基于改进各向异性扩散冲击滤波器模型的图像去噪

A modified anisotropic diffusion with shock filter model for image denoising

针对各向异性扩散冲击滤波器(ADSF)在图像增强中对噪声敏感的问题,将梯度矢量流(GVF)引入到ADSF中,提出一种新的图像去噪方法 GVF-ADSF。在改进的GVF-ADSF方法中,通过引入曲率差来区分图像的特征区域,并定义一个加权系数来控制滤波器中2个扩散项在图像的边缘区域和平坦区域的扩散程度,使得图像区域之间能够自然的平滑过渡。通过实验对比本文方法与均值滤波、Perona and Mailk(PM)模型、ADSF模型的去噪性能,结果表明所提方法能很好地去除图像噪声并保留图像丰富的纹理细节,得到更高的信噪比。

To overcome the noise sensitivity problem of the anisotropic diffusion with shock filter（ADSF） in image enhancement, a novel image denoising method was presented, which incorporates gradient vector flow（GVF） into the ADSF model. In the modified ADSF method called GVF-ADSF, the curvature difference was employed to distinguish the image region characteristics, and a weighted coefficient is defined to control the diffusion level of the two filter diffusion terms between the edge area and the flat region. Hence, the image transition area can be naturally smoothed. The denoising performance of the GVF-ADSF method was compared with that of the mean filter, the Perona and Mailk（PM） model, and the conventional ADSF model. The experimental results indicate that the GVF-ADSF method can effectively remove the image noise and retain the image texture better. In addition, the proposed method can get higher the ratio of signal to noise.