Recentring and selection of short-exposure images with photon-counting detectors.

The paper presents statistical considerations and numerical simulations which aim is to determine the minimum number of events (coming from a window encircling either a simple star or several stars used as reference for correction) necessary for 'reliable' recentring and selection. It is shown that the recentring and/or selection procedure introduces an inevitable 'over-resolution' on the profiles of the objects used as reference and a degradation of the nearby objects. Limiting the propagation of errors on the point-spread function resulting from the recentring procedure requires reaching about 100 events from the object(s) of reference. Such a number implies exposure times of a few seconds with the photon-counting detectors currently used for this purpose. Although not strictly inapplicable in practice, these exposure times are still too long for using the technique as its optimum. Much longer exposure times are required for selection, that limit its use to strongly variable conditions only. It is concluded that the best astronomical targets for which this technique is optimal are certainly star-rich fields where sub-second integration times can be applied with confidence.