Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Resolution

Again, I'm totally on board for oversampling. No issue whatsoever with that. But what's being discussed here is actually recording 6K or 8K resolution.

I'm actually discussing the benefits of resolution in general.

What I'm trying to get at is that there are many benefits over and above the headline resolution figure.

Graeme
 
Resolution certainly has a positive benefit on tones, hues and colours. Micro-detail on hues works against homogeneity in a natural manner and the proper texture adds verisimilitude. Resolution allows for better capture of the lens MTF and thus will result in more pleasing and less artificial tones. Same with the lower propensity for aliasing.

Graeme

Well one is actually capturing more colour data as for one 4K pixel there are now 4 pixels in RGGB. :)

So this technically gives full colour resolution for 4K delivery. Which affects the appearance of "thickness" and natural impression of tonal transitions.

Yes, it's a tad hard to describe but we're all having our go at it! And it's possible to demonstrate some of the technical aspects, but you really have to see it all together on your own footage to really appreciate it. And yeah, the 8k leads to super-fine micro-detail which can be rather compelling when processed through a sensitive image processing pipeline designed to bring out the best in the footage.

Graeme

The main benefits I find with 8K are clean scale to 4K delivery with minimized effect of scaling algorithm and achieving full colour resolution.
 
What I'm trying to get at is that there are many benefits over and above the headline resolution figure.


And I think what I'm trying to get at is there are some tradeoffs as well, as long as RED is making sensors that record Super35 at 8K rather than 4K or even 6K.

I appreciate that we can have a reasoned discussion about both advantages and tradeoffs. Sorry if I took that discussion off the rails at any point :smiley:
 
I wonder if the future holds different sensor purchasing options, at different resolutions... :idea:
 
I'm not sure if they graded every single shot for Gone Girl. Shooting 6K was, IIRC, for the purpose of extracting a 4K window later. A form of insurance. So shooting RAW isn't just for exporting later, it's for the ability to change your mind if you need to.

If the average person was in charge of the camera data, I doubt they'd keep all of it. I sure wouldn't. It's not worth the trouble. Maybe you'd keep it for a year, just in case you spot a composition flaw that you hadn't noticed before. But once it gets out on Blu-ray, that's it. Lock it in, and throw away the out-takes. Trying to have everything is not practical.
 
Well one is actually capturing more colour data as for one 4K pixel there are now 4 pixels in RGGB. :)

So this technically gives full colour resolution for 4K delivery. Which affects the appearance of "thickness" and natural impression of tonal transitions.
I consider 8K CMOS based camera to be the best way to deliver 4K content and this is a great example of why!

The main benefits I find with 8K are clean scale to 4K delivery with minimized effect of scaling algorithm and achieving full colour resolution.
The 2:1 down scale for 8K HD to 4K HD is nice and clean. 6K HD and 5K HD are not nearly as clean in that regard. I hope people don't misconstrue my use of the word clean. :)

Now 8K HD to 2K HD is not an ideal ratio. For that, I use a different scaling kernel. e.g. I use a box filter (with additional processing to maintain perceptual quality) for 8K HD to 4K HD while I use Lanczos for 8K HD to 2K HD. The non 2:1 ratio on the 8K HD to 2K HD may result in moire from the box filter that Lanczos does not suffer from. Box does not ring while Lanczos does. I have to use an additional anti-ringing filter when using Lanczos.

I shot 8K HD through 4K HD in 1K HD increments. I can also scale them all to 2K HD. I would like to wait until the new room / chart is done, which should have the dolly track and gear head to ensure much closer framing at each resolution. Also a room long enough to go down to 2K HD. Its a test chart, so not really ideal for this scenario (showing 8K HD scaled to 4K HD vs. 4K HD original)

The 8K HD looks really nice on the Dell 8K HD display. I am going to use a Dell UHD display next to it to tune the perceptual down scaling so they can as close as possible.
 
In computer graphics although you might render out HD, internally the math is such that it comes from a heavily oversampled internal image before being output.

This made the discussion so clear for me. To render anything in CG with any currently available raytrace renderer, everything in the scene must be pretty massively oversampled to achieve a noise free or minimal noise image that is also sharp.

