一种基于MRI图像的相干增强张量扩散方程离散方案

An MRI image-based scheme for the coherence-enhancing diffusion filtering

相干增强张量扩散方程的离散方案对图像的去噪效果尤为重要。在保持空间导数算子旋转不变性的前提下,根据大脑MRI图像的结构特征,并考虑MRI图像的噪声服从莱斯(Rician)分布的特点,以峰值信噪比(PSNR)和结构相似性(MSSIM)为目标函数对空间导数算子的取值进行优化。实验结果表明,提出的离散方案在峰值信噪比和结构相似性方面均优于其他旋转不变性离散方案。

The coherence-enhancing diffusion filtering is crucial to image denoising. Taking PSNR and MSSIM as the objective functions, we carried out an experiment on the numerical optimization of the space derivatives in the preconditions that the rotation of the derivatives remained constant and the MRI image followed Rician distribution and the results showed that the diffusion filtering proposed in this paper outperforms other schemes of rotation invariance in terms of PSNR and MSSIM.