- Title: Image Matching Using SIFT-Descriptors & Spatial Information
- Description: This thesis addresses the problem of content based image matching algorithms based on key point descriptors. The principle is to extract general information from an image without any specic predened query. First SIFT-descriptors are extracted from a large set of images. This set is then clustered into a tree structure using k-means in each node. Images can now be described using this tree as a codebook that enables fast and accurate comparisons between images. The thesis also examines the possibilities for improvements in matching accuracy and runtime performance. Some improvements compared to similar applications have been achieved in the oine calculation time with clustering done in parallel.
- Start Date: March 4, 2011
- Finished Date: March 21, 2011
- Supervisor: Håkan Ardö
- Student: Erik Södervall, (-)
Questions: webmaster
Last update: 2013-04-11
Centre for Mathematical Sciences, Box 118, SE-22100, Lund. Telefon: +46 46-222 00 00 (vx)