Change detection in very high resolution imagery based on dynamic time warping: An implementation for Haiti earthquake damage assessment

Abstract

Very high resolution (VHR) imagery exhibits high level of detailed information, but this may lead to the complicated urban structures in the spectral domain. In the context of change detection implementation, there are some issues when dealing with multitemporal change detection analysis in VHR imagery. One of the most important issues is to make a very accurate registration of two images taken at two different times. Due to some problems caused by scaling, shifting and rotation of the images, it is mostly impossible to register two images pixel-to-pixel exactly. Noise reduction, radiometric and geometric corrections are also the other factors affecting the quality of change detection analysis. All these issues may cause a great number of false alarms in addition to artifacts in change detection analysis with VHR images.