In infrared (IR) guidance, early warning and other fields, it is of great theoretical significance and application value to detect IR small target with high detection rate, low false alarm rate and high speed. An IR small target detection method based on the tri-layer window local contrast was proposed. The tri-layer window could deal with small targets of different scales by single-scale calculation, so that the detection speed could be accelerated. Meanwhile, the enhancement on the true target and the suppression on the complex background were considered before, during and after the local contrast calculation, so that the detection rate could be improved and the false alarm rate could be reduced. Experiments in some IR sequences and images show that, compared with eight existing algorithms, the proposed algorithm can achieve a better performance on detection rate and false alarm rate, and its average time consumption is only about 1/3 to 1/2 of some multiscale algorithms.