基于Hessian矩阵和熵的CT序列图像裂缝分割方法

Crack segmentation in CT image sequences using Hessian matrix and entropy

工业CT序列图像的各向不同性和伪影会影响裂缝分割精确度和准确度,因此提出一种基于Hessian矩阵和熵的各向不同性工业CT序列图像裂缝自动分割方法。首先,用基于Hessian矩阵的多尺度线状滤波增强裂缝区域,抑制非线状区域;然后,建立一种新的二维直方图,获取滤波之后层内和层间的信息;再根据直方图的最大类熵确定阈值区间,最终得到裂缝的二值化分割结果。实验表明,所提方法不仅能够满足实际工业CT序列图像裂缝分割中精确、自动的分割要求,而且相较其他4种已有方法,能够得到更完整、更准确的分割结果。

The anisotropic property and artifacts of CT image sequences adversely affect the accuracy and the precision of crack segmentation. Thus we propose an automatic crack segmentation approach for anisotropic industrial CT image sequences using Hessian matrix and entropy. Firstly, we adopt multi-scale linear filtering based on Hessian matrix to enhance cracks in the image s/ices and restrain the non-linear region. Further, we generate a novel two-dimensional histogram to capture both intralayer and interlayer information from filtered results. The threshold values are determined by the maximum class entropies according to this histogram. Finally, the binary segmentation results for cracks are derived. The experiments demonstrate that the proposed method meets the requirements of automation and high accuracy in crack segmentation of industrial CT image sequences and yields more comprehensive and accurate results, compared with other four existing methods.