基于图像块l0梯度最小化的边缘保持平滑算法
宋 昱1, 2, 3,孙文赟1, 2, 3

1)深圳大学电子与信息工程学院,广东深圳 518060; 2)深圳大学深圳市媒体信息内容安全重点实验室,广东深圳 518060; 3)深圳大学广东省智能信号处理重点实验室,广东深圳 518060

信号与信息处理; 边缘保持图像平滑; l0梯度最小化; 图像块; 局部统计特性

Edge-preserving smoothing algorithm based on l0 gradient minimization of image-patch
SONG Yu1, 2, 3 and SUN Wenyun1, 2, 3

1)College of Electronics and Information Engineering, Shenzhen University, Shenzhen 518060, Guangdong Province, P.R.China;2)Shenzhen Key Laboratory of Media Security, Shenzhen University, Shenzhen 518060, Guangdong Province, P.R.China;3)Guangdong Key Laboratory of Intelligent Information Processing, Shenzhen University, Shenzhen 518060, Guangdong Province, P.R.China

signal and information processing; edge-preserving image smoothing; l0 gradient minimization; image patch; local statistics characteristic

DOI: 10.3724/SP.J.1249.2021.03307

备注

l0梯度最小化图像平滑算法可在保持边缘的同时滤除纹理和细节,但该算法使用图像梯度判决被平滑成分时会出现包含较小图像梯度(弱边缘)的区域会被平滑,而包含较大图像梯度(强纹理)的区域被保留的现象.为克服此缺陷,提出一种基于图像块l0梯度最小化算法(image-patch based l0 gradient minimization algorithm, 简称IP-l0算法)的图像平滑算法,通过对输入图像中的图像块而非整幅图像进行平滑,动态改变图像块目标函数中的权重参数,令主要包含强纹理的图像块以较大的力度进行平滑,而主要包含弱边缘的图像块以较小的力度进行平滑,再整合平滑后的图像块得到整个边缘保持平滑图像.对IP-l0算法、原始的l0梯度最小化算法、基于局部拉普拉斯滤波器的算法、基于相对全变差算法、基于树滤波的算法,以及2种基于深度学习的边缘保持算法进行仿真实验,结果表明,使用IP-l0算法滤波后的图像能在保持较弱的边缘的同时平滑强纹理.
Image smoothing algorithm based on l0 gradient minimization can smooth details and textures of image while preserving edges. Since the algorithm uses image gradient to determine the smoothed component, the region with smaller image gradients(weak edge)can be smoothed. However, the region with larger image gradients(strong edge)will be preserved. In order to overcome this drawback, we propose an image-patch based l0 gradient minimization image smoothing algorithm(IP-l0). Instead of globally smoothing the input image, our algorithm smoothed each image patch first and then combined the smoothed patches together to obtain the final smoothed image. The weight parameters in the objective function used to smooth each image patch are dynamically changed according to the local statistics of image patch, so that the patches containing strong textures will be smoothed with greater force, and vice versa. The experimental results show that compared to the original l0 gradient minimization algorithm and several other state-of-the-art edge-preserving image smoothing algorithms including the local Laplacian filter based algorithm, the relative total variation based algorithm, the tree filter based algorithm, and two kinds of deep-learning based smoothing algorithms, the proposed algorithm can effectively smooth strong textures and well preserve weak edges or structures, and the ability of edge preservation and texture smoothing is better than other algorithms.
·