The effects of image misregistration on the accuracy of change detection for high-resolution remotely sensed images were quantitatively investigated on the basis of multi-temporal geometrical registration. The relationship of misregistration respectively with bands and land complexity were firstly analyzed and evaluated. The sensitivity of each band and the difference among each region with different land complexity were then analyzed when misregistration errors are the same. The experimental results using QuickBird true-color images of four regions indicated that the misregistration had different impact on each band, and had more evident impact on regions of high land complexity with more false information generated.