结合NSST和快速非局部均值滤波的刀具图像去噪

发布时间：2018-03-14 03:36

  本文选题：图像去噪　切入点：非下采样Shearlet变换　出处：《信号处理》2017年11期 　论文类型：期刊论文

【摘要】：为消除基于图像处理的刀具磨损检测中的图像噪声,提出了结合非下采样Shearlet变换(Non-subsampled Shearlet Transform,NSST)和快速非局部均值(Fast Non-local Means,FNLM)滤波的图像去噪方法。首先,利用基于决策的非对称剪切中值(Decision Based Un-symmetric Trimmed Median,DBUTM)方法滤除图像中的椒盐噪声;然后,对图像进行NSST多尺度分解,得到一个低频子带和一系列高频子带;最后,分别使用FNLM滤波和各向异性扩散模型调整低频和高频子带系数,并由调整后的各子带系数重构出噪声滤除后的图像。实验结果表明,与基于小波的阈值收缩方法、基于Contourlet的全变差模型结合各向异性扩散方法、基于NSST和标准非局部均值滤波方法相比,本文方法在主观视觉去噪效果、峰值信噪比、结构相似度以及处理速度等4个方面性能更优。
[Abstract]:In order to eliminate image noise in tool wear detection based on image processing, an image denoising method combining non-subsampled Shearlet transform NSST with fast nonlocal mean fast Non-local means FNLM filter is proposed. The decision Based Un-symmetric Trimmed DBUTM method is used to filter the salt and pepper noise in the image. Then, a low frequency subband and a series of high frequency subbands are obtained by the NSST multiscale decomposition of the image. The low frequency and high frequency subband coefficients are adjusted by FNLM filter and anisotropic diffusion model respectively, and the noise filtered images are reconstructed from the adjusted subband coefficients. The experimental results show that the proposed method is similar to the wavelet based threshold shrinkage method. The total variation model based on Contourlet combined with anisotropic diffusion method, compared with the standard non-local mean filtering method based on NSST, is applied to the subjective vision de-noising, peak signal-to-noise ratio (PSNR). The performance of structure similarity and processing speed is better.
【作者单位】： 南京航空航天大学电子信息工程学院;西华大学制造与自动化省高校重点实验室;
【基金】：国家自然科学基金资助项目(61573183) 西华大学制造与自动化省高校重点实验室开放课题(S2jj2014-028)
【分类号】：TG71;TP391.41
，

本文编号：1609428