一种非参数回归图像去噪方法

A Method of Image Denoising in Non-parametric Regression

为了在图像去噪时进一步提高信噪比从而增强去噪效果,提出了一种空域自适应的高稳定性去噪滤波器。将数值分析中的非参数回归模型应用于图像去噪处理,推导了基于该方法的算法模型。在双边核的基础上,提出一种自适应的Steer核滤波方法,该算法根据边缘的方向可以调整模板的方向和曲率,做到了完全自适应。最后,在相同条件下实现该算法与Nadaraya-Waatson Kernel Regression算法、Bilateral Kernel Regression算法的Matlab实验比较,实验结果表明,基于非参数回归模型的去噪算法具有良好的去噪效果,与其他算法相比,可以得到更高的信噪比。

Involved in improving the effect of image denoising so as to heighten the SNR, an adaptive denoising filter in the spatial domain is introduced. By taking the model on non - parametric regression into image denoising, the arithmetic based on this method has been presented. Based on bilateral kernel, an adaptive Steer kernel filter method is introduced, which adjusts the direction and curvature of stencil by the direction of edge and realizes the absolute adaptivity. Based on the same condition, the experiments using Nadaraya -Waatson kernel regression method and bilateral kernel regression method are compared with this method by Matlab. The results show the adaptive kernel regression method gives better effect and gets higher SNR.