[tig] Human Vision

Steve Shaw digital.praxis
Tue Dec 30 18:33:21 GMT 2003

Re-posted due to bad title in original. Sorry!
Ok, some very interesting feedback, and some real thoughts to work with.

I must thank Chuck Harrison for his very informative document, which is way
over my head, but very interesting.
Rather than use the idea of 1 arc minute per line pair, or similar, the
document Chuck supplied used Contrast Sensitivity Function, with the
resultant figures of 2500 lines at 1.5 times screen height and 1250 lines at
3.0 times screen height. That equates to 4625 pixels and 2312.5 pixels

However, and it's a big however, all the tests I have seen seem to use
'lines' as the test signal rather than actual pixels!
The human optical system is very attuned to straight lines and can 'see'
edges even if the resolution containing them is below the eye's ability to
define the detail.

So, I tried the following experiment:

I generated a test signal of alternating black and white lines with a width
of  0.6 inch. This equates to a projected pixel size from a 2K projector on
a 30ft screen. (the actual figure is 5.6888 pixels per inch). I then
generated a second 'crosshatch' test image with an alternating pixel pattern
across all lines, i.e. black-white-black-white on line one and
white-black-white-black on line two, and so on. (an example of these images
will be posted on the company website, with a short explanation, soon...)

I printed the images and illuminated them well (I don't yet have a 2K
projector at home...).

A 30ft screen showing a 1.85 image has a screen height of roughly 16ft.

When viewed from 24ft (1.5 times screen height and which subtends a
horizontal width of 64 degrees, well beyond the 'optimal' 30 to 45 degrees)
the image containing lines was still 'visible' as my eye was telling me
there was a pattern there, even if I couldn't actually see it in detail.

However, the crosshatch image had become completely grey with no discernible
detail or pixels.

Generating another text pattern with just three pixels, black-white-black,
produced the same result as the crosshatch. Nothing was visible other than a
single black point.

What was the most amazing was the distance at which the crosshatch pattern
became indistinguishable from total grey...

Also, the original math refers to a line-pair, which is 2 pixels/lines and
I'm told that there is no need to double this to two line-pairs (4 pixels).
My experiment above seems to agree with this as at a horizontal width of 30
degrees (about 56ft for a 30ft screen) even the 'lines' text pattern became
totally grey with no discernible lines.

I guess there is a middle ground as no image is ever a total crosshatch or
lines. Either way, I am becoming convinced that 2K digital projection is
well beyond what is required for a cinematic experience and well beyond what
is already perceived as cinema quality (remember the average present film
print is about 1.2K pixel equivalent).

As an aside, and as Peter brought it up, when it comes to capture I have a
certain agreement with Peter that Nyquist is important, hence my desire to
scan film at 4K and from that data Nyquist a 2K result for the DI process
(all detailed in my DI document). However, as Peter says, if the sensor is
in motion there is a form of Nyquist already taking place... have you ever
been able to keep a camera sub-pixel still? This is one reason why a digital
camera need not have as many pixels as a scanner, where the image to CCD
alignment is fixed. And as OCN film captures all the detail all the time
(fine grain) explains why the losses encountered during the chemical film
process are not an issue.

And as a final comment, for now, the following is a true-life story of

...timescales became very truncated and the decision was taken to perform
the DI work at the D-Cinema projector's native resolution of 1280x1024. This
dramatically reduced the amount of data that had to be handled, speeding up
the DI process.

One of the worries voiced by traditionalists was the choice of resolution
for the DI work. It was assumed that the D-Cinema final would look soft in
comparison to the film master.

However, the director & main actor, Roberto Benigni, was surprised to find
his 'frown lines' visible in the digital projected final when they were not
in the film print [an answer print]. As a result we had to 'soften' the
focus [blur it!] on the digital final [and I wont tell you how we did this
but is wasn't via digital technology!].

Good fun, this!


Steve Shaw
Digital Praxis Ltd
The Granary
Herts SG9 9RU
+44 (0)1763 281 699: f
+44 (0)7765 400 908: m
steve at digitalpraxis.net

More information about the Tig mailing list