A Multiscale Fusion Approach for Change Detection in SAR Images

Best performance and greatness in precise changes are vital factors of change detection. The proposed method is mutual task to deal about preprocessing and change detection of multitemporal SAR images. In preprocessing stage, Speckle Reducing Anisotropic Diffusion is implemented in each layer of multiscale pyramid transform. The speckle free images are interpreted by Absolute difference method and XOR operator to retrieve primary difference image. After that desired change detection is fused by laplacian pyramid coefficient. Fused difference image incorporates the advantages of absolute difference and XOR operation. Finally robotic threshold algorithm of Otsu is used to predict exact change detection. For experimental purposes two data sets are preferred from Envisat and TerraSAR-X images. Standard quality has been evaluated on the proposed method to quantitatively prove the performance.