A new efficient algorithm for moving objects detection in dynamic background was put forward. Registration between consecutive frames was performed by using an improved SSDA algorithm based on full use of image features. After registration between frames, sum of absolute difference of pixel intensity between consecutive frames in range of every feature and its neighboring window was calculated and the position of moving objects was determined according to the rule that the feature from moving regions shows a lager registration error. Experiments show that the algorithm is simple and fast, which can be used for detection of the moving targets in dynamic background as well as change detection.