An automatic image segmentation method applied to real-time image tracking in complex environment is proposed. In accordance with the variance of scenes

this method can adaptively make alternate use of two work modes based of scene complexity: thresholding and hierarchically clustering. Meanwhile it suppresses the interruption of background noise as much as possible. Experimental results show that the presented approach is comparable in segmentation performances to several segmentation method currently considered as better ones

but its adaptability and usable range are much superior to

and its computing cost is usually lower than the latter.