Abstract: According to the imaging characteristics of infrared variable resolution dim target in cloudy background, a method for small target detection based on scale space theory was proposed in this paper. Firstly, a Gaussian parameter model was established for small target, and then the scale space of the detected image was generated by using Gaussian kernel. Next, with optimal scale matching for scale space, the target region of interest (ROI) was extracted. Finally, the scale adaptive contrast (SAC) of ROI was calculated as detection criterion to remove the false alarms, achieving small target detection. The results of experiments show that the method proposed in this paper can not only accurately detect the target, but also estimate the size of target, and has high detection rate and efficiency.