旋转不变非局域均值算法在磁共振图像去噪中的应用
发布时间：2018-10-18 13:58
【摘要】：低场或快速成像得到的磁共振图像信噪比往往较低,临床应用中常常采用增加采集次数,将多帧图像累加平均来提高图像的信噪比。但是,在多次采集过程中,人体的自主或不自主的运动使得多幅图像之间产生相对偏移。因此,相干累加平均得到的图像边缘或细节会出现模糊。针对这一问题,我们之前的算法是基于非局域均值算法(Non-Local Means,NLM),利用图像的局部相似性计算出图像之间的局部偏移量,对图像进行局部偏移校正后再做加权平均,以达到提高信噪比的目的。本文在此基础上提出一种旋转不变的非局域均值方法(Rotation-invariant Non-local Means,RINLM)。该方法采用圆形邻域区域,并将其划分为以中心像素为圆心的一系列等面积的同心圆环,再计算邻域模式之间的相似性。与NLM算法相比,本文方法可以利用图像中发生相对旋转的相似邻域模式,提高算法的去噪性能。将旋转不变的非局域均值算法应用于图像序列的累加和去噪中,本文方法可以克服局部运动的旋转成分对计算的影响,从而更好地处理存在旋转的局部运动的情况,进一步提高图像质量。本文利用模拟数据和临床真实数据进行了实验,并采用主观和客观的方法对实验结果进行了评价和分析。结果显示,与前人方法相比,本文方法可以进一步提高图像的信噪比,更好的保持图像边缘细节信息。
[Abstract]:The signal-to-noise ratio (SNR) of magnetic resonance images obtained by low field or fast imaging is often low. In clinical application, increasing the acquisition times and adding the average of multi-frame images are often used to improve the signal-to-noise ratio (SNR) of the images. However, in the process of multiple acquisition, the autonomous or involuntary movement of the human body causes the relative deviation between multiple images. Therefore, the edges or details of the image obtained by the coherent cumulative average will be blurred. In order to solve this problem, our previous algorithm is based on the non-local mean algorithm (Non-Local Means,NLM), using the local similarity of the image to calculate the local offset between images, and then doing the weighted average after the local offset correction of the image. In order to improve the signal-to-noise ratio. In this paper, a rotation-invariant nonlocal mean method (Rotation-invariant Non-local Means,RINLM) is proposed. In this method, the circular neighborhood region is used and divided into a series of concentric rings with the center pixel as the center, and the similarity between the neighborhood patterns is calculated. Compared with the NLM algorithm, the proposed method can improve the denoising performance by using the similar neighborhood pattern of relative rotation in the image. By applying the rotation-invariant nonlocal mean algorithm to the accumulation and denoising of image sequences, the method in this paper can overcome the influence of the rotational component of the local motion on the calculation, and thus better deal with the local motion with rotation. Further improve the image quality. In this paper, simulated data and clinical real data were used to evaluate and analyze the experimental results by subjective and objective methods. The results show that compared with the previous methods, the proposed method can further improve the signal-to-noise ratio (SNR) of the image and keep the edge details of the image better.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
本文编号：2279336
[Abstract]:The signal-to-noise ratio (SNR) of magnetic resonance images obtained by low field or fast imaging is often low. In clinical application, increasing the acquisition times and adding the average of multi-frame images are often used to improve the signal-to-noise ratio (SNR) of the images. However, in the process of multiple acquisition, the autonomous or involuntary movement of the human body causes the relative deviation between multiple images. Therefore, the edges or details of the image obtained by the coherent cumulative average will be blurred. In order to solve this problem, our previous algorithm is based on the non-local mean algorithm (Non-Local Means,NLM), using the local similarity of the image to calculate the local offset between images, and then doing the weighted average after the local offset correction of the image. In order to improve the signal-to-noise ratio. In this paper, a rotation-invariant nonlocal mean method (Rotation-invariant Non-local Means,RINLM) is proposed. In this method, the circular neighborhood region is used and divided into a series of concentric rings with the center pixel as the center, and the similarity between the neighborhood patterns is calculated. Compared with the NLM algorithm, the proposed method can improve the denoising performance by using the similar neighborhood pattern of relative rotation in the image. By applying the rotation-invariant nonlocal mean algorithm to the accumulation and denoising of image sequences, the method in this paper can overcome the influence of the rotational component of the local motion on the calculation, and thus better deal with the local motion with rotation. Further improve the image quality. In this paper, simulated data and clinical real data were used to evaluate and analyze the experimental results by subjective and objective methods. The results show that compared with the previous methods, the proposed method can further improve the signal-to-noise ratio (SNR) of the image and keep the edge details of the image better.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
【参考文献】
相关期刊论文 前2条
1 石光明;刘丹华;高大化;刘哲;林杰;王良君;;压缩感知理论及其研究进展[J];电子学报;2009年05期
2 喻玲娟;谢晓春;;压缩感知理论简介[J];电视技术;2008年12期
，本文编号：2279336
本文链接：https://www.wllwen.com/shoufeilunwen/xixikjs/2279336.html
