6 Concluding remarks

The method assumes that the true photometric structure of the observed source is formed by two shifted and weighted delta functions, and that a set of short PSFs with statistically significant size do not present a marked trend during its acquisition. Also it uses an analytical approximation of the smooth light distribution in the short PSFs to fit short images, obtaining a rough estimate of the photometric parameters but a good determination of the astrometric ones. This allows us to derive by shift-and-add the mean short exposure image, characterized by an improved signal and by the near diffraction-limited resolution, since the image motion of the set images has been removed. Then the mean short image yields by a fit the final adopted features of the binary system and by a deconvolution the detailed mean PSF. The tests and the simulations performed give evidence of the robustness of the present method and of its general capability to resolve Adonis short images of close binary stars with luminosity ratio and relative position of the components spanning a reasonable range; they also enable us to understand the method's limitations and the few critical cases, thus strengthening the reliability of the features of $\tau$ CMa reported in Table 3. It is worth mentioning that the present method, thought to solve a particular problem, may be easily extended to study multiple systems and even crowded stellar fields, provided that in each case the adopted PSF parametrization adopted represents a reasonable compromise between the actual point source image and that corresponding to the full AO compensation, i.e., the diffraction pattern. For instance, dealing with the above quoted images by PUEO (see Sect. 2.1) we must use a globally elongated PSF approximation, with an elliptical core and several confocal elliptical annuli, as will be reported in a future paper. In conclusion, this method of analyzing a set of AO images is fast and robust, and its results are comparable with those obtained by IBD, which is much more time and computer memory consuming. Furthermore, it can run either in an interactive way, as done here, or automatically, since the fundamental steps (search of initial parameters, convergence attainment, image coaddition, agreement or correlation between fitted and observed images and so on) can be controlled by suitable software instructions in an expert program, thus it is amenable to an automatic iterative procedure.

Acknowledgements

The Bologna group would like to thank J. Christou for helpful discussions on IBD and for providing its IDAC code as well as images and other unpublished material. The High Resolution Data Reduction Working Group at ESO is acknowledged for punctual information on its activity and current research efforts. We also thanks David Le Mignan and the 3.6 m telescope team at La Silla for their help during the observations. We are indebted to G.M. Stirpe for correcting our manuscript and to the referee (E. Thiébaut) whose constructive criticism has much improved the presentation of this paper. This work was supported in part by Ministero della Universitá e della Ricerca Scientifica e Tecnologica (MURST).