[tig] Fw: 4K for DI - Unappealing image quality

Steve Shaw digital.praxis
Mon Apr 12 19:43:09 BST 2004


Here's something to mull over.

I have recently played with some 4K vs. 2K data images scanned from various
film stock sources. The results, from an aesthetic view point, were very
interesting.

The 2K images were more 'pleasant' to view than the 4K scans!


The reason for this, backed up by a vector scope and waveform monitor, seems
to be that the 4K images were too 'sharp' giving an edgy image with hugely
increased overall noise level. This looks to be due to the 4K scan resolving
small grain particles, increasing noise without any noticeable increase in
image content detail.


In effect, high frequency information is being maintained that has little
impact on the captured image detail, making the resultant digital image
'buzzy' and noisy when viewed moving. Something not seen in still frame
comparisons.

When OCN is optically processed a lot of the high frequency information is
removed by the lossy dupe processes caused by film grains not aligning
between the original and the copy, and especially within vertical detail,
due to 'slip' between the original and dupe film strips as they move through
a slightly different radius while being copied.

This high frequency reduction seems to be a key component in making pleasing
projected images from high detail OCN. This again suggests 4K scanning to
produce a 2K Nyquist image for the DI process and projection is the optimum
way to go, with the Nyquist resolution reduction removing unwanted high
frequency noise, while keeping full 2K image detail. Having said that I also
looked at some direct 2K scans (Spirit) and the lower colour information it
scans resulted in a noticeably lower noise image with minimal image detail
loss - although there was some.

The talk of film capturing ultra-high resolution is missing the real point.
Such resolution never makes it to the screen due to the low-pass filtering
of the lossy optical duplication process. A process that softens the harsh
and noisy OCN image, caused by film's underlying grain structure adding high
frequency noise that wasn't in the original scene.

Interestingly, digital cameras don't suffer such high frequency noise, so 4K
and 6K capture is possible without the associated problems of film. But are
such high resolutions really necessary when we are looking to improve on
print film's nominal 1.2K on-screen resolution as seen in most theatres
throughout the world?

In a separate test we shot the same image in 35mm film and HD (CineAlta in
HDcam mode so lots of compression, pre filtering and the like) and then
zoomed in 400%. While the film image, scanned at 4K, held higher frequency
image detail, but not as much as you might think, when zoomed in the
underlying grain noise made the image unacceptable far quicker than the
cleaner HD image. There are Viper/35mm examples of this test on the Digital
Praxis website.

I guess what I'm saying is there is always more then one side to a
discussion and the first answer isn't always the right one, even if the
facts are correct.

Steve

Steve Shaw
Digital Praxis Ltd.
+44 (0)7765 400 908:M
+44 (0)1763 281 699:W
www.digitalpraxis.net






More information about the Tig mailing list