Abstract: In order to improve the accuracy for detecting the density of the printed woven fabric, image repainting technology was utilized to repair textile local texture by filtering surface patterns. To verify the effects of pattern types and sizes on restoration under different organizational structures, 27 test samples were designed by pattern fusion according to the weave, pattern types and sizes. An optimized Criminisi algorithm was proposed. The priority item was redesigned, and the size of the repair module underwent self-adaptive adjustment according to the pattern size. In the experiment, the restoration effects by means of Criminisi algorithm, optimizing priority item only, and optimizing both priority and module items were compared and analyzed. Besides, the densities detected by manual method and by the algorithm in the paper were compared and analyzed. The result shows that the proposed method can effectively repair the fabric structure texture in the area of patterns, and the mean subjective and objective consistency is 99.89% for test specimens. This proposed method can reduce the interference of small printed patterns, which can be helpful to expand the application scope of image automatic inspection for fabric density.

Key words: fabric density, automatic detection, texture, image inpainting, wavelet transform

摘要： 为提升印花机织物经纬密度自动测量的准确性,利用图像修复技术滤除织物表面印花图案,复原局部组织纹理。为验证不同组织结构下图案种类、大小对复原效果的影响,采用图像融合方式,按照组织结构、图案种类、图案大小设计了27种测试样本。对Criminisi算法优先权项进行设计,并对修复模块尺寸依据图案尺寸作自适应调节。对比分析了Criminisi算法、仅优化优先权项,以及优先权与修复模块联合优化3种修复效果;同时对比分析了人工检测和本文算法检测的样本密度。结果表明:优化算法能够较好地复原图案所在区域内织物组织纹理,样本测试主客观一致率均值为99.89%。该方法的提出,能够解决部分中小尺度印花图案的干扰,进一步扩大了织物密度图像法检测的适用面。

关键词: 织物密度, 自动检测, 纹理, 图像修复, 小波变换