作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 图形图像处理 • 上一篇    下一篇

基于Mean Shift的改进型图像滤波算法

林俊杰1,蒋新华1,2,胡蓉2,郑积仕2   

  1. (1.福州大学电气工程与自动化学院,福州 350000; 2.福建工程学院信息科学与工程学院,福州 350000)
  • 收稿日期:2015-01-13 出版日期:2016-01-15 发布日期:2016-01-15
  • 作者简介:林俊杰(1989-),男,硕士研究生,主研方向为图形图像处理、机器视觉;蒋新华,教授、博士生导师;胡蓉,副教授、博士;郑积仕,副教授、博士研究生。
  • 基金资助:
    “十二五”国家科技支撑计划基金资助项目(2015BAF24B00);福建省科技重大专项基金资助项目(2014HZ0004-3);福建省高校产学合作科技基金资助重大项目(2014H6006);福建工程学院校内基金资助项目(GY-Z13004);福建工程学院科研启动基金资助项目(GY-Z13103)。

Improved Image Filtering Algorithm Based on Mean Shift

LIN Junjie 1,JIANG Xinhua 1,2,HU Rong 2,ZHENG Jishi 2   

  1. (1.College of Electrical Engineering and Automation,Fuzhou University,Fuzhou 350000,China; 2.College of Information Science and Engineering,Fujian University of Technology,Fuzhou 350000,China)
  • Received:2015-01-13 Online:2016-01-15 Published:2016-01-15

摘要:

利用传统均值漂移(MS)算法进行图像滤波时,会将噪点拟定为图像边缘,使得图像内的噪点被保留。针对该问题,设计一种新型简化双边滤波方法(ANBF),并结合MS算法,提出一种改进型MS图像滤波算法。ANBF算法采用核函数先对噪点进行查找,形成确定的噪点标识图,通过标识图定位噪点,进行噪点滤波。MS算法采用核函数实现图像模点查找,依据模点进行图像平滑滤波。将ANBF算法嵌入传统MS算法中,以改进MS算法噪点消除能力。实验结果表明,与传统MS图像滤波算法相比,改进算法能有效克服MS算法噪点消除能力弱的问题,具有较高的噪点消除率。

关键词: 均值漂移, 核函数, 双边滤波, 噪点消除, 噪点查找

Abstract: The image filtering experiment shows that the traditional Mean Shift(MS) algorithm regards noise pixel as image edge,so that the noise pixel is retained.Aiming at this problem,the paper designs a new simplified bilateral filtering method(ANBF),which is combined with the MS algorithm,and proposes an improved MS algorithm for image filtering.Kernel function is used firstly in the ANBF to find the noise and generate marking graph.Then ANBF locates and eliminates the noise according to marking graph.The MS algorithm uses the kernel function to find image mold,and then eliminates the noise according to image mold.The ANBF algorithm is embedded in traditional MS algorithm to improve MS algorithm for noise elimination capacity.Experimental results show that compared with the traditional MS image filtering algorithm,this algorithm can effectively overcome the MS algorithm’s weakly eliminating noise ability,and has higher noise eliminating rate.

Key words: Mean Shift(MS), kernel function, bilateral filtering, noise eliminating, noise searching

中图分类号: