中文摘要：

针对阈值法分割红外图像易产生误分割和水平集分割方法受初始曲线限制大,提出了一种结合模糊阈值与水平集的自适应红外图像分割方法。该方法首先采用二维Otsu方法计算阈值,利用该阈值获取模糊阈值分割法中的窗口宽度,使模糊阈值分割法具有自适应性;然后采用此自适应模糊阈值分割法预分割红外图像,利用预分割结果自动获取水平集初始曲线;最后将Chan-Vese方法与Shi方法结合提出改进的水平集方法,并用此方法分割红外图像。实验结果表明,本文方法具有较好的分割效果和较强的鲁棒性。

英文摘要：

As it is likely to generate redundant contours by fuzzy-threshold-based infrared image segmentation method and it is sensitive to initial contour by level set segmentation method, an adaptive segmentation method of infrared image based on fuzzy threshold and level set is presented. Firstly, this algorithm introduces 2-D Otsu to calculate a threshold, and then the window width of fuzzy threshold is obtained based on the threshold from 2-D Otsu. Then, the adaptive fuzzy threshold is adopted to pre-segment infrared images, and the initial contour of level set is obtained depending on the pre-segmentation result. Finally, an improved level set method by combing Chan-Vese method and Shi method is proposed, and the improved level set method is adopted to segment infrared images. The results of experiments prove that this algorithm has better segmentation effect and robustness.