Unsupervised Global Difference Modeling for Image Change Detection

Wenhua Zhang, Qihao Weng, Jia Liu, Zeng Qu, Yong Li, Jinjin Chai, Liang Xiao

Abstract: In change detection, impact of nonintrinsic changes such as those caused by illumination, season, and viewing angle variances are common in practice but also a great challenge for change detection methods. In this article, we propose a novel unsupervised image change detection method by modeling global difference information to deal with such nonintrinsic changes. Comparing global features can mitigate the impact of them due to the global consistency of them in the same scene at the same time. But global modeling for change detection also faces the challenges of feature learning with limited data and difficulty in generating pixelwise changed regions. To overcome the challenges, first, we use a backbone network to capture the global features of bitemporal images. Then, an energy function is designed with a masked difference between the two features and a margin-aware constraint in order to align the global features and meanwhile maintain detail information. To train the network with only two images, we propose an adversarial learning method by introducing a generalization network that consecutively generates two images that can minimize the energy. Then, a new loss function is derived to alternately train the feature learning network and generalization network. Second, after learning with bitemporal images, it is also important to generate the pixelwise changed regions. Then, we design a difference mapping method that maps the changed regions from global difference. Experiments on different types of data by comparing with both supervised and unsupervised methods demonstrate the effectiveness of the proposed method.

External IDs:doi:10.1109/jstars.2025.3588154