Towards a Faster Image Segmentation Using the K-means Algorithm on Grayscale Histogram

The K-means is a popular clustering algorithm known for its simplicity and efficiency. However the elapsed computation time is one of its main weaknesses. In this paper, the authors use the K-means algorithm to segment grayscale images. Their aim is to reduce the computation time elapsed in the K-means algorithm by using a grayscale histogram without loss of accuracy in calculating the clusters centers. The main idea consists of calculating the histogram of the original image, applying the K-means on the histogram until the equilibrium state is reached, and computing the clusters centers then the authors use the clusters centers to run the K-means for a single iteration. Tests of accuracy and computational time are presented to show the advantages and inconveniences of the proposed method.