Threshold segmentation is widely used in the image segmentation, and the typical method is Otsu method. However, the Otsu method can not segment the low contrast or low signal to noise ratio image exactly, so a new dynamic threshold segmentation algorithm is proposed. The dynamic background image of the original image is obtained by the smoothing filter, the original image minus the dynamic background is the difference image, and the original image segmentation is achieved by setting the threshold. The experimental results show that the method has low algorithmic complexity, can segment successfully the objects which Otsu method can not segment and reduce the noise effectively, so it is suitable for a wide range of image segmentation.