Unsupervised image segmentation evaluation based on feature extraction

Image segmentation is widely used in life. Generally speaking, the segmentation results are divided into good and bad quality, so it is very important to propose an effective method to evaluate the quality of image segmentation. This paper proposed a framework based on edge detection and feature extraction for evaluating the quality of image segmentation. The framework belongs to unsupervised evaluation, the operation is simple and easy to implement, and readers can add or subtract methods in the framework according to specific circumstances. To prove the effectiveness of the proposed framework, we tested on four different datasets. In addition, we compare the proposed framework with some classic and newer evaluation methods. Experimental results show that the proposed framework is suitable for many types of images, and its performance is better than some existing metrics.