Abstract: The detection of small single-frame image is investigated in complicated natural background. A new multi-scale fractal feature is proposed according to the fact that the fractal feature of man-made objects changes shaper than the natural background. The new feature causes the man-made objects to stand out much better from natural background than what can be done by the standard fractal dimensions, thus inhibiting well three kinds of images from different background clutters. Local gray histogram statistics is applied to object detection in the images with multi-scale fractal feature. Experimental results showed that the algorithm to detect the objects with such a feature is quite stable and generally acceptable to IR and TV images in complicated ground or sea background and, especially, it can localize small objects in a single-frame image with an accuracy up to over 95%.