Is this true??

[Pablo Moreno]

Please don't shoot the messenger. I read about it somewhere on the internet. Probably is been discussed here before. I just copied and pasted.

" The meaning of “K”: When Red refers to “4K”, they’re counting the photosites across their Bayer masked chip. That’s not the way the rest of the industry uses the term. “4K” as used elsewhere refers to pixels. Each pixel is a complete three color RGB data set for a single location in the grid. Red counts one color per location, not three. So, it’s apples and oranges. "

If that is true, then you need at least a 12K Bayer sensor in order to get a real 4K image??. Explanations are welcome..

 

[Martin Weiss]

There's a lot of marketing balooba there, but generally speaking a bayern pattern resolves above 80% of its resolution, thus a 5k Bayern pattern gives true 4k.

In photography this has never been an issue, curiously enough.

Anyway, the next generation sensor will have 6k, resolving in almost 5k resolution. More often than not, the lenses will limit the resolution, not the pattern.

Other sensors might employ 3CCD technology, which will resolve in light loss. There are always drawbacks.



Edit: What did you do with Joseph, and why are you wearing his hat?

 

[Terry VerHaar]

Yes and no. No CMOS, 4K bayer sensor will resolve to 4K, RED does well with it's Mysterium and supporting software science in that it will show measured resolution of between 80-85% of that number. I like to refer to it a "nominal" resolution (the number of photo sites or pixels in the image) vs. "measured" resolution, which is determined by shooting an appropriate chart. It is usually disingenuous for someone to say "RED isn't true 4K since others aren't either and many don't even come close."
Now that RED is shooting 5K (and soon, 6K) measured resolutions will exceed this 4K standard that Jim put up as the holy grail of film replacement a while back.

And I don't think it is accurate to say "that's not the way the rest of the industry uses the term." The truth is, no one even talked about 4K until RED came along.

 

[Vadim Bobkovsky]

All I know that 5K sensor delivers very sharp image with good glass. Debayering, high compression and OLPF eat some visual sharpness, but I still believe 4K image from 5K source resolve real 4K. I guess the Dragon is gonna be on the same level as Sony F65's "8K" 6,5K sensor with beautiful oversampled 4K output.

 

[Graeme Nattress]

There is a lot of mis-information and mis-direction on this very topic.

First off, we have to distinguish between a description of a sensor or an image's resolution, and what measured resolution a camera system can reproduce. Often similar terms are used for both.

We (as do many other camera manufacturers) say that a 4k sensor has 4096 photosites across it, and a 4k image has 4096 pixels across it. For sake of simplicity in our discussion, let's think on a large, single chip monochrome camera. Let's say it's 4k - has 4k photosites produces a 4k greyscale image. But what is it's measured resolution???? Because of the nuances of sampling theory, if you want your 4k mono sensor to produce a measured 4k resolution, you will also get some aliasing. The stronger the image contrast at a 4k level of detail, the stronger the aliasing will be. If it's too strong it will be detrimental to the image. Negligible amounts of aliasing are usually nigh on impossible to see. (All our cameras produce negligible levels of aliasing). In other words, if you try to put more than 4k worth of detail into a 4k image bucket, what spills over will manifest in the image as some level of artifacting that we call aliasing and we aim to keep that to a minimum. (see here for more info: http://www.red.com/learn/red-101/resolution-aliasing-motion-capture)

Introducing colour complexifies things..... The Bayer pattern colour filter array gives you a colour image from what would have been a monochrome sensor, but at the loss of some resolution. This was certainly an issue in the days of standard def where you have few pixels to begin with, but for digital cinema there are many advantages. We measure (on RED One M-X / Epic) around 80% of the horizontal resolution, so a 4k image would contain around 3.2k worth of image detail, with negligible aliasing. MTF is also very good with very strong contrast and detail at the 2k point and no corruption from aliasing. If you compare that to your typical 3 chip 1080p camera, they'll have visible aliasing at 1.9k (being the limit of the system) with a MTF50 much much lower. In other words, the system we use allows for a cleaner image (in terms of lack of aliasing artifacts) higher over-all resolution and higher MTF50 (which relates strongly to the human perception of sharpness / resolution).

If you go way back in time to the introduction of the original 4k RED One, there was no mis-rection on the camera specs, and I told people my conservative speculation on the measured resolution of the 4k sensor would be just over 3k.

Now, when you mention in your quote about "rest of industry" the example often given is the 2k or 4k film scan DPX. Each pixel does indeed have it's own unique R and G and B value, but that doesn't mean the measured resolution is 2k (or 4k). If you happen to open up your film scan you'll see that each colour channel actually has different levels of detail!

