We present two general methods, the so-called Locust and the generalized Hough algorithm, to search for narrow-band signals of moderate frequency evolution and limited duration in datastreams of gravitational wave detectors. Some models of long gamma-ray bursts (e.g. van Putten et al 2004 Phys. Rev. D 69 044007) predict narrow-band gravitational wave burst signals of limited duration emitted during the gamma-ray burst event. These types of signals give rise to curling traces of local maxima in the time–frequency space that can be recovered via image processing methods (Locust and Hough). Tests of the algorithms in the context of the van Putten model were carried out using injected simulated signals into Gaussian white noise and also into LIGO-like data. The Locust algorithm has the relative advantage of having higher speed and better general sensitivity; however, the generalized Hough algorithm is more tolerant of trace discontinuities. A combination of the two algorithms increases search robustness and sensitivity at the price of execution speed.