基于局部特征多轴旋转不变特性的纹理图像分类算法

Classification Algorithm for Texture Image Based on Local Characteristics of N-Fold Rotation Invariant Feature

采用非量化的局部特征设计出一个稳健的纹理描述符,以便增强旋转和尺度变化时纹理分类的鲁棒性。首先,引入了局部特征的旋转对称性的概念,提出了一种新颖的局部特征来描述纹理的旋转不变特性。为了处理剧烈的旋转、尺度等变化,利用费舍尔向量编码方法对纹理特征量进行多尺度分析,在不增加局部特征维度的同时又能结合尺度信息,由此产生的局部特征对旋转、灰度变化都有较强的鲁棒性。实验结果表明,所提方法的评估结果在许多数据集上都远远超过了现有最优算法,大大提高了纹理分类的精度。

This paper adopted a non-quantifiable local feature to design a robust texture descriptor,so as to enhance the robustness of the texture classification in the rotation and scale changes.First of all,the concept of local feature with rotational symmetry is introduced.It is found that many rotation invariant local features are rotational symmetric to a certain degree.Therefore,this paper proposed a novel local feature to describe the rotation invariant properties of the texture.In order to deal with the change of rotation and scale in texture image,Fisher vector encoding method is used to manage multiscale analysis for the texture feature,which can combine with the scale information without increasing the dimension of the local feature.The resulting local features have strong robustness to rotation and gray intensity variation.Experimental results show that the proposed method outperforms the existing algorithms on many data sets,greatly improving the texture classification accuracy.