资助项目（6087208061077011）; 中国科学院过程工程研究所多相复杂系统国家重点实验室开放基金资助项目（MPCS- 2011-D-03）

Denoising Algorithm for Dynamic Digital X-ray Image Sequences and Its Fast Implementation

YANG Min¹， MENG Fan-yong²， LIANG Li-hong³， WEI Dong-bo¹

(1.School of Mechanical Engineering and Automation, Beijing University of Aeronautics and Astronautics, Beijing 100191, China;2.State Key Laboratory of Multiphase Complex Systems，Institute of Process Engineering, Chinese Academy of Science,3.China Special Equipment Inspection and Research Institute，Beijing 100013, China)

Received:2010-03-12 Revised:2010-03-12 Online:2014-05-04
Contact: YANG Min E-mail:minyang.ndt@263.net

摘要/Abstract

摘要： 为了实现动态射线数字成像(DR）系统的快速降噪，分析了X射线数字实时成像系统的噪声特点和原始的基于图像序列的NL-means降噪算法的不足，提出了一种改进的NL-means序列图像降噪算法。同时，为了解决NL-means降噪算法计算量大、运算速度慢的问题，利用可编程图形处理单元（GPU）并行计算和高速浮点计算特性，将图像映射为GPU中的纹理，采用多线程并行计算，使得NL-means算法在GPU中加速执行。实验结果表明，改进之后的算法有效地抑制了动态DR图像中的随机噪声，同时保留了图像的细节信息。GPU加速方法可以在不损失图像信息的前提下，实现实时降噪。

关键词: 信息处理技术, 射线数字成像, 图像降噪, NL-means算法, 图形处理单元

Abstract: In order to realize the fast noise reduction in the dynamic digital radiography (DR) imaging system, the system’s noise characteristics and the disadvantages of the original NL-means denoising algorithm were studied. An improved NL-means noise reduction algorithm was proposed. Meanwhile, in order to solve its calculation complexity problem, graphic processing unit (GPU) was applied to make use of its high parallel and fast floating-point calculation abilities. In the implementation, the image was mapped to the GPU’s texture and multi-thread parallel calculation was adopted. The experimental results showed that the improved NL-means noise reduction algorithm could effectively restrain the quantum noise and improve the resolution of DR image sequences. In addition, the acceleration with GPU could realize the real-time denoising without the loss of image information.

Key words: information processing, digital radiography, image denoising, NL-means algorithm, graphic processing unit

中图分类号:

TP391

杨民，孟凡勇，梁丽红，魏东波. 动态射线数字图像序列降噪算法及快速实现[J]. 兵工学报, 2011, 32(11): 1331-1336.

YANG Min， MENG Fan-yong， LIANG Li-hong， WEI Dong-bo. Denoising Algorithm for Dynamic Digital X-ray Image Sequences and Its Fast Implementation[J]. Acta Armamentarii, 2011, 32(11): 1331-1336.

