基于改进分水岭算法的球团矿图像分割

Image segmentation of pellets based on improved watershed algorithm

针对传统图像分割算法不能对球团矿图像中粘连球团进行准确有效分割的问题,提出一种改进的分水岭分割算法。该算法将标记和分水岭算法结合起来,利用标记带来的先验知识限定图像分割区域的数目,并对分水岭算法的梯度图像进行修改,从而能够对球团边缘准确分割,同时避免产生过分割现象。实验结果表明,改进分水岭算法对图像中粘连球团的边缘有良好的分割效果。

An improved watershed segmentation algorithm is proposed to solve the problem that cannot be solved by traditional image segmentation algorithms to segment the adhesive pellets in pellet ima- ges effectively and accurately. It combines markers with watershed algorithm, takes advantage of the prior knowledge from markers to limit the number of image segmentation regions, and modifies the gradient image in order to correctly segment the pellet edges and avoid the over-segmenting phenome- non in the meantime. Experimental results show that the improved watershed algorithm has better segmentation effect.