一种遥感二值图像目标区域分割的新方法

基于线段扫描法进行二值图像连通域分割时，对数据量较多且形状复杂的遥感二值图像，容易使邻接表存储大量的等价对信息，即浪费存储空间也不利于算法合并处理。针对这一不足，提出了一种基于线段的快速标号算法，采用“双表”实时记录和修正等价标号，很好地解决了标记冲突的问题。经模拟数据和真实遥感二值图像验证表明，该算法比传统算法在处理效率上有显著提高，具有较好的应用价值。

New area segmentation method for remote sensing binary image

When the line based scanning method is used to segment the connected components of the remote sensing binary image, which has such a huge volume of data and complex shape, the adjacent table will record mass information for equivalent pairs, and result in the waste of memory and the disadvantage of algorithm running. According to this problem, this paper presents a new fast line based labeling method. The proposed method uses the double tables to record the connectivity between the adjacent pixels and corrects the labels real time, which resolves the problem of label conflict and marks all the connected components quite well. Confirmed by the simulation data set and the experimental results with real data, compared to the traditional algorithm, this method improves treatment efficiency significantly and possesses better practicability.