The entire data processing was done by using a PC 486/DX2. The software, written on PASCAL, was designed to process subframes centered on the stellar images and consists of two basic procedures - cleaning of close companions and instrumental magnitudes derivation.
To suppress stellar companions Newell (1979) has proposed a symmetry-clean
procedure, which operates by comparing the intensities of all pairs of
pixels diametrically opposite with respect to the image center. The author
included this procedure into the centering procedure aiming to increase the
speed and reliability in crowded field centering. Supposing a circular symmetry of the
stellar image and taking into account the variation of the grain noise over
its profile (Lee & van Altena 1983) we have developed another mechanism
to clean close companions. First, the image was divided into a few contiguous
annular regions with a width of about one pixel. The last few annuli should
be sufficiently far from the center of the star - several times the FWHM of
the image - so that its light contribution to the background determination
is negligible. If the image is not damaged by close companions the density
distribution within each annulus will be symmetric with coinciding values of
the mode, median and mean. In the opposite case defining the mode as the
most frequently occurring density value in a given annular region we stated
the following principle of cleaning: any pixel value exceeding the mode with
some predetermined multiple of the sample standard deviation, 
(where k is a real number), was considered as ``improper'' for the
annular distribution under consideration and replaced with the mode value.
The pixel density histogram of the regions in question is usually not
actually calculated, instead the mode of the distribution is estimated
from the formula (e.g. Kendall & Stuart 1977, p. 40):
provided that the median is less than the mean. Otherwise, the mean can be
taken. A comprehensive discussion of using the mode as the most representative
value of a sample of density values one can find in Stetson (1987) and
Da Costa (1992). Practically the scheme works as follows: first, to determine
the median and mean any strongly deviant pixel density within each annulus
is clipped iteratively until 
stabilizes; second, the mode of the distribution
is calculated and than the outlined cleaning procedure is applied. The procedure
is fully automated and does not need any investigator's interaction. Here we
have to emphasize that the efficiency of the cleaning procedure depends on
many factors and mainly on the number of companions and their closeness.
Figures 3a and 4a illustrate the effect of applying the above described
procedure in the two dimensional plane for two different cases. The frames
are centered on the stars under consideration. In both cases, the original
image is presented on the left but its cleaned version on the right
panel. In addition, we present in Figs. 3 (click here)b and 4 (click here)b profiles of
the same stars. These are not a simple image cross-section in a given
direction but rather integrated profiles which include all points found
within the image - it is a plot of the densities found on a given radius. The
presentation of the original and cleaned profiles are the same as for the
density maps.
The image cleaned in this manner was used for magnitude index formation both through a multi-component bi-dimensional Gaussian fit (using data in photographic densities) and with a synthetic square aperture with constant size (in relative intensities). The size of the aperture is 4 pixels in each dimension. The application of a one- or bi-dimensional Gaussian fit (GF) to determine MI as well as the application of synthetic aperture (SyA) photometry are well documented in the literature (Stetson 1979; Buonanno et al. 1979; Stetson 1990 etc.) and will not be discussed here. DAOPHOT photometry was performed in a standard way and the derived magnitudes will be refereed as ``PSF magnitudes'' hereafter. To complete an unique system of instrumental magnitudes, the MIs of the same colour were averaged together and then the individual plates were transformed to the average plate using a least-squares polynomial. These fits permit the mean error of a MI on a typical plate to be estimated from the transformation residuals (Stetson 1979):
Figure 5:
Mean error of the instrumental magnitudes for the two plates in the
B band as defined in Sect. 3.2. The lower panels present the comparison of
instrumental magnitudes determined with a synthetic apperture with constant
size for isolated stars (left) and stars
with close companions (right). The upper panels - the same as the lower
but for magnitudes derived with bi-dimensional Gaussian fits of the stellar
images. The observed trend of the errors on the right panels is the same as on the
left. This fact shows that
the applied cleaning procedure does not bias the cleaned image in any way
Figure 6: The same as Fig. 5 (click here) but for V magnitudes
Figure 7: Mean error of the instrumental magnitudes as an ouput from DAOPHOT
for two plates in B (left) and V (right)
where n is the number of plates used but 
is the average of the transformed
magnitude of the i-th star measured on n plates. Figures 5 (click here) and 6 (click here) demonstrate
the mean error in the B and V bands for a Gaussian fit (upper panels) and
SyA magnitudes (lower panels), respectively. The left panels show the mean
error for isolated stars, and the right pannels - the same for stars in contact.
We prefer to demonstrate both kind of instrumental magnitudes because their
determination differs. In Fig. 7 (click here) are presented
the mean error in the B (left) and V (right) bands
for PSF-magnitudes (DAOPHOT output).