There is a tradeoff between effective and computational costs in traditional change detection methods. Reducing the computational cost of the processing time results in loss of quality of detection results. We propose a new change detection method based on the temporal difference principle. The new method combines the connected component labeling, closing operation of image morphology, and human structure analysis to remove shading and noise in consecutive frames. A closed object boundary and region can be extracted with only a small additional cost in processing times. Additionally, this new method recovers the disadvantages of traditional temporal difference methods. Comparative results, in terms of computational complexity, completeness of the object and the object's outline, robustness against noise and shadow, with currently popular methods and the proposed method are also displayed. The experimental results show that the proposed method is robust in varied illuminations and effective in several predefined environments for change detection.