To produce the colour maps, the broad band images in different filters were first aligned to within a small fraction of a pixel. For the images of NGC 7469, Mrk 3, Mrk 348, Mrk 573 and NGC 1667, this was achieved using field stars as reference points. For the images of the other galaxies, there were no stars within the field. In these cases, the alignment was achieved using the galactic nucleus as the reference point. In both cases, we used the central pixel with the highest intensity as the reference point. The frames were also corrected for differences in the magnification factor between the different filters. Although determining the centroid of the light distribution is preferable to using just the central pixel as the reference point, any resulting misalignment is likely to be only a small fraction of a pixel, and so should not influence the results for structures covering a large number of pixels.
The flux calibrated images were then smoothed to the same (lowest) resolution, by application of a Gaussian filter with a width corresponding to the seeing profile appropriate for the lowest resolution frame. Then the frame corresponding to the longer wavelength was divided by the shorter wavelength frame. The resulting maps are plotted in logarithmic intervals, with darker shades indicating blue and lighter shades red emission. At the edge of the colour frames the noise dominates, but closer to the nucleus the colours are reliable.
To check that the structures apparent in the colour maps are not artefacts caused by, for example, different seeing conditions in the frames or errors in the alignment, we have constructed colour maps of the quasar 3C 273, which was observed during the November 1990 run. Reassuringly, neither 3C 273 nor the stars within its field show rings or other artefacts. Instead, for 3C 273 we find that the colour becomes gradually and smoothly bluer towards the nucleus. Also, the field stars in the other galaxy frames do not show rings or other features, but cancel out well in the colour image.
A problem in using broad band filters for continuum studies is
the contamination from emission lines within the pass band. The
B filter is devoid of strong emission lines except for
contribution from H
4861. The V filter
includes [OIII] 
4959,5007 and some of the broad and
narrow H
4861. The R filter includes a number of
strong lines, H
6563, [OI] 
6300, 6364,
[NII] 
6548, 6854 and [SII] 
6717, 6731 lines.
The I filter is virtually free of emission lines, except for
[O I] 
8446 and [SIII] 
9069 lines. Therefore,
the V and R bands are most severely affected by emission
lines, but the B and I filters are dominated by the
continuum emission. This conclusion is supported by a literature
search of EW's of major optical emission lines in the nuclear
spectra of the sample galaxies. The average contribution of
emission lines to the broad band filters are 5% (B), 17%
(V and R), and 1% (I filter).