A Maximum Likelihood Framework for Iterative Eigendecomposition
Abstract
This paper presents an iterative maximum likelihood framework for perceptual grouping. We pose the problem of perceptual grouping as one of pairwise relational clustering. The method is quite generic and can be applied to a number of problems including region segmentation and line-linking. The task is to assign image tokens to clusters in which there is strong relational affinity between token pairs. The parameters of our model are the cluster memberships and the link weights between pairs of tokens. Commencing from a simple probability distribution for these parameters, we show how they may be estimated using an EM-like algorithm. The cluster memberships are estimated using an eigendecomposition method. Once the cluster memberships are to hand, then the updated link-weights are the expected values of their pairwise products. The new method is demonstrated on region segmentation and line-segment grouping problems where it is shown to outperform a noniterative eigenclustering method.
Cite
Text
Robles-Kelly and Hancock. "A Maximum Likelihood Framework for Iterative Eigendecomposition." IEEE/CVF International Conference on Computer Vision, 2001. doi:10.1109/ICCV.2001.10091Markdown
[Robles-Kelly and Hancock. "A Maximum Likelihood Framework for Iterative Eigendecomposition." IEEE/CVF International Conference on Computer Vision, 2001.](https://mlanthology.org/iccv/2001/robleskelly2001iccv-maximum/) doi:10.1109/ICCV.2001.10091BibTeX
@inproceedings{robleskelly2001iccv-maximum,
title = {{A Maximum Likelihood Framework for Iterative Eigendecomposition}},
author = {Robles-Kelly, Antonio and Hancock, Edwin R.},
booktitle = {IEEE/CVF International Conference on Computer Vision},
year = {2001},
pages = {654-661},
doi = {10.1109/ICCV.2001.10091},
url = {https://mlanthology.org/iccv/2001/robleskelly2001iccv-maximum/}
}