基于非下采样Contourlet变换的织物疵点检测方法  

Detection Method of Fabric Defects Based on Non-subsampled Contourlet Transformation

作　　者：张莹莹[1] 韩润萍[1] 

机构地区：[1]北京服装学院信息工程学院,北京100029

出　　处：《丝绸》2014年第5期38-42,共5页Journal of Silk

基　　金：北京市教委科技创新平台项目(PXM2014_014216_000059)

摘　　要：提出了一种基于非下采样Contourlet变换的织物疵点检测新方法。首先对织物疵点图像进行非下采样Contourlet变换(NSCT)得到一个低频子带和多个高频子带;然后通过代价函数在高频子带中挑选最优子带并阈值化,同时对低频子带采用非线性增益函数进行增强及阈值化处理;最后将经上述处理后的低频子带与高频子带进行融合与分割,以二值图像的方式从织物背景中提取出疵点。对比实验结果表明,该方法具有较高的检出率及良好的适应性。This paper proposes a new method to detect fabric defects based on non-subsampled Contourlet transformation. Firstly,the non-subsampled Contourlet transformation（ NSCT） is conducted for fabric defects image to obtain one low-frequency sub-band and muleiple high-frequency sub-bands. Secondly,the optimal sub-bands are selected from these high-frequency sub-bands by using a cost function and then thresholded. Meanwhile,the lowfrequency sub-band is enhanced through a nonlinear gain function and then thresholded. Finally,these low-frequency sub-bands and high-frequency sub-bands processed are fused and then divided. The defects are extracted from fabric background in the way of binary image. The comparison of experimental results shows that this new method has higher detection rate and good adaptability.

关 键 词：织物疵点检测 非下采样CONTOURLET变换 非线性增益函数 图像分割 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]