Abstract: An infrared moving object detection method which combines Reichardt-type two dimensions Elementary Motion Detectors inspired by biological vision and region growing method was proposed to solve the two dimensions EMDs' sensitive problem in dynamic scenes. EMDs detected the most intensive motion vector signal in temporal domain which was then used as the seeds of the region growing. Region growing method was applied to make a segmentation of the target by its infrared radiation characteristic much different from background in spatial domain. The simulation illustrates that, combing the EMDs in temporal domain with the region growing method in spatial domain achieves much better detection performance in infrared frames than the original Reichardt's model. Compared with other methods, the proposed method could achieve a higher SCR.