Mosaic Image Construction

Three mosaics are constructed for each galaxy, corresponding to the J, H and Ks bands. The mosaic image size is first derived from the optical diameter, D₂₅, scaled by roughly three times this diameter, thus allowing for ample background removal. The appropriate "coadd" images are then collected from the 2MASS database. For the largest galaxies, multiple, adjoining scans comprise the image collection. In addition to the images, the calibrated point sources are also gathered from the database, and are used in the background removal operation. The construction process entails bi-cubic resampling of the coadds onto a larger grid of 1 arcsec pixels, subtracting the background after removing stars and blanking the target galaxy, adjusting the photometric zero points for adjoining scans to match the central scan, updating the FITS headers and writing the final images. The most difficult procedure is the background removal, a formidable component of ground-based NIR imaging, particularly for the H-band because of the highly variable atmospheric OH emission (the typical aggregate sky brightness at 1.6 µm is 13 mag arcsec^-2).

The background must be removed carefully to avoid subtracting the galaxy itself, particularly the low surface brightness wings. The size of the galaxy is then a critical parameter toward satisfactory background removal. For the first iteration at background removal, we assume the size is given by the optical diameter. This usually provides an adequate margin for error since the optical diameter is typically larger than the NIR diameter, except when the foreground extinction is significant (i.e., galaxies projected near the Galactic Plane). The elliptical shape of the galaxy is given by the RC3 eccentricity and position angle. For galaxies that do not have RC3 entries, we use the UGC catalog or the 2MASS XSC catalog, which should be adequate for galaxies smaller than 2 to 3arcmin. The background or "sky" can then be isolated and removed after the galaxy is elliptically masked from the images; see below. After the preliminary mosaic is constructed, we determine the actual NIR size and shape of the galaxy, and adjust the parameters accordingly, then process/iterate again using the updated parameters.

Due to the nature of the ground-based 2MASS scan methodology, and the subsequent high sky brightness and fluctuations, which occur at all scale lengths, background subtraction is in practice very challenging in the NIR. We first construct a mosaic from coadd images belonging to the same scan/tile, since sky variations are continuous within the scan. Conversely, background variations are disjointed between neighboring scans, and overall background offsets must also be removed. The image below illustrates the daunting NIR background:

• M31 scans, as seen in the H-band

With stars and the target galaxy removed from the mosaic image, the remaining "surface" is then fit with a cubic polynomial. We use a running fit, where each pixel has its own solution, first along the in-scan (the long declination axis) and then along the cross-scan (short equatorial axis); the method is detailed in Jarrett et al (2000). A critical parameter is the scale length of the polynomial fit. It must be as small as possible to accurately track the airglow variations, but no smaller than the size of the galaxy itself to avoid over-subtraction. We have found that the optimum scale length is ~two to three times the diameter of the galaxy. For H-band, the most challenging 2MASS band, there is usually some remaining residual background due to the severe "airglow" in the 1.6 µm window. The background solution is then subtracted from the mosaic image, and the next adjacent scan is then processed in the same manor until the full mosaic image of scans is constructed.

• M31 scans, H-band background solution
• M31 H-band Final Mosaic

Basic image statistics are carried out on the final mosaic after removal of stars and the target galaxy itself. The FITS header includes the mean background and standard deviation, the astrometric solution, and the photometric zero point magnitude.

More information on the fitting procedure, see the M33 and NGC 2903 constructions.