用于图像恢复的基于SVR的自适应新滤波器的研究*

提出了一种用SVR回归器识别脉冲噪声的思想，并将其应用于图像滤波和恢复，形成了用于对脉冲噪声进行滤波的SVR自适应滤波器。这种滤波器在滤波时，先用SVR对待识别像素作噪声识别，再对含噪声的像素作中值滤波。用SVR作噪声识别时，先对滤波窗口作SVR回归，通过待识别像素回归距的大小判断其是否含有噪声。在进行SVR回归时，使用鲁棒的Huber损失函数。由于更充分地利用了待识别像素点的局部背景信息，这种滤波器提高了脉冲噪声识别的正确率。实验表明，在保留原图像的细节信息方面，其滤波效果要优于基于SVC的中值滤波器。

Novel Adaptive SVR Based on Filter for Image Restoration

In this paper, a novel adaptive filter based on support vector regression is proposed, which can effectively be used to suppress impulse noise in images. The main idea of our filter is to use SVR impulse detector to judge whether an input pixel is noised. If it is noised, we remove its noise by a median filter. When detecting an input pixel, we regress the filer window of an input pixel using SVR, then judge the input pixel by its regression distance. In SVR regression, we use the Huber loss function due to its excellent robustness. Compared with the latest SVC approach, our filter can preserve more image details while effectively suppressing impulse noise for image restoration. Experimental results indicates the success of our filter.