优化贝叶斯非局部均值算法的超声图像去噪方法

Ultrasound image denoising method based on optimal Bayesian non-local mean

为了去除超声成像中可能产生的斑点噪声,提出了一种基于优化贝叶斯非局部均值算法的超声图像去噪方法,重点探讨优化算法改进的两个方面:(1)针对原始非局部均值算法运算复杂度高这一问题,应用像素预选择算法优化去噪时间;(2)针对原始算法只适用于加性高斯噪声的问题,利用伽马分布噪声模型,通过贝叶斯公式推导出适用于超声斑点噪声的非局部均值算法,引出皮尔逊距离来计算图像块之间距离,并且通过多次实验,得出噪声标准差和滤波参数的比值关系,实现参数自适应。为了评估优化算法的去噪能力,在模拟超声图像和真实超声图像上分别进行实验验证,结果表明,该优化算法去噪效果更好,图像边缘和细节保持能力更强;且在运行效率方面,较原始非局部均值算法有较大提升。

In order to overcome the speckle noise that may be generated in ultrasound imaging,an optimized Bayesian non-local mean algorithm is proposed.This paper discusses the improvement of the optimized algorithm from the following two aspects:(1)Aiming at the problem of high computational complexity of the non-local means(NLM)algorithm,the pixel pre-selection algorithm is applied optimize the denoising time of the algorithm;(2)Aiming at the problem that the NLM algorithm is only applicable to additive Gaussian noise,the Gamma distribution noise model is used to derive a non-local mean algorithm suitable for ultrasonic speckle noise through the Bayesian formula,and the Pearson distance is derived to calculate the distance between image blocks;and through multiple experiments,the ratio relationship between the noise standard deviation and the filtering parameters is obtained,and the parameters are self-adapted;finally,the algorithm is realized through the Matlab platform.In order to evaluate the denoising ability of the optimal algorithm,experiments were carried out on simulated ultrasound images and real ultrasound images.The results show that the optimal algorithm has a better denoising effect and a stronger ability to preserve image edges and details;and in terms of operating efficiency,compared with the original non-local mean algorithm,it has a greater improvement.