Perhaps I look at some things too deeply (and at others not enough :)...but
the testing method can lead to false conclusions, when testing bitmap output
at different resolutions at the same physical output size with given content
on a given device. One must be careful of how one performs the test.

Ideally one would have a VECTOR file, sharp lineart and smooth gradations
and more photorealistic renderings done in vector, approximating photography
as much as possible with vectors (not easy, but not too hard - there are
some excellent vector illustration artists out there and if you have vector
software there are examples ore clipart etc).

I guess a STN file from GF or a Mr SID file or some other
fractal/vector/flavour of month method may do too (presuming enough good
pixels in the original to encode into the 'pixel-less' state).

Or even perhaps multiple scans with locked settings, apart from the
resolution differences (no USM).

My theory being that these 'sans-pixel' files are then rasterized at the
test resolution, all at the same physical size (no USM). This will remove
variables from the test, which can slip in when using more traditional
raster/pixel/bitmap originals and resizing them.

But if starting from an image at say 360 ppi at final size, then comparing
300 ppi or 260 or any other number to see what happens at output - will
involve resampling down to the other resolutions at the same print size.
This will soften the image, the more so with greater reductions of
resolution. USM can bring back the appearance of snap - but how to evaluate
and ensure that all the resized files at different resolutions all have the
same snap without throwing off the test results (which are often
subjective).

This is not the same as the original data and may lead to less than
scientific conclusions based on the amount of pixels discarded and how this
plays with the image content at hand. Sharpening can also be hard to
evaluate, the resampled images should have some USM, but how to be objective
that all the different pixel dimensions and sharpening are the same and that
this is not what makes the image look better, rather than the resolution.

Hope I am making sense. Perhaps I am looking too deep at things or worried
about nothing, but this is the sort of stuff I have done in the past to
ensure 'objective' analysis of resolution and print output for a given image
and output process.

Digital photos and stochastic screening are a lot more forgiving than
scanned film and halftone dots - one can get away with lower resolutions
when compared to traditional methods.

P.S. In years gone by, many average trade house scans to be drum scanned at
Res12 (120 ppcm - a little more than 300  ppi) - but when doing cosmetic
work, either product shots or female faces, Res 14 or 16 scans were often
common, to help smooth out fine sharp lines and to 'oversample' and
smoothout the skintones. This was a case of intentionally providing more
data than what would be required for an 'average' image in the same
conditions.

This PDF is quite good - this is a complex issue:

http://www.ledet.com/margulis/PP7_Ch15_Resolution.pdf


Regards,

Stephen Marsh.

===============================================================
GO TO http://www.prodig.org for ~ GUIDELINES ~ un/SUBSCRIBING ~ ITEMS for SALE

Reply via email to