2 Observations and data reduction

The thirteen films of the South-East area of the Virgo cluster of galaxies were taken with the UKST between 1991 and 1994 using the Kodak TechPan emulsion with the OG590 filter. The filter has a lower wavelength cut-off around 590 nm (10% transmittance at 580 nm and 90% at 600 nm), while the emulsion has an upper wavelength cut-off around 690 nm. This particular combination of filter and emulsion has a very small colour term compared with standard R band (Phillipps & Parker 1993). The field of the plates covers an area of $6.2\hbox{$^\circ$} \times 6.2\hbox{$^\circ$}$, affected by vignetting. The films used, dates and exposure times, are given in Table 1. The first eight films were from the UKSTU archive, while the remaining five were taken specifically for this project. A sixth film (OR16137) was also taken for this project, but has double images and so was not used.

  

Table 1: Details of the films

The $\approx\! 6.2\hbox{$^\circ$}\times 6.2\hbox{$^\circ$}$ area of the films were scanned using the APM machine at the Royal Greenwich Observatory, Cambridge, at a pixel size of 32 $\mu$m ($\approx$2$\hbox{$^{\prime\prime}$}$). The size of the digitized files was approximately 10500 $\times$ 10500 pixels, which occupied about 222 MBytes.

The thirteen raw frames were combined into a single co-added image using applications from the Starlink KAPPA and CCDPACK packages (see Starlink User Notes 95 and 139, available on the World-Wide-Web through the Starlink home page at http://star-WWW.rl.ac.uk/). Some of these applications were modified to handle the very large image size by breaking the input image up into "chunks'' of $512\times512$pixels for processing. The facilities of the NDF subroutine library were used to do this (see Starlink User Note 33). The following steps were involved in creating the final co-added image:

• Conversion of the tape FITS files to Starlink NDF format.
• Alignment of the thirteen frames. For each frame, the centres of a common set of twelve stars located towards the corners of the images were found using a centroiding technique. Transformations were then found which mapped the star centres from each frame onto the corresponding star centres in the first frame. These transformations were of the form: where the coefficients C₁ to C₆ were chosen to minimise the sum of the squared residuals between the transformed star positions $(x_{\rm out}$, $y_{\rm out})$ and the corresponding star positions in the first frame. The typical values found for the coefficient values and the values corresponding to already aligned frames can be found in Table 2. A simple linear transformation was used even though the images cover a large area of the sky. This was possible because the original images were mis-aligned by a very small amount.

  
  

    

  Table 2: List of coefficient factors

  
  
  
  

  The maximum shift obtained from these values does not exceed fifteen pixels for a frame of $10500\times10500$ pixels. The accuracy of the alignment is less than a quarter of a pixel.

• Photometric normalisation. The aligned frames were then compressed by replacing each box of $21\times21$ pixels by the mean of the pixel values within the box. An exposure factor and zero point offset were then found for each of the compressed images using the makemos application from the CCDPACK package. The full sized images were then modified using the zero point offset and exposure factor for the corresponding compressed image by subtraction of the zero point offset, and division by the exposure factor. This results in the images having a common mean photometric calibration.
• Co-addition. The thirteen frames were then combined into a single stacked image by finding the median of the thirteen corresponding input pixel values at each output pixel. The median was chosen rather than the mean in order to eliminate, as far as possible, artifacts such as satellite tracks and scratches which do not occur at the same places in all thirteen images. RA and Dec. information were also stored in the final co-added array.
• Correction for the effects of vignetting. A procedure was developed to reduce the effects of the vignetting caused by the telescope in the periphery of the stacked image. The stacked image was first compressed by a factor of 21 in order to speed up the calculation of the correction. All features significantly brighter than the local background level were then removed, and the resulting image partitioned into two images; one containing all the areas judged to be unaffected by vignetting, and another containing all the vignetted areas. Smooth surfaces were then fitted to each of these two images, and used to estimate the background levels at every point in the vignetted regions. Thus the first fitted image contains an estimate of the unvignetted background level extrapolated from the unvignetted regions, and the second fitted image contains an estimate of the vignetted background level. The ratio of these two images was stored as a correction factor image and expanded to the full image size using bi-linear interpolation. The vignetted areas of the stacked image were then multiplied by this correction factor image, leaving the unvignetted areas unchanged.
• Removal of stellar features. When the stacked image is compressed, the stars in each $21\times21$ pixel bin are combined together and result in an extended surface brightness in the compressed image which depends on the number of stars in each bin. To remove this artificial surface brightness, a procedure was developed which automatically removed all stellar-sized features from the full-sized stacked image, prior to compression. This procedure band-pass filters the image to attenuate heavily all features except those of stellar size. Bright features in this image were identified and the corresponding pixels in the original image were removed by setting them to a value which indicates that the the pixels should be ignored. This basic procedure was refined to retain foreground stars which are coincident with extended background features such as galaxies.
• Gaussian filtering. The data area containing the Virgo galaxy M 89 was smoothed by a two-dimensional Gaussian function with a full-width at half-maximum (FWHM) box of $40\times40$ pixels. The resulting image was subtracted from the original data, so any "sharp" features (like shells, jets, filaments, etc) can now be clearly seen.

Multiple R-band CCD exposures of a $\approx 6^{\prime} \times 6^{\prime}$ field containing the IC 3481 and IC 3481A galaxies were taken in photometric conditions on 1995 Dec. 22-23, using the CCD camera with Tektronix CCD on the 1.0 m Jacobus Kapteyn Telescope (JKT), La Palma. The total exposure time was 70 min and the resultant co-added exposure is about as deep as the co-added array in terms of surface brightness. Standard applications from the Starlink KAPPA and CCDPACK packages were used for the reduction of this data and the resulting image (along with others) will be used for calibrating the co-added UKST data. Preliminary results from this calibration are presented below.