Towards parameter-independent data clustering and image segmentation

作者：

Highlights：

• We study how similarity measures influence the dominant sets clustering results.

• We use histogram equalization to remove the dependence on similarity parameters.

• We present a density based cluster extension method to overcome over-segmentation.

• Experiments validate the effectiveness of our algorithm.

摘要

Highlights•We study how similarity measures influence the dominant sets clustering results.•We use histogram equalization to remove the dependence on similarity parameters.•We present a density based cluster extension method to overcome over-segmentation.•Experiments validate the effectiveness of our algorithm.

论文关键词：Dominant sets,Clustering,Image segmentation,Similarity matrix,Similarity measure

论文评审过程：Received 21 October 2015, Revised 17 March 2016, Accepted 28 April 2016, Available online 17 May 2016, Version of Record 31 May 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.04.015