Solid Surface Scratch Detection with Multi-scale Wavelet Representation

摘要：

This paper proposed a set of image processing algorithms that will automatically detect scratches on a solid surface. This type of surface defect to be detected has random directions, inconspicuous gray levels and strong background noise. The method that is used to extract texture features is based on multi-scale wavelet representation (MSWAR), which has good shift invariance and remains at full resolution at every scale. We then use a controlled edge fusion model to combine the detailed wavelet coefficients (horizontal, vertical and diagonal). With the fused detail map, we apply multivariate statistics to synthesize multiple image features. Whenever the value of a region is deviated from normal texture region, defect is detected. The experimental results show that the presented algorithm can efficiently suppress the energy in the background and detect scratches accurately.

关键词： Surface defect detection wavelet decomposition multivariate statistical analysis

作者: Li Yao Ming Yao Yan Wan Bugao Xu

作者单位: School of Computer Science and Technology Donghua University Shanghai, China Department of Human Ecology University of Texas Austin, TX78712, U.S.A.

会议类型: 国际会议

会议名称: 2011 4th International Congress on Image and Signal Processing(第四届图像与信号处理国际学术会议 CISP 2011)

会议地点: 上海

会议语种:英文

页码: 1579-1586

在线出版日期: 2011-10-15（万方平台首次上网日期，不代表论文的发表时间）