Working in the transform domain opens a fan of possibilities like: testing of several kinds of filters, perturbation of the signal of the transformed image (real and imaginary parts) to search for the occurrence of specific features, and the application of different scene object models. The reader is referred to the works of Mahalanobis et al. (1994); Javidi et al. (1994); Juvells et al. (1994); Javidi & Wang (1995); Martucci (1996); Javidi & Painchaud (1996). For analysis and restoration of direct images and spectra, there are the works of Brault & White (1971); Lucy (1974); Högbom (1974) and Baade & Lucy (1990). The case of Wiener filtering, for instance, does not take into account the spatially variant statistics (the stationary hypothesis) and is minimum mean-square-error constrained. The transfer function of the Wiener filter has a large magnitude at those spatial frequencies for which the signal power is large if compared to the noise power. These facts are partially responsible for the smooth look of the so restored images, in contrast to the human eye, which accepts more noise whenever associated to sharp intensity changes (STSDAS help page on "restore" and Arp & Lorre 1976). Incomplete deconvolution due to the cutoff of certain high frequencies may form haloes around bright regions against darker background.
Deconvolution using the point spread function of frame stars in the field of the galaxies is also useful in the search for structures in galaxies (Zavatti et al. 1990). Some artifacts introduced during the deconvolution of images of E-S0 galaxies have been reported by Michard (1996), but this can happen during any step of the processing in the transform domain. Some other ideas about deconvolution techniques can be found in the works of Yaroslavsky & Caulfield (1994) and Thiébaut & Conan (1995).
The idea of computer-generated holograms is recent (Mori & Ohba 1994; Nagashima 1996, and references quoted therein), and so is the digital recording and mathematical reconstruction of Fresnel holograms (Schnars & Jüptner 1994). The computer-generated Fourier holograms of the present work are simple; they have been built from the original images to create a transfer function for feature enhancement. Problems of artifacts may be present, also because the traditional methods generally require many iterations of the Fourier and inverse Fourier transforms (see Gabel & Liu 1970; Hauck & Bryngdahl 1984; Ersoy et al. 1992 and Chang & Ersoy 1993). See Schnars & Jüptner (1994), Mugnier (1995), Dorsch et al. (1994), and Nagashima (1996) for interesting reading.
We have shown that the mixing of different procedures may be necessary to enhance morphological details of extended objects. Care must be taken when choosing the parameters so that artifacts are reduced to a minimum. Creating transfer functions and filtering images are easily and quickly performed in the STSDAS and IRAF environments. We believe that these operations performed in the transform domain may improve automated galaxy classification. Improved filtering techniques could also be useful to enhance cooling flows' signatures in galaxy clusters (see Friaça 1996 for detection problems), and, why not, contribute to the effort of the primeval galaxy search (see Pritchet 1996).
Detailed studies on HRG 54103 will be presented in a forthcoming paper (Faúndez-Abans & de Oliveira-Abans 1998).
Acknowledgements
This work has been partially supported by the Conselho Nacional de Desenvolvimento Cientifico e Tecnológico (CNPq, Brazil). This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, Caltech, under contract with the National Aeronautics and Space Administration. We would like to thank Dr. J. Sulentic for his comments and interesting suggestions to the early version of the manuscript.