分形维-最大熵阈值的弱边缘工业CT图像分割算法

Weak Edge Industrial CT Image Segmentation Algorithm Based on Fractal Dimension-Maximum Entropy Threshold

含有弱边缘的工业CT图像在分割时易产生过分割现象,对此提出了一种分形维定位和最大熵阈值分割的图像分割算法.首先,采用中值滤波和高斯平滑对图像进行预处理;其次,对预处理后的图像进行分块,并求取每个分块的分形维;再次,根据背景和缺陷处分形维的差异对缺陷进行粗定位,并根据连通域分析精确定位缺陷区域;最后,利用最大熵阈值法对精确定位后的局部区域进行分割.仿真实验表明:所提算法具有良好的分割能力,可以准确地分割出含有弱边缘的缺陷目标,并有效排除轮廓背景对分割的干扰,避免了过分割.

Industrial CT images with weak edges are prone to over segmentation in segmentation.This study proposed an image segmentation algorithm to cope with the over segmentation problem based on fractal dimension location and maximum entropy threshold segmentation.Firstly,the images were preprocessed by using median filtering and gaussian smoothing;Then,the preprocessed images were divided into blocks and fractal dimension of each block was calculated;After that,the rough positioning of defects was done according to the difference of the fractal dimension between the background and the defect,and accurate positioning of defect areas was done based on connected domain analysis;Finally,the maximum entropy threshold method was used to accurately segment the positioned defect areas.The simulation results show that the proposed algorithm has good performance in segmentation.It helps to accurately segment the defect target with weak edges,effectively eliminate the interference of contour background to segmentation,and avoid over segmentation.