基于频带增强和像素能量的纸病检测

Paper Defect Detection Based on Band Enhancement and Pixel Energy

关键词:  频带增强  像素  方向  能量  纸病

Key Words:band enhancement  pixel  direction  energy  paper defect

摘要:采用频带增强和像素能量算法以提高纸病检测的效果。首先非下采样Contourlet变换把彩色纸划分为不同频带，采用不同方法增强低频子带、高频子带；然后建立彩色纸纹理检测窗口与纹理在经度、纬度方向上的自相关函数关系，低频子带需要大的检测窗口以获得图像的轮廓，高频子带需要小的检测窗口以获得图像的细节信息；然后通过纹理像素在水平分量、垂直分量、对角线分量的能量异常获得纸病，纸病的分割阈值为各个方向分量最大值的均方值；最后给出了算法流程。结果表明，该算法对彩色纸表面的纸病检测清晰，混合纸病检测准确率评价指标高于其他算法。

Abstract:In order to improve the effect of paper defect detection， band enhancement and pixel energy algorithm are adopted. Firstly， the color paper was divided into different bands using Non-subsampled Contourlet Transform， and different enhancement methods were adopted for low-frequency sub-bands and high-frequency sub-bands. Secondly， the relationship between the texture detection window of color paper and the autocorrelation function in longitude and latitude direction was established， low-frequency band needed a large detection window to obtain the image contour， and the high-frequency band needed a small detection window to obtain the image detail information. Thirdly， paper defect was obtained by the energy anomaly of texture pixels in horizontal component， vertical component， and diagonal component， and the mean square value of the maximum value of each direction component was used as the segmentation threshold of paper defect. Finally， the algorithm process was given. The results showed that the algorithm could clearly detect paper defects on the surface of colored paper， and the evaluation index of the accuracy of mixed paper defect detection was higher than other algorithms.