基于边界几何特征的重叠颗粒图像分割算法

Overlapping Grain Image Segmentation Algorithm Base on Boundary Geometrical Characteristics

图像分割是图像处理中的重要问题,也是计算机视觉领域低层次视觉中的主要问题,是成功进行图像分析、理解和描述的关键技术。该文构造了一种针对二值图像边界几何特征的提取的算法,通过对图形边界的凹特性分析,实现了对不规则颗粒图像的分割。该算法仅对提取的图像数据作处理,不对图像本身作任何运算,避免了传统的基于图像形态学的分割算法所导致的图像信息丢失。

Image segmentation is an important problem in image processing, the main of vision bottom in computer vision field, and the key technology of analyzing, understanding and characterizing images. In this paper an algorithm of extracting boundary geometrical characters from a bi-value image is constructed. Anomalous image segmentation is completed by analyzing concave properties of graph. The algorithm processes image data extracted without operating the image itself to avoid the loss of image information by traditional segmentation algorithm based on image morphology.