A Comparison of Measures for Detecting Natural Shapes in Cluttered Backgrounds

  • Authors:

  • Lance R. Williams;Karvel K. Thornber

  • Affiliations:

  • Dept. of Computer Science, University of New Mexico, Albuquerque, NM 87131, USA;NEC Research Institute, 4 Independence Way, Princeton, NJ 08540, USA

  • Venue:
  • International Journal of Computer Vision - Special issue on computer vision research at NEC Research Institute
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

We propose a new measure of perceptual saliency and quantitativelycompare its ability to detect natural shapes in cluttered backgroundsto five previously proposed measures. As defined in the new measure,the saliency of an edge is the fraction of closed random walks whichcontain that edge. The transition-probability matrix defining therandom walk between edges is based on a distribution of natural shapesmodeled by a stochastic motion. Each of the saliency measures in ourcomparison is a function of a set of affinity values assigned to pairsof edges. Although the authors of each measure define the affinitybetween a pair of edges somewhat differently, all incorporate theGestalt principles of good-continuation and proximity in some form. Inorder to make the comparison meaningful, we use a single definition ofaffinity and focus instead on the performance of the differentfunctions for combining affinity values. The primary performancecriterion is accuracy. We compute false-positive rates in classifyingedges as signal or noise for a large set of test figures. In almostevery case, the new measure significantly outperforms previousmeasures.