摘要： 砾石图像的复杂多样使得每颗完整的砾石往往具有非均匀多彩色的性质，用一般的分割方法很难准确检测出砾石边缘。该文提出用高精度的边缘流和水平集分割方法提取图像中所有强弱边缘，用一种基于彩色梯度和邻域均值方差相结合的方法去除大砾石内的虚假边缘，使每个颗粒边缘尽量完整。实验证明该算法效果良好，便于后续处理。

关键词: 边缘流, 水平集, 图像分割, 梯度

Abstract: The whole gravel stone often has non-uniform and multi-color nature because of its complexity and variety, so it is difficult to accurately detect its edge with original method. This paper presents a high-precisioned method of segmentation named edge flow and level set to extract all the strong and weak edges. In order to get rid of feigned edges in big gravel stone, it proposes a method combining color gradient and mean-variance, which makes each edge of the stone more complete. Experimental results show that the the method is good and lays a good foundation for follow-up treatment.

Key words: edge flow, level set, image segmentation, grads

中图分类号: 

• TP391.41