Grain Aliasing

[carllooper]

When scanning noise (such as grain) there is a phenomena called "grain aliasing" where grain appears larger than it actually is.

But it isn't. As you increase the definition of the scan, the apparent size of the grain shrinks! This is because what we're seeing in a scan is not the native grain but an "alias" of such.

There isn't any known way of alleviating this grain aliasing other than by increasing the resolution of your sensor. This has nothing to do with the resolution of the original signal. One could start with a completely out of focus signal and the situation would be the same.

[S8 Booster]

There is too long since I dealt with related details but I am frustrated by never being able to upload any of my Fuji F500T S8 footage as it actually is shown on my computer screen to any web site on the ned so I can actually prove what i wrte because it is actually good.

Something goes wrong. The original SD scan of this neg film show grain but very homogenous grain sure with no apparent aliasing.
Whenever "digitised" to any web place i have ever tried the stuff block quarks and quirks but can still appear quite good.

When viewed on the computer it kind of looks like this - the grain size is almost identical bu they "flow". theu reach the same amplitude but so ofnever simoultanesly og at the exact same spot. Probably much smoother at 25fps.

Coarsely remote imagery look of my MAC screen S8 reproduction just to try to show how:


Now, if one could get rid of all aliasing without interfere with the image itself - is that possible? Just a thought?


The other day I visited a TV shop with all sorts of TVs and stuff and for the first time I saw 4k stuff that was good - very good - 4k Sony imagery on a 4k Sony TV - it looked good enough to just walk in without opening a door or window.

However, this was a beach shot probably a place near you (Pacific and where the waves break at the horizon there was some aliasing - so it seems no not be there yet - at 4k. My impression from watcing all the 2k and all xHD stuff still battle with aliasing mor or less.

Well, anyway, just struck me when reading your post that if one could get rid of the aliasing all together would it be good for the film(grain)?

One film that looks good on the screen but on the net not. http://vimeo.com/35142252

[carllooper]

There is too long since I dealt with related details but I am frustrated by never being able to upload any of my Fuji F500T S8 footage as it actually is shown on my computer screen to any web site on the ned so I can actually prove what i wrte because it is actually good.

Something goes wrong. The original SD scan of this neg film show grain but very homogenous grain sure with no apparent aliasing.
Whenever "digitised" to any web place i have ever tried the stuff block quarks and quirks but can still appear quite good.

Yeah, nice fine grain looks brilliant. But most social media video codecs just can't handle it. They inadvertently re-introduce grain aliasing. This is because they match low frequency signals with less bits. But the grain which otherwise occupies the low frequency signal requires more bits - not less, and so the result creates grain aliasing, ie. ugly clumpy grain (which for some works could be exploited as a virtue).

Now you can't alleviate grain aliasing by using a low pass filter. It is low pass filtering which creates grain aliasing! Video codecs use low pass filtering if they consider a given area of the signal doesn't have much information in it (relative to other areas of the image which they estimate have more information). The low pass filter they use ends up turning the grain into aliased grain (clumpy grain).

So you have to start with a fine grain original (hires scan - the higher the better) and then, if it's destined for social media, you have to use a proper noise filter (or "degrainer") on it. The higher the scan res, the better a degrainer can degrain it! And you have to experiment relative to how the video codec will subsequently process it (including delivery size). In other words you tweak the degrainer and then apply the video codec to the degrained signal, and check the result for whether you've removed too much grain or too little during the previous degrain step. In other words you have to go back and forth between the degrainer and the video codec result, to find the right setting for the degrainer.

As strange, or counter-intuitive as this sounds, you need higher res scans the smaller the film gauge. That doesn't mean your deliverable has to be a giant scan. The giant scan is purely for mastering the work: for making a lower res deliverable.

The appeal of small gauges for me isn't this additional work that's required, but the fact that the cameras are smaller - and so much easier to carry around. If I have to do additional work back in the studio, at least I can sit on a chair while doing so, or sleep on the couch while it's oterwise rendering. So much more relaxing than running around with so much heavy equipment. Especially since I don't have a car and use public transport.

C

[S8 Booster]

That makes sense all tough some workload to make it swim.

Around 1990 i did do som film work on my MAC Quadras and then the details/artifacts you describe above was easily visual and verifyable.

The MAC had analog in and outputs - built in A/D and D/A cards.

The fun stuff was that there were almost unlimited number of software compressors etc with all different charateristics and speed.

The fastest capture was possible capturing uncompressed. Low load on the machine but it "ate" disk space like mad. one PAL image was 17-18 MB.

To get reasonable quality the best was was to capture "raw" and master with at slow processing but good quality fro reasonable playback/export.

Anyway, some fast compressors tried to "lock" (duplicate) the backgrounds if they were static while concentrate on processing the "moving" areas.
If one sett it "coarse" a great degree of blockiness (like bad streaming) appeared with the things in motion while the background was x-tal clear.

To overcome the capturing challenges I set up a RAM disk in memory to make raw captures to and then copy it to th HD. This had to be repeated if one needed high quality but it was of course possible to record directly to disc if the demand was like VHS solution - or mix the two when suitable.

It was so fun those days when it was so easy to set ANYTHING one wanted.

The early version of Adobe Premiere could handle 4096x4096 size images if I recall correctly.
Not possible with standard hardware.

It was real fun to verify all sorts of artefacts with the different settings and modules.

Shoot.....

[carllooper]

That makes sense all tough some workload to make it swim.

Around 1990 i did do som film work on my MAC Quadras and then the details/artifacts you describe above was easily visual and verifyable.

The MAC had analog in and outputs - built in A/D and D/A cards.

The fun stuff was that there were almost unlimited number of software compressors etc with all different charateristics and speed.

The fastest capture was possible capturing uncompressed. Low load on the machine but it "ate" disk space like mad. one PAL image was 17-18 MB.

To get reasonable quality the best was was to capture "raw" and master with at slow processing but good quality fro reasonable playback/export.

Anyway, some fast compressors tried to "lock" (duplicate) the backgrounds if they were static while concentrate on processing the "moving" areas.
If one sett it "coarse" a great degree of blockiness (like bad streaming) appeared with the things in motion while the background was x-tal clear.

To overcome the capturing challenges I set up a RAM disk in memory to make raw captures to and then copy it to th HD. This had to be repeated if one needed high quality but it was of course possible to record directly to disc if the demand was like VHS solution - or mix the two when suitable.

It was so fun those days when it was so easy to set ANYTHING one wanted.

The early version of Adobe Premiere could handle 4096x4096 size images if I recall correctly.
Not possible with standard hardware.

It was real fun to verify all sorts of artefacts with the different settings and modules.

Shoot.....

Yes, working with film and digital video in the late 80s and early 90s provided a lot of insights that are probably lost on a new generation. The video compressors were quite "naked" in terms of how they handled information. So one could literally see how they worked, even if one didn't quite comprehend such in any more sophisticated way. One could get a really good intuitive grasp of such. And being quite novel at the time there was something quite exciting about seeing even the most muddy video playing back on a computer. It had this psycho look to it. And it also seemed quite impossible as well. It was hard to imagine how one could get any image at all given the data rates at the time. It was like magic. And it still is. It remains an incredible thing: video compression.

Getting small gauge film to work within the same domain remains an ongoing challenge. But an interesting one for me.

C