Welcome to our community

Be a part of something great, join today!

Log in
  • 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 :)

IBC '09? There's a new Canon camera to be announced soon (Canon source)

Steven Caesare said:
Interesting that those cams downscale and don't even give you the native frame if you want it?
Click to expand...

In a RAW camera, there is no "native frame," at least none that gives you a coherent image. If you make a 4K image from a Red, you're making an interpolated 4K image because a demosaic reconstruction is imperfect. Even Graeme will tell you that. So if anything, the Red "upscales" (in your interpretation) to give you a 4K RGB image, while the other devices do not. The notion of oversampling for better results only exists at lower (than 4K) resolutions for the Red. The decision as to what sensor(s) to use and the size image to output at the camera's final stage are engineering decisions that are intended to get the best results. Different manufacturers have different ways of looking at that.
 
M Most said:
Canon has redefined the meaning of the term for its own purposes. Just like Red did.

The moral of the story is that it's silly to blame companies for misleading marketing, because it's done all the time at all levels by all companies, intentionally or unintentionally, depending on your point of view.
Click to expand...

While I do agree that this does happen. I just find it hard to accept that wrapping that situation up in the blanket of "redefinition of the meaning of a term" justifies it.

Say they would have crippled it to 15, maybe 10 FPS. Even MORE unusable, would that also be simply redefining the meaning of a term?

Its one thing when a camera like RED produces that caliber of image that it does, and shoots their interpretation of 4K. Resolutions references can be ambiguous terms can they not? (aren't both 1920 x 1080 and 1440 x 1080 both considered 1080?)

But my issue in that instance, was that, it was not a matter of interpretation but usability. The mass majority of people do not accept 1080p20 as a acceptable resolution. And I highly doubt Canon does either. But when they are competing with Nikon (who often times has a slight step up in the SLR world), it helps to slap a selling point, and devote less attention to functionality.

To me simply saying that's an "interpretation", is dangerous as it prevents acknowledging faulty aspects about a product, and writes them off to interpretation.

Imagine an automobile that's marketed as a "turbo car", you buy it, hit the gas, and are incredibly disappointed. You get home and pop the hood to see if something is possibly wrong, only to discover that the turbo is there, mounted under the hood, but the Manufacturer decided not to integrate it with the engine. Its just sitting there. You have a car, and inside that car, there's a turbo. So you have a "turbo car". Perhaps that's just that Manufacturers redefinition of the meaning of the term for its own purposes?

No, that's just a marketing ploy, and not dissimilar to what Canon did with the T1i. Sure it has "1080 capture", but its unusable in almost every situation....

So while RED may have a slightly different idea of what 4K is to you, or to me, or someone else (and resolution terms already can be ambiguous). They still aren't saying 4K, but giving us crippled features in an attempt to "steal looks" from a competitors product, etc.

That's the "RED difference", I was talking about in my first post..
 
M Most said:
Yes. Sony and Panavision with the F35 and Genesis (both have 5.7K sensors, but are advertised as HD resolution cameras, which is what they output), and Arri (whose D21 has a 3K Bayer sensor, but is advertised as a 2K camera). Not to mention Thomson/Grass Valley with the Viper, which has three 9 megapixel chips (the pixel equivalent of 12 times HD resolution) but is advertised as an HD camera.
Click to expand...

Because in the F35/Genesis/Viper cases, they were designed only ever to deliver HD resolution - no more, but use different techniques to achieve that.

You can also think of the HVX200 which uses three sub-HD chips with a pixel-shift to deliver a HD output. Or Juan's ReelStream cameras that used the offsets on those cameras with more advanced algorithms.

What's nice about this industry is that there are many, many ways to take sensor(s) and make a nice image - and you can bias towards various different image parameters, and also balance them with how the human visual system perceives images. (See previous threads on this aspect). YOu can pick the one that suits your needs.

Graeme
 
M Most said:
In a RAW camera, there is no "native frame," at least none that gives you a coherent image. If you make a 4K image from a Red, you're making an interpolated 4K image because a demosaic reconstruction is imperfect. Even Graeme will tell you that. So if anything, the Red "upscales" (in your interpretation) to give you a 4K RGB image, while the other devices do not. The notion of oversampling for better results only exists at lower (than 4K) resolutions for the Red. The decision as to what sensor(s) to use and the size image to output at the camera's final stage are engineering decisions that are intended to get the best results. Different manufacturers have different ways of looking at that.
Click to expand...

"native" in the sense of the actual resolution, which is what is being discussed here.

A DSLR has what's conidered a "native resolution" which is it's sensor resloution, even tho it may be Bayer.

And while I understand what you were trying to say, de-bayer and upscaling are typically thought of as seperate process. Debayering a given image does not imply scaling the resolution up to a higher pixel count, whereas upscaling typically does.

-sc
 
Last edited:
Michael Vera said:
So while RED may have a slightly different idea of what 4K is to you, or to me, or someone else (and resolution terms already can be ambiguous). They still aren't saying 4K, but giving us crippled features in an attempt to "steal looks" from a competitors product, etc.

