Quality control in silicon steel manufacturing process is a crucial step. The application of image processing techniques is very useful in steel inspection and manufacturing. It has established to be the most reliable and promising solution for the development of an automatic defect detection. Since the surface of the silicon steel strip has a cluttered background and defects with small sizes, flaws detection becomes a complex task. In this paper a novel rapid algorithm based on anisotropic diffusion and saliency map is proposed for detection of defects in images of hot rolled silicon steel. The algorithm first adopted a saliency map to enhance defects. Then the computed saliency map was employed in the anisotropic diffusion coefficient function as an orientation guide of the diffusion flow. The aim behind using salient feature is that a small defect can frequently attract attention of human eyes which permits to identify defects in high textured image. Finally, the defects were extracted using a local threshold operator. To verify the validity of the proposed algorithm, extensive experiments were realized on an image database of silicon steel strip then a comparison with traditional diffusion algorithms was given. Experimental results show that this method achieves accuracy and outperforms traditional methods in terms of accuracy and robustness.

硅钢制造过程中的质量控制是至关重要的一步。图像处理技术的应用在钢铁检测和制造中非常有用。它已被确定为开发自动缺陷检测的最可靠，最有前途的解决方案。由于硅钢带的表面具有杂乱的背景和小尺寸的缺陷，因此缺陷检测成为一项复杂的任务。提出了一种基于各向异性扩散和显着性图的快速算法，用于热轧硅钢图像的缺陷检测。该算法首先采用显着图来增强缺陷。然后，将计算出的显着性图用于各向异性扩散系数函数中，作为扩散流的定向向导。使用显着特征的目的在于，小的缺陷可以经常引起人眼的注意，从而可以识别出高纹理图像中的缺陷。最后，使用局部阈值运算符提取缺陷。为了验证该算法的有效性，在硅钢带图像数据库上进行了广泛的实验，然后与传统扩散算法进行了比较。实验结果表明，该方法在准确性和鲁棒性方面都达到了优于传统方法的水平。在硅钢带的图像数据库上进行了广泛的实验，然后与传统扩散算法进行了比较。实验结果表明，该方法在准确性和鲁棒性方面均达到了优于传统方法的水平。在硅钢带的图像数据库上进行了广泛的实验，然后与传统扩散算法进行了比较。实验结果表明，该方法在准确性和鲁棒性方面均达到了优于传统方法的水平。