基于FABEMD的纹理图像分类算法  

Texture image classification algorithm based on FABEMD

作　　者：胡明娣 臧艺迪 徐家慧 HU Mingdi;ZANG Yidi;XU Jiahui(School of Communication and Information Engineering,Xi＇an University of Posts and Telecommunications,Xi＇an 710121,China)

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121

出　　处：《西安邮电大学学报》2018年第1期53-58,共6页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金资助项目(61502386);陕西省教育厅科学研究计划资助项目(2013JK1074)

摘　　要：针对二维经验模式分解算法应用于纹理图像分类中存在运算时间长、准确率低的问题,将快速自适应二维经验模式分解(fast and adaptive bidimensional empirical mode decomposition,FABEMD)方法应用到纹理图像分类中。该方法首先将纹理图像分解成3个二维固有模态函数(bidimensional intrinsic mode function,BIMF)和1个余量;其次使用灰度共生矩阵提取各BIMF的能量、熵、对比度和相关性这4个纹理特征参数,组成特征向量,最后采用最小距离分类器进行纹理图像分类。实验采用Brodatz纹理图像库,选取10幅纹理图像作为样本图像,仿真实验结果表明,与二维经验模式分解方法相比,所提的算法查准率为86.28%,同时缩短了运算时间。Due to the problem that the two-dimensional empirical mode decomposition algorithm has a long computing time and low accuracy in texture image classification,a fast and adaptive bidimensional empirical mode decomposition（FABEMD）algorithm is proposed.In this algorithm,firstly,the texture image is decomposed into three bidimensional intrinsic mode functions（BIMFs）and a residual.Secondly,the gray-level co-occurrence matrix is used to extract the energy,entropy and contrast of each BIMF Relevance of the four texture feature parameters,which are constituted as an eigenvector.Finally,the minimum distance classifier is used to classify the texture image.Brodatz texture image database is used in the experiment,and ten texture images are selected as sample images.Simulation results show that compared with the bidimensional empirical mode decomposition method,the proposed algorithm has a precision of86.28% and can shorten the computation time.

关 键 词：纹理图像分类 快速自适应二维经验模式分解 灰度共生矩阵 最小距离分类器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]