The sensitivity discussion also makes sense to me. Dealing with tiny cell phone image sensors is an entirely different ballgame with all sorts of other compromises. In that same regard, I have a question, how much does size of sensor relate directly to quality of image captured? Or maybe pixel density is what is important? Would doing a 16k sensor at S35 size have any meaningful improvement over 8k for example? Just curious if you guys are anywhere near diminishing returns in terms of pixel density.
 
This made the discussion so clear for me. To render anything in CG with any currently available raytrace renderer, everything in the scene must be pretty massively oversampled to achieve a noise free or minimal noise image that is also sharp.

The sensitivity discussion also makes sense to me. Dealing with tiny cell phone image sensors is an entirely different ballgame with all sorts of other compromises. In that same regard, I have a question, how much does size of sensor relate directly to quality of image captured? Or maybe pixel density is what is important? Would doing a 16k sensor at S35 size have any meaningful improvement over 8k for example? Just curious if you guys are anywhere near diminishing returns in terms of pixel density.

I'm not happy predicting in theory what is going to work best on the sensor side of things. There's too many variables and pixel design advances along with the electronics and processing. Unfortunately that means we'll have to wait on good empirical data before I can answer that! But my guess is that the size of the sensor is intimately tied to the lenses used with it, and that is a major influence on the image we see. Pixel density is going to interact with the lens MTF too, and that is partly why Helium sensors look so good. It's a very interesting question.

Graeme
 
19222732_10211027676661021_6483367987082731668_o.jpg

Just adding a little visual reference, I think the extra rez helps tonal range a ton!
EpicW 7K still, 35mm Xeen @T2
 
If you wanted to shoot a Netflix drama using an FS-7 as the "A" camera, my guess is that Netflix would not be happy about that and at the very least would encourage you to use a "better" device. Or they might simply say "no."...
I just glanced through the latest Netflix specs I have and I didn't see anything there that says you can't use a shitty 4K camera -- only that you can't uprez 2K or HD. What if they used a Blackmagic Ursa Mini? I think there's a slippery slope when it comes to actually specifying what camera is acceptable and what isn't. There's always somebody who might use a 10-year-old Red One and then try to jack it up to ISO 6000 and say that's the look they want. Or they use a modern 4K camera and use it badly.

I wouldn't have a problem if they had some kind of generic statement that said, "in addition to source resolution being 4K or higher, care must be taken to ensure reasonable picture quality and noise levels according to industry standards." It's vague, but at least it would preclude somebody trying to shoot 4K on an iPhone.

I do agree with you about the FS-7. For $10K, it's still noisy and doesn't quite hack it for dynamic range. On the other hand, it's better than a GoPro, and I know of Amazon and Netflix projects that have occasionally used GoPro cameras for brief cutaways.
 
19222732_10211027676661021_6483367987082731668_o.jpg

Just adding a little visual reference, I think the extra rez helps tonal range a ton!
EpicW 7K still, 35mm Xeen @T2
That's a great looking shot. Another thing I notice with red is having the nice dynamic range (i.e. basically 16 bits) with the extra resulution really gives the nice feel like I see in this still.
 
I just glanced through the latest Netflix specs I have and I didn't see anything there that says you can't use a shitty 4K camera -- only that you can't uprez 2K or HD. What if they used a Blackmagic Ursa Mini? I think there's a slippery slope when it comes to actually specifying what camera is acceptable and what isn't. There's always somebody who might use a 10-year-old Red One and then try to jack it up to ISO 6000 and say that's the look they want. Or they use a modern 4K camera and use it badly.

I wouldn't have a problem if they had some kind of generic statement that said, "in addition to source resolution being 4K or higher, care must be taken to ensure reasonable picture quality and noise levels according to industry standards." It's vague, but at least it would preclude somebody trying to shoot 4K on an iPhone.

A spec IS a general statement. It's a starting point of what is expected. Anybody doing a Netflix original production has meetings and discussions with Netflix in which things are proposed. Some of those things are accepted, some are rejected. As a company, you don't (and can't) put every single possible contingency into a spec document. That's why meetings and discussions take place at which specifics are put on the table. At that point decisions are made based on the needs of the particular production, the expected production circumstances, the artistic considerations, and Netflix's own goals. Those decisions are very sober and sensible. They are almost never ridiculous.
 
