一种新的图像分割算法

A New Image Segmentation Algorithm

在形态学梯度边缘检测算子的基础上,综合多尺度和多结构元算法的特性,提出了一种新的图像分割方法。利用桥梁的明显特征预测桥梁目标的位置,获得感兴趣区域,通过多阈值法简化原图;采用多尺度形态滤波和区域标记得到目标的初始轮廓,构造5个不同方向的结构元素,对这些结构元素运用多尺度形态学方法来检测目标边缘。实验结果表明本算法能够比较准确地完成图像分割。

Based on edge detection operator of morphological gradient, this paper presents a new segmentation algorithm for image based on multi-scale and multi-structure element method. Firstly, it forecasts bridge position using the obvious characteristics, and obtains the interesting region, and simplifies the original map through multi-threshold method. Then, the target initial contour can be obtained through the multi-scale morphological filtering and region labeling. Finally five different orientation structure elements are constructed to detect target contour by multi-scale morphology. The experimental results show that the proposed method can complete the image segmentation.