一种基于灰度级连通性的红外图像分割方法

本文提出一种新的基于灰度级连通性的红外图像分割方法。灰度级连通性认为在某个灰度级以下的所有级集合是连通的，则灰度图像是连通的。提出的图像分割方法使用k级特征开运算将图像中包含目标的k级以上的连通成分保留下来，结合图像弱小目标的特征进行k级连通成分分解运算，提取出包含目标的k级连通成分实现图像简化和目标提取，最后结合简单的二值化处理就能够准确地分割出目标。通过仿真结果的比较，证明在红外图像中这种方法可以实现提高信噪比、提取目标和分割图像的目的。

A Segmentation Approach Based on Grayscale Level Connectivity in IR Image

A new segmentation approach is proposed based on grayscale level connectivity in IR image. In this connectivity, a grayscale image is connected if all level sets below a prespecified threshold are connected. First, the level-k characteristic opening using appropriate thresholding "keeps" only the regional maxima of image that are at or above level k, and flattens the rest. Second, level-k connected component decomposing based on target size can extract connected component that contain the objects of interest. The end step can segment target combining with simple thresholding. Experimental results in real IR images demonstrate the effectiveness and robustness of the proposed approach to enhance signal noise ratio, extract target and segment image.