基于方向自适应相对高斯引导的图像纹理滤波  

Texture filtering based on direction adaptive RoG guidance

作　　者：韩雨轩 高银 李琦铭 李俊 HAN Yu-xuan;GAO Yin;LI Qi-ming;LI Jun(Advanced Manufacturing College,Fuzhou University,Fuzhou 350116,China;College of Computer Science and Technology,University of Chinese Academy of Sciences,Beijing 100049,China;Quanzhou Institute of Equipment Manufacturing,Fujian Institute of Research on the Structure of Matter,Chinese Academy of Sciences,Quanzhou 362200,China)

机构地区：[1]福州大学先进制造学院,福建福州350116 [2]中国科学院大学计算机科学与技术学院,北京100049 [3]中国科学院福建物质结构研究所泉州装备制造研究中心,福建泉州362200

出　　处：《计算机工程与设计》2024年第10期3002-3009,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(62001452);中国光电子信息福建科技创新实验室(闽都创新实验室)基金项目(2021ZZ116);泉州市科技基金项目(2020C07)。

摘　　要：针对联合双边滤波具有固定空间核尺度,不能随局部特征灵活变化的问题,提出一种基于方向自适应相对高斯算法的结构保持双边纹理滤波算法(DARG)。通过检测方向场中8个方向的相似像素构造自适应相对高斯引导映射图,引导联合双边滤波过滤纹理。大量实验结果表明,DARG算法可以很好地过滤复杂及大比例的纹理,以优于目前经典算法的方式保留高质量的小结构,简单有效,易于应用。A structure-preserving bilateral texture filtering algorithm(DARG)based on the directional adaptive relative Gauss algorithm was proposed to solve the problem that the joint Bilateral filter has a fixed spatial kernel scale and cannot change flexibly with local features.An adaptive relative Gaussian guidance map was constructed by detecting similar pixels in 8 directions in the direction field and then a joint bilateral filter was guided to filter the texture.A large number of experimental results show that the DARG algorithm can effectively filter complex and large-scale textures while retaining high-quality small structures in a way superior to current classic algorithms.It is very simple,effective,and easy to apply.

关 键 词：纹理过滤 保边平滑 结构保持图像平滑 自适应相对高斯 联合双边滤波 引导映射 核尺度自适应 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]