Comparison of ITPCA and IRMAD for Automatic Change Detection Using Initial Change Mask

In change detection analysis, the computation of the no-change distribution is affected when changed pixels are in large number in the scene. Because of this, the performance of several techniques are compromised. In this paper we compare two well known automatic change detection techniques (ITPCA and IRMAD) by performing an initial elimination of the strong changes in order to minimize the contribution of the changed pixels to the radiometric normalization computation. These two techniques are ineffective in correctly estimating the distribution of the no-change pixels when this kind of scenario is encountered. The strong changes are identified by building an initial change mask (ICM), which is based on the statistical analysis of the given data set. In this paper we show two simple algorithms for building the ICM. From the experiments on a data set characterized by a high amount of changes due to the agriculture activity, the improvement in quality of the map of changes obtained by the proposed approach with respect to the ones obtained without using the ICM has been observed.