Edge based technique to estimate number of clusters in k-means color image segmentation

K-Means algorithm is an unsupervised clustering algorithm that classifies the input data points into multiple classes based on their inherent distance from each other. Success of k-means color image segmentation depends on parameter k. If numbers of clusters are estimated correctly, k-means image segmentation can provide good results. This paper proposes a novel method based on edge detection to estimate number of clusters automatically. Edges are detected in terms of phase congruency. Short edges reflect the local character in image while long edges are more important to estimate number of clusters. The short edges are eliminated. Edge line clustering is used to group long edges based on color similarity. For grouping color similar edges average color of each edge is calculated. Euclidean distance on average color of each pair of edges is calculated. Long edges assigned same label if Euclidean distance on average color is less. We have estimated the number of clusters in image using edge line clustering. The number of edges left after edge line clustering is thought as number of clusters in image. This value is used as value of k for k-means image segmentation.