基于L0范数视觉显著性的织物疵点检测算法研究

A Novel Fabric Defect Detection Algorithm Using L0 Norm Visual Saliency

提出了一种基于L0范数视觉显著性的织物疵点检测算法。首先将测试图像分块,针对每一个测试图像块,用随机选择的K个其他图像块与该图像块变异构建字典库,并利用该字典库对测试图像块进行稀疏表示。然后采用L0范数优化方法来求解稀疏系数,将变异图像块对应的稀疏系数作为该图像块的显著度,从而生成最终视觉显著度图。最后通过迭代最优阈值分割算法定位出疵点区域。该算法相对已有方法能更有效地检测出疵点区域。

This paper proposes a novel fabric defect detection algorithm using sparse representation-based visual saliency. In the proposed scheme, an input image is first divided into blocks, then each image block is represented based on a dictionary constructed using some randomly selected blocks and the dithered test block, using L0-rninimization solves the equation to obtain weights. Based on the corresponding weight value of dithered test block, a saliency map is produced. Finally, saliency map is segmented by using an optimal threshold, which is obtained by an iterative approach. Experimental results demonstrate that generated saliency map using our proposed method outperforms state-of-the art, and the defect can be efficiently localized by the optimum threshold segmentation.