So why do we, and ever other digital cinema camera manufacturer (which means "rest of industry" is false because Sony, BMD, Arri, Canon, Nikon, Aaton, etc. etc. etc. ) use a bayer (or bayer-like) colour filter array? Because 3chip doesn't work for digital cinema. 3 chips need a prism, and that'd mean all your standard cinema glass doesn't work any more. Although bayer patterns may look awful at low resolutions, at high resolutions their benefits outweigh their detriments, and because all cameras need appropriate optical filtering to avoid the worst of aliasing artifacts, you're resigned to never getting the fullest resolution possible out of the system anyway, as measured resolution will always be lower than sensor resolution. Bayer patterns are currently the most effective image sensor compromise for all the variables.

Your concept of the 3 pixels - one each for R,G and B is basically what the Sony F35 used. They had effectively nearly 6k pixels horizontally, but only produced a 1080p image, and the actual measured resolution of the system was less than 1080p because of how the spatial offset of the R, G, B channels by 1/3 pixel caused rainbow artifacts and detail cancellation. Vertically they just had strong luma aliasing which was very visible on charts and fine clothing detail.

Of course, we understand the small resolution losses I mention above, which is why Epic is 5k as that is enough extra resolution to offset them, while still retaining excellent performance against the artifacts of aliasing.

Graeme

 

[Graeme Nattress]

All I know that 5K sensor delivers very sharp image with good glass. Debayering, high compression and OLPF eat some visual sharpness, but I still believe 4K image from 5K source resolve real 4K. I guess the Dragon is gonna be on the same level as Sony F65's "8K" 6,5K sensor with beautiful oversampled 4K output.

As F65 only has 4k worth of green pixels, it cannot be said to over-sample on resolution. At best it can be said to "just" sample. Math says it's sensor is equivalent to around 5.7k if you rotate the pixels back around to a normal Bayer. Of course, when they say 8k, they are really into the realm of marketing pixels, as it neither refers to their sensor, nor any where close to measured resolution.

Graeme

 

[Les Dittert]

I beg to differ.

Yes, a finely tuned 5K sensor with a fully optimized OLPF and a high performance debayer can deliver close to a 4K image.
But in the Red camera the OLPF is rather 'strong' for lack of better words, as is needed if you want to optically remove any trace of digital aliasing and moire patterns appearing on motion pictures.
Still cameras can get away with lighter blur from the OLPF becase the artifacts do not call attention to themselves if they are not moving.
Also the Red debayer is not on the sophisticated side on top of the blurred image that makes it through the OLPF. It does not have to be ! The light is so spread out to avoid aliasing optically that there is not much chance a fancy debayer can recover and reconstruct any detail anyhow.

There are image processing solutions that present themselves, because we are gifted the motion aspect that the still picture guys don't get to use.
Debayer without considering the temporal information is sub optimal.
See my posts that include full resolution 4K results of my R&D post processing. You will see true 4K . I challenge anyone to post full 4K jpg's that hold more detail. And I use a $600 lens !

Is there a reason that 4K full res. non-r3d image posts are rarely, if ever posted on the forum ? I find it very odd. People have 56K modems , maybe ?

 

[Graeme Nattress]

Les, indeed the low aliasing requirements for a motion camera are more stringent than a stills camera, not least it's feasible to hand-correct a still that has issues and not so the case on motion. And of course, as you mention, aliasing in motion looks doubly worse than aliasing that is still.

Graeme

 

[Brian Merlen]

dude with all do respect you know Red recently wanted to have a shoot out between all the cinema brands, and the other ones were too shook to even accept the invites. i was looking forward to seeing it, but sony/arri/etc were not. so you wanna throw up some charts and do this, i'm down i'll use dslr glass even, but don't throw around conspiracy theories lets just shoot it out and see where it stands!

 

[Blair S. Paulsen]

Its just not that simple...

Its just not that simple...

Bear in mind that just because a particular photo site has a color filter in front of it that is not green, does not mean it cannot read luma. I'll grant you that the debayering algorithm is interpolating some values, but if you can resolve it on a zone plate (resolution chart) then it exists - however it was created. That said, more native spatial resolution is a plus and the F65 image owes some of its IQ to it.

FWIW the marketing "pixies" raps are a labyrinth of half truths and curious definitions from every manufacturer to one degree or another. Its also worth noting that the OLPF (optical low pass filter) compensation portion of the decoding algorithm can have a more dramatic impact on delivered resolution than one might imagine.

I must also point out that the optics are a huge piece of the puzzle, particularly in terms of genuine mid band resolution that exceeds Nyqvist and is therefore perceivable by humans. One of my areas of curiosity has been to properly test the way particular lens designs interact with different OLPF implementations. I theorize that in a perfect world you would actually swap OLPF filters to match the "pitch" characteristics of the exit pupil of the light tube (lens).

