SAR image change detection based on saliency region guidance and SIFT keypoint extraction

Lu Wang, Bailiang Sun, Chunhui Zhao, Suleman Mazhar, Tomoaki Ohtsuki, P. Takis Mathiopoulos, Fumiyuki Adachi

Abstract: Highlights•A saliency guidance fusion strategy is proposed. We determine saliency regions and leverage clustering information from both contrast and spatial cues. This effectively suppresses background and noise in SAR images, thereby reducing their interference in subsequent detection processes.•We propose a collaborative enhancement mechanism that integrates SIFT keypoints with saliency mapping, leveraging the keypoints’ concentration in significant gradient changes, as well as their scale-adaptive property, to effectively complement the saliency intervals.•We propose a novel patch-level pseudo-label processing and spatial structure preservation mechanism. Its advantage is exploiting ViT’s patch-based image processing capability, the method efficiently models regional samples of different pseudo-label categories. Combined with positional encoding, retaining spatial structure.

External IDs:dblp:journals/pr/WangSZMOMA26