Change detection for mutil-temporal remote sensing images based on NSCT and hierarchical clustering

Change detection has many applications in remote sensing, such as urban development, environment and damage monitoring and so on. Some typical methods are difficult to maintain detail information and the detection accuracy is also not satisfied. In this paper, a detail-injecting algorithm conducted by non-subsampled contourlet transform (NSCT) and hierarchical clustering is presented to preserve the detail information and increase the separability of the intermediate classes to improve the accuracy. The strategy of detail-injecting based on NSCT is to extract the detail information, and then inject the detail to difference image. After that, the residual image which have been highlighted by histogram contrast (HC) model is used as input of the strategy of hierarchical clustering to obtain the final result. Compare with some tradition methods, the experiments indicate that the proposed outperforms others in detection accuracy for remote sensing image.