Cluster-based point set saliency

• Flora P. Tasse
• Jiří Kosinka
• Neil A. Dodgson

Cluster-based point set saliency on a range scan. From left to right: normal map, adaptive fuzzy clustering (each cluster a different colour), cluster uniqueness metric, cluster spatial distribution metric, cluster saliency calculated from uniqueness and distribution metrics, point saliency derived from cluster saliency.

Abstract

We propose a cluster-based approach to point set saliency detection, a challenge since point sets lack topological information. A point set is first decomposed into small clusters, using fuzzy clustering. We evaluate cluster uniqueness and spatial distribution of each cluster and combine these values into a cluster saliency function. Finally, the probabilities of points belonging to each cluster are used to assign a saliency to each point. Our approach detects fine-scale salient features and uninteresting regions consistently have lower saliency values. We evaluate the proposed saliency model by testing our saliency-based keypoint detection against a 3D interest point detection benchmark. The evaluation shows that our method achieves a good balance between false positive and false negative error rates, without using any topological information.

Video

Example

• Navigable version of the Bremen City data set
  The dataset has 11.1M points. The default is to display 0.4M points (4% of the dataset). Use the slider at top right to change the number of points displayed.

Downloads

• Paper (PDF, 5.7 MB)
• Source: Soon

Bibtex

    @inproceedings{Tasse2015,
     author    = {Tasse, Flora P. and Kosinka, Ji{\v{r}}{\'{i}}
                  and Dodgson, Neil A.},
     title     = {Cluster-based point set saliency},
     year      = {2015},
     booktitle = {ICCV 2015},
     url       = {http://www.cl.cam.ac.uk/research/rainbow/projects/pointsetsaliency/},
    }