Compared with other targets, it is more difficult to detect infrared small targets due to several aspects such as the low signal to noise ratio, low contrast, small size, the lack of shape and texture information of the targets, especially under complex background. A novel infrared small target detection method based on peer group filter (PGF), bi-dimensional empirical mode decomposition (BEMD) and local inverse entropy (LIE) was proposed to overcome these difficulties. The PGF is implemented to remove noise and improve signal-to-noise ratio of the initial image. The proposed BEMD algorithm is able to estimate background effectively, which gets target image by removing background from original image and segmenting the Intrinsic Mode Functions (IMFs) by local inverse entropy. Experimental results demonstrated that the novel method can extract the small targets validly and accurately.