高阶PDE模型中的松弛中值图像去噪方法

A relaxed median image denoising method in high order PDE model

针对平均曲率极小化模型在去噪过程中产生斑点的问题,提出了一种平均曲率和松弛中值滤波结合的迭代算法。首先,使用平均曲率模型对噪声图像处理,根据局部方差信息,利用阈值确定斑点的位置。其次,利用具一定边界保持性质的松弛中值滤波器消除斑点噪声。最后,为更有效地消除斑点,在每一次随着时间的迭代后都使用松弛中值滤波。对曲线和图像进行去噪仿真实验,结果表明,相对于平均曲率模型,本文算法在客观指标和主观视觉效果上均有更好的去噪效果和更低的时间复杂度。

To remove speckle effect produced by the mean curvature model, we propose an iterative image denosing algorithm, which combines the mean curvature model and the relaxed median filter. Firstly, we denoise the image by the mean curvature model. Then the speckle noise is located using a threshold according to the variance of the image. Finally the relaxed median filter is introduced. To remove the speckle noise more efficiently, the filter is applied at every iteration step. Denoising simulations on curves and images show that the proposed algorithm has better denoising effect in terms of objective indicators and subjective visual effect, as well as lower time complexity in comparison with the mean curvature model.