A novel scheme for infrared small target detection in compressive domain is presented. First, the original image is separated into two components, i.e., the target and the background. Next, we compress them individually. Finally, the compressed target image is utilized to construct the corresponding compressive detector to perform detection in compressive domain. Compared to conventional detection methods based on compressive sensing(CS), it not only can detect infrared small target images under diverse backgrounds in compressive domain without any reconstruction of original signal, but also possess faster detection speed and favorable detection performance, and require much less storage and computation. Experiment results indicate the scheme is effective.