Abstract for johnson_stable06

Cambridge University Engineering Department Technical Report TR20060916


Matthew Johnson and Roberto Cipolla

September 2006

Local interest points and descriptors have been used very successfully to achieve accurate and efficient image retrieval and matching performance which is robust to occlusion and limited viewpoint change. Currently, these systems tend to be initialized from still images and require that a thousand or more points be stored in a retrieval data structure for each object. Many of these points are rarely if ever used, and thus unnecessarily limit the number of reference images that can be stored effectively. We propose a method for determining the stability of local interest points and their descriptors such that an efficient and effective subset of points can be stored. This technique has been shown to reduce the number of required points by an order of magnitude while improving performance, allowing for significantly smaller data structures for use in retrieval and matching.

(ftp:) johnson_stable06.pdf (http:) johnson_stable06.pdf

If you have difficulty viewing files that end '.gz', which are gzip compressed, then you may be able to find tools to uncompress them at the gzip web site.

If you have difficulty viewing files that are in PostScript, (ending '.ps' or '.ps.gz'), then you may be able to find tools to view them at the gsview web site.

We have attempted to provide automatically generated PDF copies of documents for which only PostScript versions have previously been available. These are clearly marked in the database - due to the nature of the automatic conversion process, they are likely to be badly aliased when viewed at default resolution on screen by acroread.