北京生物医学工程

基于RCSCT变换的DR图像去噪及加速

DR image de-noising and acceleration based on recursive cycle spinning contourlet transformation

作者：林芳宇罗海周荷琴

单位：中国科学技术大学信息科学技术学院(合肥 230027)

关键词： DR图像去噪；Contourlet变换；递归循环平移；计算统一设备架构

分类号：

出版年·卷·期（页码）：2012·31·3（245-250）

摘要：

目的数字化X线摄影(digital radiography，DR)图像中的高斯噪声对图像质量影响大，消除此类噪声有利于提高图像质量以辅助医生做出正确的诊断。方法为抑制DR图像的高斯噪声，首先采用递归循环平移与Contourlet变换结合的（recursive cycle spinning Contourlet transform，RCSCT）方法变换分解DR图像，接着采用连续的二元软阈值函数处理变换系数防止系数被过度扼杀，然后基于CUDA（compute unified device architecture,计算统一设备架构）平台对去噪方法加速。结果该方法提高了去噪后的图像峰值信噪比，有效抑制了伪吉布斯现象,保留了更多的图像细节信息，并且加速处理后运算耗时较短。结论本文方法比小波变换和Contourlet变换在保留视觉细节信息方面效果更优，算法耗时少，实用性好。

Objective The influence of Gaussian noise on digital radiography(DR)image is great，and the removal of Gaussian noise is beneficial to the image quality and clinical diagnosis. Methods To suppress Gaussian noise of DR image，this paper first decomposes DR image based on recursive cycle spinning Contourlet transform(RCSCT)that combines recursive cycle spinning and Contourlet transform，then adopts continuous binary soft threshold function to process the transformed coefficients，which can prevent coefficients over killed，and subsequently accelerates the de-noising method based on compute unified device architecture(CUDA)platform. Results The experimental results show that the suggested method can obtain higher PSNR value，inhibit Gibbs-like phenomena，and preserve more image details with shorter time-consuming after acceleration. Conclusions The proposed method based on RCSCT is better than wavelet transform and Contourlet transform in practicability，time consuming and preservation of visual information.

参考文献：

［1］丰国栋，何祥彬，周荷琴. 基于Laplace-Impact混合模型的DR图像去噪算法［J］. 中国医疗器械杂志，2009，33(4)：247-250.
Feng Guodong, He Xiangbin, Zhou Heqin. DR image denoising based on Laplace-Impact mixture model［J］. Chinese Journal of Medical lnstrumentation, 2009, 33 (4): 247-250.
［2］Candes EJ，Do DL. Curvelets：A surprisingly effective non-adaptive representation for objects with edges［C］. Curve and Surface Fitting. Nashvil-le，TN，USA：Vanderbilt University Press，1999.
［3］Do MN，Vetterli M. Contourlets：A directional multiresolution image representation［C］. Proc of IEEE International Conference on Image Processing. Rochester，NY：2002： 357-360.
［4］Coifman RR，Donoho DL. Translation invariant denoising［C］. Wavelets and Statistics，Springer Lecture Notes in Statistics 103. New York：Springer Verlag，1995： 125-150.
［5］Fletcher AK，Ramchandran K，Goyal VK. Wave-let denoising by recursive cycle spinning［C］. Proc IEEE International Conference Image Processing. Rochester，NY:2002： 873-876.
［6］Fletcher AK，Ramchandran K,Goyal VK. Iterative projective wavelet methods for denoising［J］. Proc Wavelets：Appl in Sig & Image Proc X，part of SPIE Int Symp on Optical Sci & Tech，2003，5207(1)：9-15.
［7］Eslami R，Radha H. The contourlet transform for image de-noising using cycle spinning［C］. Asilo-mar Conference on Signals，Systems，and Compu-ters. Pacific Grove，USA：2003. 1982-1986.
［8］梁栋，沈敏，高清维，等. 一种基于Contourlet递归Cycle Spinning的图像去噪方法［J］. 电子学报，2005，33(11)：2044-2046.
Liang Dong,Shen Min, Gao Qingwei, et al. A method for image de-noising based on the Contourlet transform using recursive cycle spinning［J］. Acta Electronica Sinica, 2005, 33(11): 2044-2046.
［9］杨晓慧，焦李成，牛宏娟，等. 基于多阈值的非下采样轮廓波图像去噪方法［J］. 计算机工程，2010，36(4)：200-204.
Yang Xiaohui, Jiao Licheng, Niu Hongjuan, et al. Image Denoising Method for Nonsubsampled Contourlet Based on Multi-threshold［J］. Computer Engineering, 2010, 36(4): 200-204.
［10］何祥彬，周荷琴，李方勇. 基于CUDA平台的DR图像增强处理加速算法［J］. 中国医疗器械杂志，2010，34(1)：9-11.
He Xiangbin, Zhou Heqin, Li Fangyong. A new approach to accelerate DR image enhancement based on CUDA［J］. Chinese Journal of Medical lnstrumentation, 2010, 34(1): 9-11.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录