OpenCV分水岭算法的改进及其在细胞分割中的应用

Improvement on watershed algorithm of OpenCV and its application in cell image segmentation

分水岭算法是一种基于形态学的图像分割算法,能快速准确地确定连通区域的边界。将基于标记的分水岭算法用于细胞图像的分割,较好地解决了粘连细胞的分割问题。在该细胞分割算法的实现过程中,发现了OpenCV分水岭算法实现的缺陷,通过对相关代码的分析,发现该缺陷存在的原因是算法流程中对相邻像素相对关系的描述存在问题。将OpenCV分水岭算法中对相邻像素取差的绝对值,改为对相邻像素取差值,对该算法进行了改进。实验证明,改进后的OpenCV分水岭算法对细胞图像的分割效果明显好于直接使用OpenCV分水岭算法得到的结果。该方法在不影响分割速度的情况下,提高了OpenCV分水岭算法分割的准确度。

Watershed is an image segmentation algorithm based on morphology, which can determine the boundary of connected section efficiently and effectively. Application of marker watershed algorithm in cell image segmentation leads to the solution of adhesion cell segmentation. When implementing this algorithm, a flaw of the watershed algorithm in OpenCV has been found. The cause of this bug is the wrong description of the difference between two adjacent pixels. An improvement of OpenCV watershed was proposed by replacing the absolute difference between adjacent pixels with the difference between them. The experiment shows that the improved OpenCV watershed algorithm has a better segmentation performance than that of original algorithm with same processing speed.