一种自适应改进曲率扩散超声图像滤波方法

Ultrasound Image Filtering Using Adaptive Modified Curvature Diffusion Equation

针对图像尤其是医学超声图像的边缘模糊或者缺失对图像进一步处理带来的影响,同时由于传统改进曲率扩散方程无法自适应及迭代时间长,提出一种结合高斯滤波的参数自适应的改进曲率扩散方法,并且使用统计学中的绝对偏差中值自动调整梯度阈值参数。该方法能在保留改进曲率扩散方程去噪能力的同时有效地保留图像边缘细节信息,同时减少滤波过程中的迭代次数,从时间复杂度的方面提高了方法效率。另外,自适应梯度阈值的采用进一步提高了改进曲率扩散方程保留图像细节信息的能力。实验结果表明,结合高斯滤波的参数自适应改进曲率扩散方法继承了改进曲率扩散的优点,同时又减少了人工干预,提高了去噪效率。

Blurred or missing edges in image,especially ultrasound image could affect further processing,and traditional modified curvature diffusion equation's parameters cannot be adaptive and iteration time is long. For these reasons,an adaptive modified curvature diffusion equation combined with Gaussian filter was proposed and median absolute deviation in robust statistics was used to adjust the gradient magnitude threshold parameter automatically. This method can preserve image details effectively while remain MCDE's capacity for removing noise,and lower the time of iterations during filtering which improves efficiency in terms of time complexity. Moreover,MCDE's ability of preserving image details is further enhanced by applying adaptive gradient magnitude threshold. Experiments proved that this method inherits the advantages of MCDE,reduces human intervention and improves the efficiency of denoising.