The images were flatfielded by the observer (P. P.) using a combination of
dome and sky flats. Cosmic
rays were then removed with the FILTER/COSMIC procedure in the
MIDAS reduction package.
The sky background was then fitted using FIT/FLAT
[3]_SKY
with parameters set such that the background was approximated by a tilted
plane, preventing the inclusion of higher order effects due to the sky
subtraction itself.
One of the greatest challenges with dwarf galaxy photometry is to cope with
the extremely low signal-to-noise ratio, as even the central surface
brightnesses of these galaxies often represent only 
of the
night sky brightness.
The night sky itself then introduces a considerable amount of Poisson
noise comparable to the galaxy signal (see below).
A very good method to improve the signal-to-noise ratio, and hence the accuracy of the photometry, is to employ azimuthal averaging of a galaxy image around a pre-chosen centre. This can be done by calculating "growth curves'' (GC) where the light is integrated in concentric circles of increasing radius, corresponding to a simulated aperture photometry. The (azimuthally averaged) surface brightness profile then results from differentiating back the GC (see Sect. 4.1 (click here)).
Growth curves were constructed in one-pixel-steps centred on the galaxy centre using INTEGRATE/APERTURE, after nearby or overlaying stars had been removed.
The removal of stars from the frames was done with MODIFY/AREA. This procedure replaces a pre-defined area with a constant, a plane or a second order polynomial surface fitted to the surrounding areas depending on the parameters set. A proper point spread function fitting and removal of the bright stars was made impossible by the non-rotationally symmetric psf.
The galaxy centres were determined by fitting elliptical isophotes to the faint outer parts of the galaxies with FIT/ELL3 so as to centre on the old population, or by using an intensity weighted 2-dimensional centring procedure CENTER/MOMENT where the ellipse fitting was too perturbed to be used. In a few cases the centring was done by eye, for the very irregular galaxies.
A further great advantage of the GC-method is that one can easily
control the sky subtraction by noting that the GC should be
asymptotically flat at large radii if the background is correctly
removed (Binggeli et al. 1984). The sky subtraction is critical for
the correct
photometry of the dwarfs, due to the extremely low signal compared to
the background. Indeed, a typical sky background value is 
in the B band, whereas a typical value
for the central surface brightness of the galaxies is
and the outskirts
of the galaxies are here traced down to approximatively 
. The accuracy of the sky determination
as well as its final flatness must on average be, in a region around
the dwarfs, well with in fractions of a percent, typically 0.05% of
the original sky background level, if one is to obtain GC's that
flatten out correctly at large radii.
Setting the magnitude scale was done using standard stars from the
photometry of 
(De Vaucouleurs et al. 1994). These are
situated within frames centred on the core of M 81, which was
imaged twice per night. Aperture photometry of the stars was done
with the MAGNITUDE/CIRCLE procedure. As the extinction curve
was under-sampled, the standards were reduced to an airmass of 1.2, so
as to minimise the errors caused by the uncertainties in the slopes of
the extinction curves.
Due to the non-photometric conditions and the absence of standard star
sequences, the main contributor to the errors remains the
uncertainties in the photometric zero point determination. The
calibrations rest upon aperture photometry of stars superposed on
M 81. It appears realistic to estimate a magnitude scale
error of up to 0.2 mag for our objects. This is in accord with the
results of Paper II (see Paper II), which show deviations in 
(see Sect. 4.2 (click here), this being the most reliable comparison
due to the different reduction methods employed) of the order of
in B and 
in R. See Sect.
4.1 (click here) for typical errors on the radial profiles due to the
low signal-to-noise.
All the magnitudes in this paper are corrected for galactic absorption
using the 
law given in Sandage & Tammann (1987) for the
B band, which we transformed into the R band by way of the
interstellar absorption curve (e.g., Mihalas & Binney 1981):
Typical values for the region around M 81 are AB = 0.08 mag and AR = 0.05 mag, hence the effect is small.