Evaluating image thresholding techniques for land cover modification mapping

Document Type

Conference Paper


Change detection is an important application of remote sensing in environmental monitoring. This study examined landcover modification in San Diego, California using two Landsat images, one acquired in 1990 and the other in 1996. These images were enhanced using the Kauth Thomas (KT) transformation to produce brightness, greenness and wetness features. Change detection techniques used were the standard global thresholding change detection technique and the new localized thresholding change detection (LTCD) method. Unlike the global thresholding method, in which upper and lower threshold limits are the same number of standard deviations about a given mean, this new method uses threshold limits that vary in the number of standard deviations from the mean for the upper and lower threshold limits. Hence a typical threshold used in change detection may have a lower limit of 2.5 standard deviations about the mean of the image with an upper limit of 1.5 standard deviations from the mean. Change/no-change areas were more accurately mapped using the LTCD method. Accuracies obtained using the LTCD in mapping change were observed to be consistent with those obtained when change/no-change pixels were mapped using the global approach in mapping change in brightness and change in greenness. However the LTCD did not perform quite as well in mapping change in wetness where a decrease of 30% in accuracy was observed.

Publication Title

American Society for Photogrammetry and Remote Sensing - Annual Conference 2005 - Geospatial Goes Global: From Your Neighborhood to the Whole Planet

Publication Date




First Page


Last Page





luminance, remote sensing, signal detection, statistics

This document is currently not available here.