结合分块模糊熵和随机森林的图像分类方法

摘要

To improve the performance of image classification,a new image classification method is proposed.The basic idea is:by considering the uncertainty of the content of the image as a stochastic process,blocked fuzzy entropy is used to extract image features,and random forest method is used to execute feature classification.First,it divides an image into multiple image blocks by considering the global and local properties of the image.Then,it executes fuzzy c-means clustering on each image blocks,and extracts fuzzy entropy features.And then,it obtains a fuzzy entropy feature vector of an image after feature normalization.Finally,a classifier of random forest is built,and feature classification is realized for fuzzy entropy feature vectors.Experimental results show that the method has low error-classification rate,and less time-consuming of classification.%为提高图像分类性能,提出了一种图像分类方法.其基本思想是将图像内容的不确定性描述看作是一个随机过程,采用分块模糊熵来提取图像特征,采用随机森林方法进行特征分类.首先,考虑全局和局部特性,将图像划分为多个图像子块;然后,对每一个图像子块进行模糊c均值聚类,提取模糊熵特征;接着,通过归一化处理,得到图像的模糊熵特征向量;最后,构造随机森林分类器,实现模糊熵特征向量的分类.实验结果表明,该方法的错分率低,分类耗时少.