Automatic estimation of clusters number for K-means

Abstract

At the present time, clustering algorithms are popular analysis tools in image segmentation. For instance, the K-means is one of the most used algorithms in the literature and it is fast, robust and easier to understand and to implement but, the main drawback of the K-means algorithm is that the number of clusters must be known a priori and must be supplied as an input parameter. This paper discusses the problem of the estimation of the number of clusters for image segmentation and proposes a new approach which is based on histogram to find a suitable number of K (the number of clusters). Experimental results demonstrate the effectiveness of our method to estimate the correct number of clusters which reflect a good separation of objects for each image.