Advanced Density Peak and K-means Clustering on Image Segmentation

K-means is a classical and widely-used data clustering algorithm. Despite its effectiveness, the drawbacks are obvious that it needs to know k value previously and not suitable for complex situations. Density Peak clustering can practice on irregular data sets with a higher accuracy and better performance than K-means and doesn't need to get prior knowledge. However, few concentrated on their performances on image segmentation. In this paper, we propose novel image segmentation approaches based on K-means and Density Peak clustering which greatly reduce running time. Compared with current methods, our methods have improved aspects as following: 1) The methods could have much shorter run time performance than other current normal methods. 2) Unlike other current image segmentation methods, our method could save the original colors of the pictures and provide a rather real image segments. Experiments on test data will testify the validity of the methods and a detailed description based on empirical results will be provided as conclusions.