Whereas data quality is of paramount importance and obviously limits the angular resolution that can be ultimately obtained in the reconstructed images, the nature of the data reduction methods subsequently employed to extract the scientific information contained in these images plays a key role in ensuring the overall success of the scientific programs.
The ``Aperture Synthesis'' team at OMP has been mainly involved during the last few years in the theoretical aspects of aperture synthesis and related problems such as deconvolution, wavelets and multi-resolution methods (Lannes et al. 1987; Lannes 1988, 1989, 1991). Our approach to these problems is deterministic and based on a least-squares scheme that allows error analysis, hence a good understanding of the stability of the image restoration process. The initial theoretical work on single aperture interferometry, i.e. speckle interferometry, has been successfully extended to multi-aperture devices (Lannes 1989, 1991, Anterrieu, 1992)
We have written some specialized software to process data from the various detectors we have used (cf. §4), and the different observing modes: speckle imaging, aperture synthesis with pupil masks, speckle spectroscopy, and coronagraphy.
For speckle imaging, a few observers have independently reached the conclusion that ``the bispectrum combined with a constrained iterative deconvolution of amplitudes produces the highest quality imagery'' (Beletic and Goody, 1992). Nevertheless, we have used various programs ranging from Knox-Tompson (1974) to full bispectrum methods (Weigelt, 1977, Roddier, 1986, Lannes, 1989) (and even partial bispectrum methods, i.e. using only a subset of all possible closure relations) and found little differences, if any, on the restored phasor image of double stars. The pre-processing of the original data (correction of geometric distortion, of flat-field, and various calibrations) is for us the crucial step in the whole image restoration process.