Brief

Researchers have developed HashSCD, a novel framework for scene change detection that utilizes patch-wise image hashing. This method allows for efficient identification of changes within images by encoding spatially aligned patches into compact hash codes. HashSCD enables both global and localized change detection directly in Hamming space, reducing computational costs and storage needs compared to existing methods. The unsupervised contrastive learning approach demonstrates competitive performance against state-of-the-art techniques. AI