An automatic pattern recognition method for the processing of long slit spectral images

A pattern recognition method for the reduction of stellar and nebular spectra is presented. This method automatically corrects the geometrical and optical distortions of long-slit spectra and more specifically of bidimensional photographic image-tube spectra. Any photographic plate surrounded by two comparison spectra can be processed. The presented spectral reduction method is fully independent of the spectrograph and of the image-tube characteristics; no a priori knowledge of the different distortions is required. SPECTRA software, developed on VAX 11/780, has been included in the ESO Observatory´s image processing system (MIDAS), and was successfully applied to plates originating from the Haute Provence Observatory (OHP), the ESO Observatory at la Silla, and the Canada-France-Hawaii telescope