基于统计法的纹理矢量周期描述

Texture Victor Periodicity Description on the Basis of Statistics

本文提出了纹理矢量周期的描述方法:T{θ,Tθ,Vθ},分别从八个方向上(即0°,45°,90°,135°,180°,225°,270°,315°)讨论纹理的周期性、方向性以及周期成分所占的比例等方面。提出了分析纹理周期的d-θ分析方法,比较全面地实现了对纹理周期描述和分析。针对纺织布的纹理特征,经过大量实验,筛选出能反映周期方向的5个特征参量和周期大小的4个特征参量。最后,在本文提出的纹理周期矢量描述方法和分析方法基础上,采用二值共生矩阵及其纹理特征实现对纺织布的纹理周期描述。结果表明,对于含有一定周期的纹理图像,采用本文纹理周期描述和分析方法,实现了比较全面的描述。和直接采用灰度共生矩阵进行分析相比,本文分析方法极大降低了计算量。

This paper presents a description of the texture vector periodicity method：T{θ,Tθ,Vθ}.From eight directions （that is,0°,45°,90°,135°,180°,225 °,270°,315°） to discuss the periodicity of texture,direction,and the proportion of periodicity components,etc.The d-θ analysis method is proposed to achieve a more comprehensive description and analysis of the texture periodicity.For fabric textures,after a large number of experiments,we select five textural features to reflect the direction and reflect the periodicity value of the four textural features.Finally,Based on the vector periodicity description methods in this article,we use the binarized co-occurrence matrix（BCM＇s） and textural features to describe periodicity.The experimental results show that a certain periodicity texture image,using the periodicity described and the analysis method in this paper,achieves a more comprehensive description.Compared to the method of directly using gray level co-occurrence matrix（GLCM＇s） for analysis,our analysis method greatly reduces the computational complexity.