Boundary outlier centroid based reduced overlapping image segmentation

Abstract

This paper proposes a novel approach to image segmentation that integrates the improved K-means algorithm with Otsu Segmentation and Watershed algorithm. The conventional K-means algorithm is widely used for image segmentation because of its ease of use and small computation time, but has certain drawbacks such as strongly affected by outliers and may settle at the local minima, thereby leading to poor scaling. Similarly, the traditional watershed algorithm is widely used but has certain drawbacks such as over segmentation and high sensitivity. So the shortcomings of the conventional K-means algorithm and the hide-bound Watershed algorithm are addressed by integrating them such that better segmenting results are obtained. The proposed BOCRO algorithm utilises the combination of clustering algorithm with initial centroids by Centroid Calculation and Overlapping Reduction with Boundary Outliers and is efficient in terms of complexity and produces accurate results. It can be incorporated with a device to detect and segment images to obtain meaningful information.