基于分水岭和K-均值的半自动眉毛图像分割

摘要

To extract a pure eyebrow image from an original image rapidly and steadily, an eyebrow segmentation method based on watershed and K-means algorithm was presented, which was called W-K algorithm. First, a number of eyebrow pixels and non-eyebrow pixels by manually scratching several simple lines on an original eyebrow image were labeled; Second, the watershed algorithm was used to produce catchment basins, and them were clustered by K-means algorithm; Finally, a pure eyebrow image was extracted by eyebrow pixel filtering. Experiment results show that it can segment pure eyebrow images in high speed and good performance for preprocessing to improve eyebrow recognition accuracy.%为了从原始图像中快速、稳定地提取纯眉毛图像，提出了一种融合分水岭和K-均值算法的眉毛图像分割方法，即W-K算法．首先通过手工在眉毛图像上画上几条线标注部分眉毛点和非眉毛点，其次利用分水岭算法产生蓄水盆，再使用K-均值算法对蓄水盆进行聚类，最后通过眉毛点筛选实现纯眉毛图像的分割．实验结果表明，该方法在分割纯眉毛图像的过程中具有速度快、效果好的优点，可用于眉毛识别的前期预处理，并有助于提高识别结果的准确率．