Superpixel results

Most existing superpixel algorithms only consider color intensity and position coordinates, while ignoring local neighborhood factors. This limitation leads to low applicability in noisy and cluttered environments. To address this issue, we propose a seminal and novel Fuzzy C-Means clustering with Region Constraints for Superpixel generation (RCFCMS). First, employing region constraints to prevent boundary crossing. Second, adopting spatial information to mitigate noise interference. Third, utilizing soft membership to convert labels. Finally, applying connected components to optimize superpixels. The qualitative and quantitative evaluation demonstrates that the proposed RCFCMS outperforms state-of-the-art superpixel algorithms, as shown in the attached dataset.