基于Labelme的参考图像的手工分割

Manual segmentation of ground truth image based on Labelme

图像分割算法虽然已经有了大量的研究,但没有适用于所有图像的通用分割算法。因此针对不同图像的实际情况,需要进行大量的实验来选择最优的分割算法。通过比较算法分割得到的结果图像与对应手工分割得到的参考图像之间的差异,可以得到一种图像分割算法的性能评价,因此本文提出一种参考图像的获取方法。通过结合算法和Label Me在线注释软件,该方法能够方便地完成图像中各种目标边界的定位并保存图像。

Although a large amount of research has been done to image segmentation algorithm, there is no general segmentation algorithm that applies to all images. Therefore, for the actual situation of separate images, a serious of experiments are essential to choose the optimal algorithm for image segmentation. Performance evaluation of image segmentation algorithm can be achieved by comparing the differences between the automatic segmented image and manually segmented image. So this paper proposed a method to obtain ground truth image. By combining algorithm and LabelMe open annotation tool, the method can locate the boundaries of various targets easily and save images.