Abstract: In order to detect small targets with changing size in infrared scene, aiming at the problems in traditional filtering method with fixed size filter, a method for small and dim infrared targets detection based on scale-space theory was proposed. First, the target characteristic was analyzed and point spread function form was used to represent the target model. Then, in order to suppress background clutter and enhance the power of target, fixed adaptive neighborhood method was used in image preprocessing, on the basis of Laplace scale-space after scale standardization which has different filtering responses for different elements, the suspicious targets were obtained in the images, then with the gradient relationship between suspicious target point and its surrounding pixels, the coordinates of the suspicious targets centers and its size were got; Finally, each suspicious targets gained an adaptive window to obtain segmentation threshold and true targets. Experiments results show that, compared with traditional methods, the new method proposed in this paper has a better performance to detect small targets, and has a lower false alarm rate.