基于二维属性直方图的铸件X射线图像的缺陷提取

Defect Extraction of X-Ray Images of Castings Based on Two-Dimension Bound Histogram

针对铸件X射线图像对比度低、边缘模糊、噪声多等特点,提出了基于二维属性直方图的最大相关准则图像分割算法。首先利用最大熵法得到的阈值构造图像的属性集,然后由原始图像及其属性集构造二维属性直方图。通过最大化图像的二维属性直方图中目标和背景分布的相关量来选择阈值向量。同时,为了节省二维阈值算法的计算时间,给出了递推算法,此算法减少了大量的重复计算,有利于该算法的实时应用。对铸件中气孔、缩孔和杂质三种缺陷进行了分割,试验结果表明,该算法能够快速、准确地分割出铸件图像中的缺陷。

In this paper, an image segmentation method using maximum correlation criterion based on two-dimensional bound histogram is proposed to detect defect in X-ray image, which possesses low contrast, blur edge and is corrupted by noise. First, the bound set is constructed by thresholding the image using maximum entropy method. Then, the two-dimensional bound histogram is built according to original image and the constructed bound set. The best threshold vector is obtained by maximizing the sum of correlation of the object and background. Meanwhile, a recursive algorithm, which avoids much repeated computation, is proposed. The proposed method is used to segment the blowhole, shrinkage and impurity and experimental results show that it can correctly detect the defect at a high speed.