Dynamic spectral residual superpixels

作者：

Highlights：

• A novel superpixels approach, which incorporates a saliency function.

• Saliency prevent unnecessary over-segmentation to focus on relevant objects.

• A new seeding strategy based on the inhomogeneity of a saliency function.

• We report the lowest undersegmentation error and highest boundary recall.

摘要

•A novel superpixels approach, which incorporates a saliency function.•Saliency prevent unnecessary over-segmentation to focus on relevant objects.•A new seeding strategy based on the inhomogeneity of a saliency function.•We report the lowest undersegmentation error and highest boundary recall.

论文关键词：Superpixels,K-means,Spectral residual,Segmentation

论文评审过程：Received 20 January 2020, Revised 27 August 2020, Accepted 13 October 2020, Available online 21 October 2020, Version of Record 30 January 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107705