Change detection for remote sensing images based on wavelet fusion and PCA-kernel fuzzy clustering

A change detection method is proposed to improve the robustness, detection accuracy and noise immunity. Wavelet fusion is employed to combine the difference image obtained by subtraction operator with that obtained by ratio operator. Then, the fused image is partitioned into non-overlapping blocks, and an orthonormal basis is extracted from them through principal component analysis (PCA). Each pixel in the fused image is represented by a feature vector which is the projection of neighborhood patch onto the orthonormal basis. Finally, the change detection image is achieved by clustering the feature vectors using kernel based fuzzy C means (kernel-FCM) clustering algorithm. Experiments show that the strategy of image fusion enhances the robustness of the algorithm when compared with those based on single difference image, and kernel-FCM improves the accuracy further. In addition, due to the use of feature extraction technique, the method performs well on combating noise. ©, 2015, Chinese Institute of Electronics. All right reserved.