GEOSCAN Search Results: Fastlink


TitleOn the Hausdorff distance used for the evaluation of segmentation results
DownloadDownloads (Preprint)
LicencePlease note the adoption of the Open Government Licence - Canada supersedes any previous licences.
AuthorBeauchemin, M; Thomson, K P B; Edwards, G
SourceCanadian Journal of Remote Sensing vol. 24, issue 1, 1998 p. 3-8,
Alt SeriesEarth Sciences Sector, Contribution Series 20042944
PublisherInforma UK Limited
Mediapaper; on-line; digital
File formatpdf
Subjectsremote sensing; Hausdorff distance
Illustrationsdiagrams; graphs; tables
Released2014 07 31
AbstractImage segmentation tools are widely used in remote sensing. Their implementation in most commercial image analysis packages demonstrates the interest in them. A considerable effort has been directed toward the development of segmentation algorithms, resulting in a profusion of different techniques. In comparison, the number of tools developed for the quantitative evaluation of segmentation outputs is very limited. In fact, visual inspection is still considered, by many, as the ultimate quality test. In this papel; a new quantitative method for the evaluation of segmentation results is presented. It relies on the partial directed Hausdorff distance. The method focuses on boundary point matching between a segmentation output and a reference partition. The method is illustrated with a case taken from the literature and is shown to provide useful information to assess the quality of a segmentation.

Date modified: