A new algorithm based on third-order cumulant was proposed to detect a moving dim infrared target under heavy background.Because the grey-scale value of a pixel in an image has fluctuation when a target passes by,this fluctuation can be viewed as a non-Gaussian weak transient signal,and its third-order cumulant can be calculated to detect this signal.The process of average-value-subtraction in the estimation of the third-order cumulant also suppresses the background clutter,and leads to a great improvement of the signal to clutter ratio(SCR) and the probability of the detection in single frame.The experimental results show that the algorithm can effectively suppress heavy clutter background and reliably detect dim infrared targets when SCR is larger than 1.