2.2. Data analysis
All the images were corrected for bias and flat-field using standard IRAF procedures. The relative photometry was carried out with the PSF-fitting code DoPhot (Schechter et al. (1993)). Each frame was independently reduced. A 3 threshold above the background noise was adopted for the search for sources on the frames and the spatial variations of the PSF were modeled with a quadratic polynomial function. Only the sources classified as stars by the code were retained. All the V and I catalogues of a given field were reported to the (instrumental) photometric system of the best-seeing image acquired under photometric conditions. The catalogues were then cross-correlated, the magnitudes were averaged and the standard deviation adopted as the photometric uncertainty of the individual stars. In the final catalogue of instrumental magnitudes and positions we have retained only the sources that have at least three valid measures of the magnitude for each passband (e.g. at least 3 V and 3 I). Moreover, all the stars with an associated error (either in V or I magnitude) larger than three times the average uncertainty at their magnitude level were excluded from the sample. The final F1 and F2 catalogues contains 26399 and 1937 sources, respectively. Accurate aperture corrections have been obtained for each field on a few tens of bright and isolated stars.
The stellar crowding is quite low in the considered fields that sample external low-surface brightness regions of the galaxy. Even in F1 the average stellar density is as low as 0.08 stars/arcsec2 (considering all the detected stars). For the applications presented in this paper the effects of incompleteness are not an issue since we always compare subsets of stars that are homogeneous in magnitude and color (see Sect. 3.3 and 3.4 below). No attempt is made to determine the true Luminosity Functions of the observed sequences. The TRGB level, the main target of the present study, occurs more than 3 magnitudes above the limiting magnitude, a range in which the considered sample is likely ~ 100% complete. Since the photometric uncertainties of individual stars (in each passband) are empirically estimated as the standard deviations of 3 to 8 independent measures of the magnitudes there is no need for artificial star experiments to characterize photometric errors.