计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (19): 143-146.

• 图形图像处理 • 上一篇    下一篇

基于轮廓波变换的织物疵点图像消噪新方法

槐向兵,厉征鑫,刘建立,高卫东   

  1. 江南大学 纺织服装学院,江苏 无锡 214122
  • 出版日期:2014-10-01 发布日期:2014-09-29

New fabric defect image denoising method based on contourlet transform

HUAI Xiangbing, LI Zhengxin, LIU Jianli, GAO Weidong   

  1. School of Textile and Clothing, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2014-10-01 Published:2014-09-29

摘要: 为了实现织物疵点图像的有效消噪,使其更有利于特征提取和疵点检测,提出了基于轮廓波变换的织物疵点图像消噪新方法。综合考虑轮廓波方向子带能量的大小与织物疵点图像轮廓细节之间的关系,对Donoho多尺度分解阈值进行修正,改进了Donoho多尺度分解阈值对图像细节“过扼杀”的缺点。实验结果表明,对织物疵点图像进行基于轮廓波变换改进阈值消噪时,该方法更好地保留了织物疵点图像的轮廓细节,峰值信噪比显著提高。采用改进的轮廓波Donoho多尺度分解阈值消噪后的图像,可以更好地应用于织物疵点图像的特征提取和疵点识别。

关键词: 消噪, 小波变换, 轮廓波变换, 织物疵点, 峰值信噪比(PSNR)

Abstract: In order to achieve effective denoising of fabric defect images, making them more conducive to defect detection and feature extraction, a new fabric defect image denoising method based on contourlet transform is proposed. Considering the relationship between directional sub-bands energy of contourlet and the outline details of fabric defect images, the Donoho multi-scale decomposition threshold is corrected to improve its shortcoming that “over kill” to image detail. Experimental results show that the fabric defect images retain better outline details and the peak signal to noise ratio is improved significantly when contourlet transform denoising based on improved threshold is used. The fabric defect images which are denoised by improved Donoho multi-scale decomposition threshold can be better applied to feature extraction of fabric defect images and defect detection.

Key words: denoising, wavelet transform, contourlet transform, fabric defect, Peak Signal Noise Ratio(PSNR)