工业生产场景下的图像成像环境复杂, 在基于机器视觉的图像处理任务中要对该类图像实施精确分割很不容易。针对这一问题, 应用信息论中的距离测度理论结合高斯分布提出一种基于信息论距离的图像阈值分割方法。在提出方法中运用信息论距离度量分割前后图像信息的损失程度, 通过最小化该距离获取最佳分割阈值, 然后应用该阈值对图像进行分割。最后在工业无损检测图像、红外图像以及医学脑血管造影术图像上与几种经典及较新的图像分割方法进行了实验比较。结果表明, 提出方法获得的结果视觉效果好, 分割精度高, 具有较好的应用推广前景。

In the task of image processing based on machine vision, the segmentation of images in the production scenarios in complex environments is not easy. A new method based on an information-theoretic distance measure combined with a Gaussian distribution is presented to combat this problem. The information-theoretic distance is regarded as the measuring tool of loss of information in segmented images. The optimal segmentation result is determined by minimizing the distance measured in the proposed method. Finally, a series of experiments on non-destructive testing images, infrared images, and cerebral angiography images is conducted based on the comparison with some classical and state-of-the-art methods. The results demonstrate that the proposed method has a better visual effect, higher precision with regards to segmentation, and can meet the demand in practice.