计算机与现代化

• 图像处理 • 上一篇    下一篇

一种基于MAD的鲁棒性分形维数计算方法及图像识别应用

吴海燕,张陈陈   

  1. 深圳大学数学与计算科学学院,广东深圳518060
  • 收稿日期:2013-04-17 修回日期:1900-01-01 出版日期:2013-12-18 发布日期:2013-12-18

A MAD-based Robust Fractal Dimension Calculating Method and Its Application in Image Recognition

WU Hai-yan, ZHANG Chen-chen   

  1. College of Mathematics and Computational Science, Shenzhen University, Shenzhen 518060, China
  • Received:2013-04-17 Revised:1900-01-01 Online:2013-12-18 Published:2013-12-18

摘要: 传统鲁棒差分盒计数法(RDBC)已成功用于高斯噪声图像的分形维估计,但由于对椒盐噪声较敏感,因此不再适用于椒盐噪声图像的分形维估计和图像分类。本文提出一种基于中值绝对偏差(MAD)的分形维数计算方法(MAD-DBC)。该方法利用MAD进行差分盒计数,对椒盐噪声具有很好的鲁棒性特点。实验结果表明,利用小波多分辨率的DBC、RDBC和MAD-DBC对椒盐噪声的16种Brodatz纹理图像进行分类,MAD-DBC具有更高的识别率和更好的噪声鲁棒性。

关键词: 分形维, 差分盒计数法, MAD, 图像分类

Abstract: The traditional robust differential box-counting method (RDBC) has been successfully used for calculating fractal dimension of an image degraded by Gaussian noise. However, it is not suitable for estimating fractal dimension of salt & pepper noisy images and classifying those images. This paper presents a MAD-based method (MAD-DBC) for calculating fractal dimension of an image. The method uses MAD for differential box-counting, which is robust against salt & pepper noises. Classification experiments on Brodatz texture images show that, compared with DBC and RDBC, the MAD-DBC achieves higher classification rate and better noise robustness.

Key words: fractal dimension, differential box-counting, MAD, image classification