For the difficulty of moving object detection by a moved camera, a method for detecting dynamic background caused by camera motion was proposed by combining a Scale Invariant Feature Transform(SIFT) and a differential multiplication. Firstly, the image registration based on SIFT features was applied to calculate transformation parameters, including translation factors, rotation angles and scaling coefficients,and to provide a robust matching for realizing the motion compensation precisely. Then,the object segmentation based on the differential multiplication was introduced to detect moving objects accurately. Finally, the experiments on the video sequences were performed to verify the robustness and validity of the method.Experiments show that instead of 6 or more frames, 4 frames are enough for the differential multiplication in practical applications. The detection method is robust and denoising in dynamic scenes and has good adaptability to changing illumination, occlusion and camera movement.