Invariant Features from Interest Point Groups

Matthew Brown, D Lowe

Research output: Contribution to conferencePaper


This paper approaches the problem of ¯nding correspondences between images in which there are large changes in viewpoint, scale and illumi- nation. Recent work has shown that scale-space `interest points' may be found with good repeatability in spite of such changes. Further- more, the high entropy of the surrounding image regions means that local descriptors are highly discriminative for matching. For descrip- tors at interest points to be robustly matched between images, they must be as far as possible invariant to the imaging process.

In this work we introduce a family of features which use groups of interest points to form geometrically invariant descriptors of image regions. Feature descriptors are formed by resampling the image rel- ative to canonical frames de¯ned by the points. In addition to robust matching, a key advantage of this approach is that each match implies a hypothesis of the local 2D (projective) transformation. This allows us to immediately reject most of the false matches using a Hough trans- form. We reject remaining outliers using RANSAC and the epipolar constraint. Results show that dense feature matching can be achieved in a few seconds of computation on 1GHz Pentium III machines.

Original languageEnglish
Number of pages10
Publication statusPublished - Sept 2002
EventBMVC 2002: 13th British Machine Vision Conference - Cardiff
Duration: 2 Sept 20025 Sept 2002


ConferenceBMVC 2002: 13th British Machine Vision Conference


Dive into the research topics of 'Invariant Features from Interest Point Groups'. Together they form a unique fingerprint.

Cite this