参考文献

［1］ Lee H Y, Park D S, Lee S D, et al. An effective image enhancement filtering for noisy image sequences[J]. Proc of SPIE-IS&T Electronic Imaging, 2006, 6069: 1-9.
[2] Aleksandra Pizurica, Vladimir Zlokolica, Wilfried Philips. Noise reduction in video sequences using wavelet-domain and temporal filtering[J]. Proc of SPIE, 2004, 5266: 48-59.
[3] Samy R. An adaptive image sequence filtering scheme based on motion detection[J]. Proc of SPIE 1985, 596:135-144.
[4] Sezan M I, Ozkan M K, Fogel S V, et al. Temporally adaptive filtering of noisy sequences using a robust motion estimation algorithm[C]∥Proceedings of the Int Conf Acoustics Speech and Signal Processing, Toronto: 1991：2429-2432.
[5] Kokaram A C. Motion picture restoration[D]. Cambridge: Cambridge University, 1993.
[6] Martinez D M. Model-based motion estimation and its application to restoration and interpolation of motion pictures[D]. Massachusetts: Massachusetts Institute of Technology, 1986.
[7] Buades A, Coll B, Morel J M. A non-local algorithm for image denoising[C]∥Proc of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE, 2005:60-65.
[8] Buades A, Coll B, Morel J M. Denoising image sequences does not require motion estimation[C]. Proc of the IEEE Conf on Advanced Video and Signal Based Surveillance. IEEE, 2005:70-74.
[9] 李保磊,杨民,李俊江. 基于改进NL-means算法的显微CT图像降噪[J]. 北京航空航天大学学报, 2009,35(7): 833-837.
LI Bao-lei, YANG Min, LI Jun-jiang. Denoising of micro-CT image based on improved NL-means algorithm[J]. Journal of Beijing University of Aeronautics and Astronautics, 2009,35(7):833-837. (in Chinese)
[10] Mueller K, Neophytou N, Xu F. MIC-GPU: high-performance computing for medical imaging on programmable graphics hardware (GPUs)[J]. SPIE Medical Imaging, 2007:1-46.
[11] Harada T. Real-time rigid body simulation on GPUs[M]. NVIDIA. GPU Gems3, County of Santa Clara: 2007: 611-632.
[12] Deschizeaux B, Blanc J Y. Imaging earth’s subsurface using CUDA[J]. NVIDIA GPU Gems3, 2007: 831- 850.
[13] Xu F, Mueller K. Accelerating popular tomographic reconstruction algorithms on commodity PC graphics hardware[J]. IEEE Trans Nuclear Sci, 2005, 52(3): 654-663.

[1]	李静, 刘哲, 黄文准. 基于快速非局部均值和超分辨率重建的图像降噪算法[J]. 兵工学报, 2021, 42(8): 1716-1727.
[2]	曾溢良，蓝金辉，邹金霖. 滑动置信度约束的红外弱小目标跟踪算法研究[J]. 兵工学报, 2017, 38(9): 1771-1778.
[3]	李关防，崔杰，袁富宇. 基于线谱瞬时频率估计的被动声纳目标运动分析[J]. 兵工学报, 2017, 38(7): 1395-1401.
[4]	朱甦，薄煜明，何亮. 基于冗余字典的多特征压缩感知目标跟踪算法[J]. 兵工学报, 2017, 38(6): 1140-1146.
[5]	王啸, 韩太林, 张永立, 刘轩, 王义君, 宫玉琳. 多通道瞬态信号自适应变频算法[J]. 兵工学报, 2017, 38(6): 1161-1167.
[6]	张石，张百海，王飞帆，关子霄. 基于跳数量化的无线传感器网络节点定位算法[J]. 兵工学报, 2017, 38(5): 932-939.
[7]	夏平, 任强, 吴涛, 雷帮军. 融合多尺度统计信息模糊C均值聚类与Markov随机场的小波域声纳图像分割[J]. 兵工学报, 2017, 38(5): 940-948.
[8]	曹俊，郑翠娥，孙大军，张殿伦. 基于凸优化的水下载体定位研究[J]. 兵工学报, 2017, 38(3): 520-526.
[9]	向颉，茅永兴，郭才发. 船载雷达测速数据的船速修正新方法[J]. 兵工学报, 2017, 38(11): 2268-2273.
[10]	潘江怀，乔慧. 多传感器空间配准的病态性及其稳健估计[J]. 兵工学报, 2017, 38(10): 1965-1973.
[11]	王清，丁赤飚，付琨，任文娟. 基于膨胀运算的移动对象兴趣点检测方法[J]. 兵工学报, 2017, 38(10): 2041-2047.
[12]	李子高，李淑秋，闻疏琳. 一种基于独立成分分析的水下宽带信号波达方位估计方法[J]. 兵工学报, 2017, 38(1): 114-122.
[13]	刘泳庆，蒋硕，于栋，邵晓田. 窄带干扰抑制对码相位测量零值的影响分析[J]. 兵工学报, 2016, 37(2): 293-298.
[14]	管四海, 李智，黄辉，王哲. 带零吸收项的变步长l₀范数归一化最小均方误差算法[J]. 兵工学报, 2016, 37(11): 2170-2176.
[15]	焦传海，李永成，谢恺，杨运甫. 基于期望偏差和广义似然比检验的非重构宽带压缩盲感知[J]. 兵工学报, 2016, 37(10): 1837-1843.