Feature-Cut: Video Object Segmentation Through Local Feature Correspondences

Abstract

Accurately segmenting objects in video is a difficult and\ntime consuming process in modern post-production houses.\nAutomatic systems may work for a small number of frames,\nbut will typically fail over longer video shots. This work\nproposes a semi-automatic, feature-based system to perform\nobject segmentation over longer sequences. The user\nmanually extracts masks from representative instances of\nthe object, which are then propagated to the remaining unsegmented\nframes and used to bootstrap the automatic segmentation\nfor these frames. The presented work dramatically\nreduces the manual workload required to segment a\nvideo sequence, allowing longer and more accurate object\nmattes.

Cite

Text

Ring and Kokaram. "Feature-Cut: Video Object Segmentation Through Local Feature Correspondences." IEEE/CVF International Conference on Computer Vision Workshops, 2009. doi:10.1109/ICCVW.2009.5457644

Markdown

[Ring and Kokaram. "Feature-Cut: Video Object Segmentation Through Local Feature Correspondences." IEEE/CVF International Conference on Computer Vision Workshops, 2009.](https://mlanthology.org/iccvw/2009/ring2009iccvw-featurecut/) doi:10.1109/ICCVW.2009.5457644

BibTeX

@inproceedings{ring2009iccvw-featurecut,
  title     = {{Feature-Cut: Video Object Segmentation Through Local Feature Correspondences}},
  author    = {Ring, Dan and Kokaram, Anil C.},
  booktitle = {IEEE/CVF International Conference on Computer Vision Workshops},
  year      = {2009},
  pages     = {617-624},
  doi       = {10.1109/ICCVW.2009.5457644},
  url       = {https://mlanthology.org/iccvw/2009/ring2009iccvw-featurecut/}
}