Land cover change detection using saliency and wavelet transformation

Abstract

How to obtain accurate difference map remains an open challenge in change detection. To tackle this problem, we propose a change detection method based on saliency detection and wavelet transformation. We do frequency-tuned saliency detection in initial difference image (IDI) obtained by logarithm ratio to get a salient difference image (SDI). Then, we calculate local entropy of SDI to obtain an entropic salient difference image (ESDI). The final difference image (FDI) is the wavelet fusion of IDI and ESDI, and Otsu thresholding is used to extract difference map from FDI. Experimental results validate the effectiveness and feasibility.