How will 2K material be refinished for UHD?

[Carl Looper]

I've scanned Super8 at 4K and played it back on a 4K (UHD) screen. And it certainly looks a whole lot better than the same film scanned at 2K.

 

According to certain theories it shouldn't look any better, but it does. So it's how one describes this better looking result, in a theoretical way, that is more the point, rather than whether (theoretically) it does or does not look (or can or can not look) any better.

 

While resolution is often discussed, it's not just resolution that is a higher res scan can improve. A higher res scan can also improve the dynamic range of the result - since there are more bits per area of film being captured. If a 2K capture obtains, say 24 bits per area of film, a 4K capture would capture 96 bits per same area of film. Four times as much information. So even if there were not any appreciable increase in detail (resolution), there will still be an appreciable improvement in the tone and colour.

 

In film there is no scale at which the image just stops and pure grain (or noise) begins. Both image and noise are present at all scales. While the frequency response of the image becomes increasingly difficult to distinguish at increasingly smaller scales, the tonal response of the film, while noisy, is still precisely entangled with the magnitude of the light from which it was derived. It is statistically precise. Or to put it another way there's no barrier beyond which the film just decides to arrange itself according to some concept of neutral noise. It is under the control of light all the way down, regardless of how finely tuned or not that light might be in terms of resolving power.

 

Carl

[Roberto Pirodda]

 

At 60 lp/mm, S35 should resolve about 2880 lines and S16, about 1502. Even Velvia would only get (coincidentally) 3840 lines from S35, right?

 

Come on, it is already has been discussed hundred of times here ! As the word "motion picture" says, we have an averaging of multiple frames, and the higher the framerate the better (have you ever heard about showscan?) So you cant make equal a single still 35mm film frame with digital world. I am not able to calculate the equivalent 35mm film in motion at 24 fps, but for sure it is far beyond the 2880 lines

[Carl Looper]

Re.compression.

 

In "lossless compression" there will be (must be) certain types of images which, when 'compressed' would actually produce a larger file than a corresponding uncompressed file. But in the case of pictorial images (as in conventional photography/cinematography) the majority of images will tend to produce smaller files (compared to an uncompressed file).

 

The pictorial world produces a distribution of frequencies, from the very soft to the very sharp. But in an uncompressed file (so called) the data is arranged according to the assumption of a worse case scenario: that everywhere were just high frequencies, but if this were the case the image would have to be pure noise. Since most images are not pure noise, there is room for a more optimised encoding of pictorial data: so called "compression".

 

Instead of reading off each pixel as a standalone value, the image (called a signal) is transformed into the frequency domain (or hybrid versions of such) where the 'pixels' of such are now frequency values rather than amplitude values. At this point there is no change in the data size, but there is now an opportunity to vary the precision of the representations according to the frequencies they represent. This does not remove necessary information from the data - it removes any redundant information from the data. Another way to think of this is that an "uncompressed" file (unless it is pure noise) will have redundant information in it. It will be bigger than it actually needs to be. It will be bloated. The compressed file can be regarded as a deflated version of the bloated file.

 

Display technology, however, can't directly display such a deflated file. The file needs to be re-inflated with redundant information again, in order to obtain individual values that can update the individual pixels of a display.

 

C

[Carl Looper]

Re. Super8

 

While a 4K scan of Super8 (over lower-res scans) appreciably improves the tone and/or colour of the result, it is arguable that it does also, in fact, improve the resolution of the result. The resolution of film, and in particular motion picture film (where time as much as space plays a role) is quantified according to observer based concepts. One shoots an MTF chart, for example, and one asks a range of observers (human or otherwise) to decide the distinguishability of certain frequencies in the chart. From a human perspective (as much as a machined version of the same) one tends to be looking at a small area of the image and fighting the noise to discern anything of the target. At some point one decides one can't see the target - only noise - and one draws a line there.

 

One simply decides that if it can't be seen it's not there.

 

However perception is a complex thing. While looking at an isolated area of an abstract target one can be forgiven for not seeing anything more. There's nothing much there to hold onto in the target signal, let alone deciding if it's still there in a haze of noise.

 

But if one goes back to a pictorial image, where the image isn't abstract but a recognisable reality, there will be subconscious correlations between the macro-scale signals and very fine details (even if wallowing in noise), and we can often see, quite clearly, whether such details (even if noisy) are consistent or inconsistent with the apprehended signal. There will be very fine biases in the noisy micro-details that reinforce the macro-signal. And it is clear from understandings of the way film is manufactured and otherwise works that this must be the case. There is no scale at which the film switches off. It is sensitive to light all the way down.

 

It is particularly the case in motion pictures where not only the observer is engaged in a dynamic relationship with the image (eyes scanning the screen) but the film is also engaged in a dynamic relationship with the the reality it otherwise observes. Each and every interaction, both macro and micro, contributes to the image being created.

 

To switch any of these interactions off is purely for practical utiltarian purposes rather than any properly theoretical reason. Methinks

 

C

[Chris Millar]

Whatever eventuates, it will be the net result of mostly negative forces - ignorance born of mostly laziness on the behalf of those who don't see anything to gain combined with the purposeful misuse of engineering terminology for marketing purposes from those who do.

 

I applaud those who educate themselves and then disseminate that information (as abstractly as any forum post can afford) ... and there certainly is something to be said for the feeling one gets from being around (at least online) others who 'understand'.

 

I but wonder, perhaps ignorance is bliss :rolleyes:

Edited January 21, 2015 by Chris Millar