Evaluating Image Segmentation Algorithms: Mean Shift, K-means, and Multi-Otsu on BSDS500 with Human-Annotated Reference

In this article, we conduct a comparative study of three image segmentation algorithms: Mean Shift, K-means, and Multi-Otsu. The primary objective of this study is to evaluate the performance of these algorithms using measures based on visual results obtained from the BSDS500 dataset. We apply the three algorithms to the test images of BSDS500 and generate image segments. Subsequently, we visually compare the segments obtained by each algorithm with the reference annotations provided by BSDS500. The advantages and limitations of each algorithm are discussed based on the alignment of the produced segments with the reference annotations. This study provides valuable insights to researchers and practitioners interested in image segmentation by highlighting the relative performances of Mean Shift, K-means, and Multi-Otsu in terms of visual agreement with the reference segments of the BSDS500 dataset.