ad on
arget
nined
Inter-
; with
r fre-
s/mm
original
insfer
‘erent
for a
00. In
es on
the Durst enlarger with a Rodenstock optic and then
scanned with an effective pixel size of 10 um. This image
was then resampled to a pixel size of 40 um, correspond-
ing now to 10um with respect to the original. The effective
transfer function for 12 um is about 1096 lower. It is how-
ever astonishing that a very similar transfer function was
obtained with the PS1 with a pixel size of 15 um as with
the enlarged photographs. An other scan on the PS1 with
a pixel size of 7.5 um at a private firm gave a much more
unfavourable result. The result of the Rastermaster RM1
of Wehtli (121m pixel size) and the Agfa Horizon Plus (20
um pixel size) are very similar.
The results do not show the importance of the pixel size
as one would expect; but it must also be stated that these
results are very preliminary and have still to be verified. It
will also be interesting to compare these results from the
MTF determined earlier form the noise. Furthermore it
most be kept in mind that the MTF determined here is the
sum of the taking camera, the performance of the film and
the performance of the scanner. The MTF of the scanner
could be deduced when knowing or making asumptions
for the MTF of the other components. The force of this
test lies however in the comparative analysis. By using
an enlargement of the test film (curve F and G in fig. 2)
one also realisis that the quality of the scanner limits the
resolution of the scanned images.
4.2 Image Noise and the Dynamic Density
Range
The determination of the image noise is rather simple. It is
sufficient to scan the Kodak Photographic Step Tablet
(Test film 2) and to compute the RMS of the pixel values
within homogenious areas. The variance of the gray val-
ues is in general constant if one uses a logarithmic look-
up table over a certain range of density values. If one
analysis the variance of the initally obtained gray values,
which are in general proportional to the tansparency of
the film on observe a dimiuation with increasing density.
In order to obtain a comparable result which also gives in-
formation on the dynamic range it is strongly recom-
manded to transfer the variance of gray values into the
optical density. The image noise in density values is
given for the different scanners in table 1 and 2.
One remarks that the image noise is rather uniform up to
the density 1.0 and remains beneath +0.02 D. May be the
RM1 of Wehrli shows somewhat higher values whereas
the Agfa Horizon seems more favorable. A noise level of
+0.03 D is reached for 1.5 D and increases then rapidly.
That would mean that the tested scanners have a dy-
namic range of about 1.5 D nearly independently whether
a logarithmic or a linear look-up table is used for the trans-
formation of the gray values. Also the pixel size influ-
ences hardly the image noise, only for the Helava DSW
200 one has the impression that the image noise is more
favorable for a pixel size of 25 um then for 12 um.
It is understood that this test was run on material with
very fine grain. It must be suspected that effects like the
illumination used in the scanner will also influence the
graininess, an effect which should be proportional to the
pixel size. However this effect could not be demonstrated
here.
The limited dynamic range of the scanner explain the dif-
ficulties encountered when scanning negatives with a
contrast greater then 1 - 1.5 D.
4.3 Analysis of the Colour Reproduction
When working with colour photographs one is in general
not too strict as for colour fidelity like in printing industry.
Furthermore one is aware that haze degrades consider-
able the colour reproduction; even worth very often one
prefers false colour photographs instead of true colours
as they give more information. All this might explain why
the colour reproduction was not really analysed up to
now.
The Kodak Ektachrome colour table Q60 gives an effi-
cient tool for the testing of the colour reproduction. The
table includes batches which are only transparent in the 3
bands (red, green, blue) and transparency in the other
bands is below 196. When controlling the pixel values af-
ter scanning one has the impression that the Agfa Hori-
zon scanner gives a very adequate response whereas
RM1 DSW200 PS1 Horizon RM1 DSW200 PS1 Horizon
Density Wehrli Helava Intergraph | Agfa Density Wehrli Helava Intergraph
12 um 12 um 15 um 23 um 24 um 25 um 30 um 23 um
linear linear log. log. linear linear linear log. log. linear
0.04 0.014 0.014 0.010 0.012 0.006 0.04 0.013 0.008 0.008 0.013 0.006
0.21 0.011 0.015 0.017 0.020 0.005 0.21 0.011 0.007 0.015 0.011 0.005
0.37 0.018 0.021 0.016 0.020 0.008 0.37 0.012 0.009 0.012 0.012 0.008
0.52 0.024 0.018 0.020 0.021 0.007 0.52 0.012 0.015 0.012 0.012 0.007
0.68 0.016 0.014 0.020 0.024 0.012 0.68 0.014 0.019 0.017 0.014 0.012
0.84 0.027 0.031 0.022 0.030 0.010 0.84 0.018 0.014 0.020 0.018 0.010
0.99 0.036 0.029 0.026 0.042 0.010 0.99 0.019 0.018 0.021 0.019 0.010
1.15 0.037 0.022 0.056 0.031 0.011 1.15 0.029 0.021 0.018 0.029 0.011
1.30 0.051 0.032 0.022 0.034 0.020 1.30 0.029 0.022 0.018 0.029 0.020
1.45 0.031 0.036 0.027 0.041 0.025 1.45 0.050 0.024 0.018 0.048 0.025
1.60 0.078 0.049 0.032 0.031 0.027 1.60 0.055 0.032 0.028 0.055 0.027
1.75 undef. 0.044 0.049 0.093 0.050 1.75 0.070 0.055 0.019 0.072 0.050
1.91 undef. 0.068 0.086 0.088 undef. 1.91 undef. 0.071 0.028 undef. undef.
Table 1 Table 2
Determination of the image noise for different scanners in
function of the optical density. The image noise is also
given in density values; small pixel sizes specified in the
header.
59
Image noise for larger pixel size; other elements as in
table 1.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B6. Vienna 1996
