Image Segmentation: Automatic Cluster Number Generation in K-Means
Anju Bala, Aman Sharma · 2019
Nowadays, image analysis has become popular in every field, and the process is different for the different field. This leads to a huge advancement in the area of image analysis. One of the major steps of image analysis is image segmentation that plays a vital role in extracting the information from an image. Although there are several segmentation algorithms present in literature, however, there is no standard algorithm which works equally in all fields. Clustering algorithms are a new way of identifying and classifying regions in an image using mathematical patterns. K-means is one of the most popular clustering algorithms because of its simplicity and efficiency. However, it has three main drawbacks: 1) cluster number, 2) initialization and 3) noise sensitivity. This paper focuses on the cluster number problem and proposes a novel threshold-based approach to find out the cluster numbers. Experimental results show the effectiveness of the proposed cluster number generation method.