计算机工程与应用 ›› 2013, Vol. 49 ›› Issue (4): 206-208.

• 图形图像处理 • 上一篇    下一篇

一种加权的多重分形特征提取算法

谢雅婷1,梁光明2,石跃祥1,柳佳雯1,丁建文3   

  1. 1.湘潭大学 信息工程学院,湖南 湘潭 411105
    2.国防科技大学 电子科学与工程学院,长沙 410075
    3.长沙爱威科技,长沙 410013
  • 出版日期:2013-02-15 发布日期:2013-02-18

Weighted multi-fractal algorithm for feature extraction

XIE Yating1, LIANG Guangming2, SHI Yuexiang1, LIU Jiawen1, DING Jianwen3   

  1. 1.College of Informationr Engineering, Xiangtan University, Xiangtan, Hunan 411105, China
    2.College of  Electronic Science and Engineering, National University of Defense Technology, Changsha 410075, China
    3.Ai Wei Technology of Changsha, Changsha 410013, China
  • Online:2013-02-15 Published:2013-02-18

摘要: 为解决多重分形维数不能够很好地反映图像强度信息和对图像尺度有强依赖的问题,在研究q 阶广义维数D(q)基础上,提出两种改进方法。通过分析影响生长概率的因子,提出一种结合强度信息的加权子数计算方法,提出一种基于网格强度与均值的二维分形维数计算方法。实验表明改进的多重分形算法提了特征区分度,计算特征更加鲁棒和有效,将改进方法用于血细胞识别系统,改善了白细胞分类准确性。

关键词: 特征提取, 多重分形, 网格, 生长概率

Abstract: In order to solve the drawbacks of the multi-fractal dimension can not be a good reflection of the image intensity information and highly dependes on the image scale, this paper presents two improvement based on q-order moments of general dimension theory, marked D(q). This paper proposes a new weighted calculation of boxes number method conbined intensity information by analyzing the factors affecting the probability of growth, then proposes a two-dimensional method of calculating fractal dimension based on the gride intensity and mean. Experiment shows that the new method improves features differentation, computes features more robust and more effective, and improves the classification accuracy by putting the new method in the identification systerm of blood cells.

Key words: feature extraction, malti-fractal, gride, growth proability