基于各向异性高斯滤波器的改进引导滤波算法

Improved guided filtering algorithm based on anisotropic Gaussian filter

针对传统引导滤波中矩形均值加权窗口会引入过多干扰,导致图像边缘两侧模糊以及产生光晕伪影的不足,提出一种基于各向异性高斯滤波器的改进引导滤波算法。将各向异性高斯滤波器融入传统引导滤波算法,利用其较强的边缘分辨力更准确地识别图像的边缘结构。同时,对窗口内的像素信息进行各向异性加权,使其具有更强的边缘保持性。实验结果表明,改进算法比传统引导滤波算法精细边缘的保持性更好,并在图像平滑、图像增强和图像亮度调整等图像处理应用中的处理结果更优,在有效保留图像边缘结构的同时减少了边缘伪影和边缘模糊。

In view of the shortcomings in traditional guided filtering that the rectangular mean weighted window will introduce too much interference,resulting in halo artifacts and blur on both sides of the image edge,an improved guided filtering algorithm based on anisotropic Gaussian filter is proposed.The algorithm integrates the anisotropic Gaussian filter into the traditional guided filtering algorithm,which can more accurately identify the edge structure of the image with its edge resolution.The pixel information within the window is weighted and summed by anisotropy,so as to have stronger edge retention.The experimental results show that the improved algorithm has better retention of fine edges than the traditional guided filtering algorithm,and has better processing results in image processing applications such as image smoothing,image enhancement and image brightness adjustment.It can effectively preserve the image edge structure and reduce edge artifacts and edge blur.