Abstract: To begin with an important visual feature of texture, such as periodicity, a new approach is proposed for fabric defect detection based on texture periodicity analysis according to feature of fabric texture image. By different fabric defect images detection experiments, this approach is proved to enjoy the good features of reliability and validity to detect fabric defects and to be able to detect more kinds of fabric defect.

Key words: texture periodicity, auto-correlation function, texture primitive, fabric defect detection

摘要： 根据织物图像纹理自身特点，从图像纹理的周期性这个重要的视觉特征入手，提出了基于纹理周期性分析的织物疵点检测方法。通过对大量不同疵点图像检测实验，证明提出方法对织物疵点检测具有较好的有效性和可靠性，而且具有检测的疵点种类多、实用性好的特点。

关键词: 纹理周期性, 自相关函数, 纹理基元, 织物疵点检测