THE REPRODUCTION OF BOTANICAL ART - Page 2

Hahnemühle papers from their Art range set against a neutral, artificial grey of the same brightness value. The creaminess should remain in the final printing of the magazine.

The surroundings to the image are pure, paper ‘white’. The petals of the lily, in the wild, are probably whiter than the paper and have to be 'suggested' by creating the folds, using subtle highlight gradations. The creamy, underlying texture of the paper is an essential element in the beauty of the original, but presents the greatest challenge. If the digital file is made accurately, it will have a yellow dot all over the ‘white’ area, to recreate the paper cream that the scanner can ‘see’. If this image is now printed on a cream, art paper then the background and highlights will receive a double cream layer – that of the paper, plus the yellow dots of ink. This yellow dot must, therefore, be eliminated in the digital file. Normally we might be tempted to go to Levels and drag the slider down from 255 to around 247 (the value which creates the 1–2% dot). However, this noticeably degrades the other highlights in a way that might pass unnoticed on a fullcolour acrylic, but stands out in a delicate watercolour. We are, by now, prowling around near the limits of what colour calibration equipment can achieve, as we will illustrate. The Epson V750 Pro is the most accurate scanner we have tested to date. This is based on making a scanner profile and then scanning the Macbeth Chart, applying the profile and analysing the error. One problem is that we see more variation between Macbeth Charts than the error limits we are attempting to work to, although we can overcome this to some extent with bespoke, hand-measured reference files. These are not needed for run-of-the-mill profiling but can help at the accuracy levels we are seeking.

Dealing with the Whites

In trying to understand the way that scanners and cameras respond to targets, we needed a constant, even and repeatable set of patches which were close to neutral. This was provided by a set of paint swatches, we tried both Dulux and Berger swatch sets, which provide very subtle differences in shades of off-white. This required a mass of statistics to describe what was going on, but in summary, the untagged as-scanned data was too dark and a little green. This was lighter and more neutral after assigning the V750 profile, but the highlights were a little compressed. This only confirmed the need for selective adjustments, the background was better dealt with by erasing it, the highlights were a little better in the as-scanned state but the greens were better with the assigned profile. A combination of all three treatments was therefore needed – messy, but better overall. These variations would be barely noticeable on a scan of a full-colour acrylic but such is the delicacy of watercolour that you have to work extra hard.

TOP : Hahnemühle papers from their Art range set against a neutral, artificial grey of the same brightness value. The creaminess should remain in the final printing of the magazine.

Middle and Bottom: The two scans show an assembly of art papers from the Permajet range. They are very subtle shades of cream apart form the Matte proofing paper which stands out as being very much cooler. To the right are the Berger paint swatches of the near-white, cream-biased paints. In the graph directly above, the blue plot shows the measured lightness of the paper using a spectro, plotted against the resulting RGB value in the scan. The blue line from an unprofiled scan has more contrast than the profiled scan. This is frustrating as the unprofiled scan was not as accurate overall, but was actually giving better separation. Such anomalies are part and parcel of working at really high colour precision; it can be a mystifying business.

The Greens of the Leaves

On the face of it, the green of the leaves should be relatively easy to reproduce. As can be seen from the table, the colour used by Margaret Stevens was very close to the Macbeth ‘foliage’ swatch. This is hardly surprising because the original Macbeth colour was specified to mimic the front of a typical leaf. As the table shows, there is a bigger difference between the same foliage patch of the Macbeth SG and Color Checker charts than between scans! This is despite the claims in the blurb that the ‘industry standard’ Macbeth Chart is contained with in the Macbeth SG Chart. On the other hand, the difference between the two SG charts we have was very small (0.66 ΔE Lab). The differences between the X-Rite profiling reference file and the actual charts was quite small at 2.4 and 3.0 ΔE Lab. However, these errors are of the same order as that we are trying to achieve (after profiling) over the whole gamut, so it is hardly surprising that we sometimes struggle to get really low error averages.

ABOVE: Although it makes for slightly tedious reading, if you are persuing the ultimate in precision, you are obliged to check every measurement. The table shows the size of various errors between targets and reference data and between scans and originals. The one assumtion this makes is that the spectro (left) is calibrated – ours is actually past its due date, although it self-calibrates before use each day. The subject area we chose is boxed in red (to the left) and was quite close to the Macbeth Foliage swatch. Our Digital Swatchbook is shown far left.

BELOW: The best overall profiles we could make with the Epson 10000XL and V750 did not produce the most accurate foliage greens – it's called Sod's Law we think! The errors were around 7 ΔE Lab, the poorest colour of the Macbeth swatches.

Photo Quote: The ear tends to be lazy, craves the familiar and is shocked by the unexpected; the eye, on the other hand, tends to be impatient, craves the novel and is bored by repetition. - W. H. Auden