IAC - shoot everything with the Leica C primes, hire a jock 1st AC, use a technically adept colorist and I highly doubt you will be disappointed with the resolution of the resulting 4K image from Epic, Scarlet or the F65.

Cheers - #19

 

[Les Dittert]

Hi Graeme,

Indeed, aliasing can be our foe, and possibly our friend as well.

Something interesting that I found: If I play one of my ultra detailed 4K test clips on a 30" monitor full screen ( 2560 pixels across ) it is chock full of aliasing issues. Horrible. But if I play it with a 2560 crop out of the 4K, so that the pixels are represented one-to-one ( no scaling or downsample by the player ) it has no aliasing.
So I feel that if I play it on a future 4K monitor or projector ( one-to-one ) I won't see the issue either.

 

[Jon Thomasberg]

What did you do with Joseph, and why are you wearing his hat?

Love it! Thanks for the laugh Martin.

 

[John Marchant]

Les I don't think there's any conspiracy on posting 4k+ jpegs, its just that they're not very practical for forum-based / web delivery.

I upload R1 / EPIC stills to my flickr account regularly, and often link to them - the full res is downloadable. My recently posted Monochrome tests also linked to full res .psd files

Just for you, an old Red One 4.5k WS still from a cheap wide angle - click

 

[Les Dittert]

There will be a high end shoot out at my company's stage in Los Angeles as soon as I get the Black Magic camera.
It will be science based, with posted results at full resolution. Not just charts ( they make little sense because modern debayer can sense obvious hard edges and snap them up ) , but also a set with various objects in it.

dude with all do respect you know Red recently wanted to have a shoot out between all the cinema brands, and the other ones were too shook to even accept the invites. i was looking forward to seeing it, but sony/arri/etc were not. so you wanna throw up some charts and do this, i'm down i'll use dslr glass even, but don't throw around conspiracy theories lets just shoot it out and see where it stands!

 

[Stephen Matthews]

Charts and repeatable non subjective tests are always necessary in my opinion. Even if they are not perfect they are a great point of comparison with each other, and a way of showing any errors in the methodology of the testing or how the cameras are being run. As well it is always nice when the raw files are provided for people to test themselves.

 

[Blair S. Paulsen]

Before making any blanket assumptions about aliasing on the display end you need to consider the downscaling choice - Lanczos is my favorite, Mitchell is popular, there are plenty of others. Bit depth at the panel and GPU design are also potential sources of aliasing. Just sayin'

Cheers - #19

 

[Les Dittert]

Thanks John, that still does look very nice.
Maybe we should start a thread just for 4K+ samples at full res. ?

Charts: Yes, I agree we must have charts for any shootout. As well as some 'everyday' footage as well.
Can't wait to do it !
My personal challenge is to out resolve the Monochrome Epic with my color Scarlet . How does that sound ?

Les I don't think there's any conspiracy on posting 4k+ jpegs, its just that they're not very practical for forum-based / web delivery.

I upload R1 / EPIC stills to my flickr account regularly, and often link to them - the full res is downloadable. My recently posted Monochrome tests also linked to full res .psd files

Just for you, an old Red One 4.5k WS still from a cheap wide angle - click

 

[Scott Crawley]

Thanks John, that still does look very nice.
Maybe we should start a thread just for 4K+ samples at full res. ?

Charts: Yes, I agree we must have charts for any shootout. As well as some 'everyday' footage as well.
Can't wait to do it !
My personal challenge is to out resolve the Monochrome Epic with my color Scarlet . How does that sound ?

It sounds as if you have something up your sleeve already. ;-)

 

[Vadim Bobkovsky]

As F65 only has 4k worth of green pixels, it cannot be said to over-sample on resolution. At best it can be said to "just" sample. Math says it's sensor is equivalent to around 5.7k if you rotate the pixels back around to a normal Bayer. Of course, when they say 8k, they are really into the realm of marketing pixels, as it neither refers to their sensor, nor any where close to measured resolution.

Graeme

Thanks a lot for the correction again, Graeme. That's actually interesting in the context of Sony claiming they're gonna release a firmware update to F65 that unlocks 8K mode... If it's really only 5.7K, then how on Earth they would do that without some kind of sensor upgrade program? Questions, questions...

 

[Graeme Nattress]

Thanks a lot for the correction again, Graeme. That's actually interesting in the context of Sony claiming they're gonna release a firmware update to F65 that unlocks 8K mode... If it's really only 5.7K, then how on Earth they would do that without some kind of sensor upgrade program? Questions, questions...

Well their unevidenced claim is that their current pixel arrangement gives them 8k. Of course, the straight numbers and experience with sensors say otherwise....

Graeme