So does this mean we'll be seeing 12k full frane helium soon?? :ihih:
 
When Jim said "Resolution matters" and made that a cornerstone of RED's mantra, he was right. It mattered then and it matters now. I remember the "HD is not Good enough" threads and he was right about that as well.

AS one example, I heard that IMAX is accepting 8K Helium SUPER 35, but not 6K Dragon (which has a slightly larger frame size over the Helium)

Pixels. We are in a pixel game. Pixels matter. I am not going to get into a more pixels are better discussion, but it roughly equates to 16mm, 35mm, 65mm, 70mm etc.

I don't know, but I feel better about having shot at least 4K for the last 10 years because 2K does not really cut it.

David

Resolution matters, but color reproduction, DR and overall image quality are much more important. Which is why Arri is king with their Alev sensor. Putting resolution first was a mistake imo.
 
Locally adaptive downsampling - interesting!

I consider 8K CMOS based camera to be the best way to deliver 4K content and this is a great example of why!

The 2:1 down scale for 8K HD to 4K HD is nice and clean. 6K HD and 5K HD are not nearly as clean in that regard. I hope people don't misconstrue my use of the word clean. :)

Now 8K HD to 2K HD is not an ideal ratio. For that, I use a different scaling kernel. e.g. I use a box filter (with additional processing to maintain perceptual quality) for 8K HD to 4K HD while I use Lanczos for 8K HD to 2K HD. The non 2:1 ratio on the 8K HD to 2K HD may result in moire from the box filter that Lanczos does not suffer from. Box does not ring while Lanczos does. I have to use an additional anti-ringing filter when using Lanczos.

I shot 8K HD through 4K HD in 1K HD increments. I can also scale them all to 2K HD. I would like to wait until the new room / chart is done, which should have the dolly track and gear head to ensure much closer framing at each resolution. Also a room long enough to go down to 2K HD. Its a test chart, so not really ideal for this scenario (showing 8K HD scaled to 4K HD vs. 4K HD original)

The 8K HD looks really nice on the Dell 8K HD display. I am going to use a Dell UHD display next to it to tune the perceptual down scaling so they can as close as possible.
 
Resolution matters, but color reproduction, DR and overall image quality are much more important. Which is why Arri is king with their Alev sensor. Putting resolution first was a mistake imo.

There are two interesting counter-examples:

Arri 65 is not 3k but 6k. If you were correct, then Arri would have used the larger sensor and kept the resolution the same, making a bigger pixel. They didn't.

Canon, Sony, Panasonic, Phantom, Ikegami, Hitachi.

Graeme
 
Resolution matters, but color reproduction, DR and overall image quality are much more important. Which is why Arri is king with their Alev sensor. Putting resolution first was a mistake imo.

IF the Alev 2 actually had significantly better DR than Dragon or Helium then you'd have a point. Color reproduction includes so many factors beyond the performance of the sensor itself, starting with debayering algorithms/color matrices, that claiming one is "better" than another ignores the complexity and inherent subjectivity of such evaluations.

The Alexa does many things well and deserves it's reputation as a excellent digital imaging device. Alexa's success can be attributed to a number of factors, too numerous for this thread, but it's only fair to note that many of those factors are not strictly about image quality.

As many have noted just in this thread, the impact of higher resolution and downsampling has a powerful effect on overall image quality. Resolution isn't a magic bullet, but starting with high resolution as the basis for a great motion imaging system is hardly a "mistake".

Cheers - #19
 
The most cynical part of me attributes the Alexa's success to several factors - yes, this is on-topic:

1. It produces a nice image, so right there, it cannot be a failure

2. The ARRI brand (they make amazing film cameras - but I have no interest in their digital products personally)

3. Virtue signalling by the operator - the ability to say that resolution isn't everything and that it's the story that matters

4. Related to point 3 is the feeling of maturity - the Alexa is more expensive yet generally less capable, giving the operator a feeling of sober decision making

5. It's not made by RED

6. Future shock - the inability to grasp more than one revolution in an industry (look at how many photographers really don't like mirrorless cameras because they will make their favourite DSLRs look old-fashioned - even though these same photographers happily rode the digital revolution)

Edit: very much to ARRI's credit, they understood downsampling from the beginning, hence the Alexa is a genuine 2K camera (as much as a Bayer sensor can be), unlike many craptastic DSLRs which are 1080p on paper only.
 
Back
Top