Abstract: Automatic object detection from natural scene clutters is fundamental to object detection. When the scale is changed, the fractal features of man-made objects in a natural scene change much more violently. In terms of the characterization, this paper presents a new method for automatic target detection using fractal maximum. Experimental results show that the method based on fractal parameter maximum map (FPMM) performs very well for small man-made object detection from a natural scene clutter.