摘要
传统的滤波算法在重噪声区域易出现噪声抑制能力下降和边缘光晕效应,且在滤波过程中矩形滤波模板的两边总是分别平行于离散图像的坐标轴,必然导致模板跨边缘平滑滤波。为解决这些问题,提出一种窗口形状自适应的双边滤波算法。采用边缘梯度图作为引导图像,通过每个像素的梯度方向确定该像素处的边缘方向。然后使模板窗口的两边分别平行或垂直于当前像素处的局部边缘方向,实现模板“沿边缘”平滑滤波,在去除噪声的同时更好地保持边缘。实验结果证明,该改进算法滤波性能和保边能力较好,PSNR平均值提升13%左右,SSIM平均值提升4.6%左右,且对受到高强度的噪声污染的图像具有较好的噪声抑制和边缘保持能力。
Traditional filtering algorithms are prone to noise suppression reduction and edge halo effects in heavy noise areas.In the filtering process,the two sides of the rectangular filter template are always parallel to the coordinate axis of the discrete image,which inevitably leads to the cross-edge smoothing filtering.To solve these problems,this paper presents a window shape adaptive bilateral filtering algorithm.We used an edge gradient map as a guided image,and determined the edge direction of each pixel by the gradient direction of each pixel.Then,the two sides of the template window were parallel or perpendicular to the local edge direction of the current pixel,so that the template“along the edge”smoothing filter could be realized,and the edge could be better kept while the noise was removed.The experimental results show that the improved algorithm has better filtering performance and edge-preserving ability.The average PSNR and SSIM are increased by about 13%and 4.6%respectively,and it has better noise suppression and edge-preserving ability for images polluted by high-intensity noise.
作者
喻恒
胡海峰
耿则勋
Yu Heng;Hu Haifeng;Geng Zexun(College of Information Engineering,Pingdingshan University,Pingdingshan 467000,Henan,China)
出处
《计算机应用与软件》
北大核心
2019年第12期201-208,213,共9页
Computer Applications and Software
基金
河南省科技厅科技攻关项目(201700812)
关键词
双边滤波
梯度引导
窗口形状自适应
边缘保持
Bilateral filtering
Gradient guidance
Window shape adaptive
Edge preservation
