用于DR图像缺陷检测的改进的LBP算法

针对传统LBP（Local Binary Pattern）算法在DR图像缺陷检测中对噪声异常敏感而导致的缺陷识别率低的问题，在已有的韦伯LBP算法（Weber Local Binary Pattern，WLBP）的基础上，提出改进的WALBP（Weber Adapted Local Binary Patterns）算法。WALBP算法保留了WLBP算法最后生成二维直方图的特点，对其所用的LBP算子和LoG（Laplacian of Gaussian）方法进行了改进。WALBP算法更加有效地描述了DR图像的纹理特征，同时有效解决了WLBP算子在进行缺陷检测时直方图维数较多及分类能力不强的问题。通过对多幅铸件DR图像进行实验分析，结果表明，相对于已有的WLBP算法和传统的LBP算法，WALBP算法在缺陷检测上具有更高的识别率，在缺陷识别技术中具有很高的应用价值。

Improved LBP used for detection of defects of DR images

For traditional LBP（Local Binary Pattern） algorithm is sensitive to noise and lead to the lower recognition rate in DR image defect detection, an improved WALBP（Weber Adapted Local Binary Patterns） algorithm is proposed on the basis of the existing of Weber LBP（WLBP） algorithm. WALBP algorithm retains the characteristics which form the two-dimensional histogram, the LBP operator and LoG（Laplacian of Gaussian） method that used are improved. WALBP algorithm is more effective to describe the DR image texture feature, and effectively solves the WLBP（Weber Local Binary Pattern） operator in defect detection histogram more dimension classification and capability is not strong. Through analyzing casting DR image experiments, the results show that compared with the existing WLBP algorithm and the traditional LBP algorithm, WALBP algorithm has higher recognition rate on defect detection, in the defect recognition technology has a high application value.