Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6937737 | Image and Vision Computing | 2018 | 24 Pages |
Abstract
Superpixel has become an essential processing unit in many computer vision systems, and superpixel segmentation of images is one of the most important step. In this paper, an efficient superpixel segmentation algorithm was proposed. We introduce a new compact-aware minimum barrier distance for superpixel segmentation (MBS), and a propagation scheme for the cluster centers between adjacent levels on a hierarchical architecture. Experiments show that it achieves state-of-the-art performance and can be configured with simple trade-off between performance and efficiency. Furthermore, the compactness of segmented superpixel could be flexibly controlled continuously by only one parameter, which could be easily integrated in other computer vision tasks. The source code of MBS is available at https://github.com/YinlinHu/MBS.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Vision and Pattern Recognition
Authors
Yinlin Hu, Yunsong Li, Rui Song, Peng Rao, Yangli Wang,