About the Questions in the Paper

[slchenchn]

Hi, I am reading your paper associated with this reporeportd and I have a problem. 
In the Sec. II ( RELATED WORK ), "unlike scene change detection, it is unnecessary to estimate correspondences between input images and the change mask because the change regions between the input images are common. However, for street-level scene change detection, the estimation is necessary because scene objects can appear, disappear, and move", could you please describe the differences between scene change detection and remote sensing change detection in detail? There are also moving objects with remote sensing images, why only the scene change detection needs to estimate correspondence?