We describe the development and the implementation of a new nonlinear image processing methodology to detect dim targets in infrared images. The methodology is based on mathematical morphology. In a first phase, the image is filtered in order to enhance positively contrasted, isotropic objects with a specified size. Then, a simple motion analysis is carried out, and finally, detection is performed by thresholding. A complete study of the algorithm performance on NATO test sequences is presented. A comparison is made with the Holmes filter and the scale-subtraction filter, a wavelet-based method.