Dawction of targets in in a by signal to clutter plus noise ratio (SCNR) is a problem of increasing interest. In this paper we present two ellaptive algorithms for the detection of smell targets (of the order of one pixel) in images using reference correlated frames (the reference frames can be obtained either from frequency bands of the same scene or from different sequential observation in time) in a by signal to clutter plus noise ratio (SCNR) environment (of the order of -14.5 dB). They both have the ability to track the nonstatimery image signals (targets and clutter plus noise) and suppress the clutter plus noise background. Both detictors are based on time varying autoregressive models to model image background and on correlation canceling concept. The first one uses an order recursive least squares (ORLS) lattice filter, stile the second one is based on a normalized version of the two dimensional least Mall square (TDLMS) algoridnu. The influence of the osier of the detectors on then detection performance is studied. The performence of the two algorithms are evaluated using an optical smell* image, as a clutter beckgrotuel, with compubr generawd target and noise added to it.