In this paper we introduce a general purpose graphical processing unit (GPGPU) based method for performing a sweep across a set of the scale invariant feature transform (SIFT) parameters for pairs of images. The focus of the paper is the analysis of the data generated using information visualisation techniques including a cross brushing technique between parallel coordinates, scatter plots and histograms. Results have shown us the importance of carefully selecting some parameters depending upon the properties of an image pair while other parameters are shown to be robust to variation. The parameters chosen by analysis of the sweep data have then been compared to the previously published SIFT's values and a consistent improvement in accuracy is shown. © 2010. The copyright of this document resides with its authors.
|Title of host publication||British Machine Vision Conference, BMVC 2010 - Proceedings|Br. Mach. Vis. Conf., BMVC - Proc.|
|Publication status||Published - 2010|
|Event||2010 21st British Machine Vision Conference, BMVC 2010 - Aberystwyth|
Duration: 1 Jul 2010 → …
|Conference||2010 21st British Machine Vision Conference, BMVC 2010|
|Period||1/07/10 → …|