That's the "RED difference", I was talking about in my first post..
Click to expand...

I'm going to quote myself from my previous post:

People here seem to think that I'm arguing something. I'm not arguing anything other than to leave personal emotions out of logical conclusions about business.
Click to expand...

Now I rest.
 
The difference between up-sampling and demosaicing is that in upsampling you have no data at the pixel location you're interested in. With demosaicing you don't have no data at that location - you have less than you need, but more than none. Mathematically they can be viewed as similar problems but they are not the same problem or the same thing.

Graeme
 
Steven Caesare said:
while I understand what you were trying to say, de-bayer and upscaling are typically thought of as seperate process. Debayering a given does not imply scaling the resolution up to a higher pixel count, whereas upscaling typically does.
Click to expand...

Semantics.

In either case, you're creating RGB pixels that did not exist in the original captured information. Whether you're using that to fill out the image to the size you want to represent, or taking a smaller image and presenting it in a larger size, you're still creating information based on "best guess" methodologies rooted in various mathematical theorems. So you are, in effect, doing similar things, regardless of how you want to guise it.
 
Oops, must have been typing at the same time Graeme was. Fortunately, we both said pretty much the same thing.
 
M Most said:
Oops, must have been typing at the same time Graeme was. Fortunately, we both said pretty much the same thing.
Click to expand...

Really?

Graeme said:
Mathematically they can be viewed as similar problems but they are not the same problem or the same thing.
Click to expand...

Wikipedia: A video scaler is a device for converting video signals from one size or resolution to another: usually "upscaling" or "upconverting" a video signal from a low resolution (e.g. standard definition) to one of higher resolution (e.g. high definition television).

I disagree.

Demosiac uses luminence values from each individual pixel location, and mathematically interpolates chorma vaules for the pixels where information is missing. However it is doing so at the source resolution.

Ask 100 people what upscaling means, I and I almost guarantee you hear something incorporating the idea of a resolution change 99% of the time.

Ask the same 100 (if you can find them) what a debayer involves, and I suspect you'd not hear the word "upscaled" even once.

Different things.

-sc
 
You're quote mining me :-) What I said was what I said. The whole thing I said was in context, the above quote is not.

Graeme
 
Graeme Nattress said:
You're quote mining me :-) What I said was what I said. The whole thing I said was in context, the above quote is not.

Graeme
Click to expand...

I know. But the larger point I'm trying to make (and everybody seems to miss it) is that you can't have the "native" resolution of 4K in a 4K image created from a 4K Bayer pattern sensor because the image represented by the RAW data must be interpolated (call it demosaic, call it whatever you want - you're still creating information that isn't explicitly there in the original) to create an RGB image with the pixel dimensions of the sensor. So if someone is going to state that a camera such as the Genesis, for example, is "downsampling" - which you and I know it is not - then you could, in a sense, say that Red, for example, is "upscaling" - which is, technically, just as untrue. I also pointed out that the idea of "oversampling," often used as one of the prime motivating factors for using a higher resolution original, only comes into play with Red One images at resolutions lower than 4K, which is true.

Is that better?
 
You're getting there, but most of the reason you can't have 4k from a 4k sensor is the very same reason it would be for any camera - if you have significant MTF at rated resolution, you must also be having aliasing. The higher the MTF at rated resolution, the higher the aliasing. The 5D2 has very high MTF (vertically at 1080p) resolution, and is also almost totally corrupted with aliases vertically.

For the most part, the information interpolated in a demosaic is in the original image, just somewhat burried. It takes a fair bit of work to extract that out. In demosaicing you have a sample at each and every pixel position. With 2x upsampling, say, you have a known value at 1 out of pixel locations, the other three you have no value at all. As I said, related problem, but a different problem.

Both the Genesis and RED One sample the image - both using a colour filter array. However RED uses quite complex image reconstruction algorithms and AFAIK, the Genesis uses none. The optical low pass filter in the RED is set appropriately for it's 4k output resolution. I have no idea on the Genesis - I need to point it at a chart to see what it's doing. I'd assume the OLPF was set more for 1080p output, but I'd not stake my reputation on that without measuring actually.

The Arri D21, however does generally downsample - from the 3k demosaic to HD output. In the RED has a nice 2k from 4k high quality downsample mode in post that is similar in concept in that you reconstruct then downsample for best quality. The image quality doing that is visibly superior to a simple extraction (which is similar to how basic reconstruction of an RGB stripe works).

Graeme
 
Graeme Nattress said:
You're getting there, but most of the reason you can't have 4k from a 4k sensor is the very same reason it would be for any camera - if you have significant MTF at rated resolution, you must also be having aliasing. The higher the MTF at rated resolution, the higher the aliasing. The 5D2 has very high MTF (vertically at 1080p) resolution, and is also almost totally corrupted with aliases vertically.
Click to expand...

Thanks, good explanation.

Both the Genesis and RED One sample the image - both using a colour filter array. However RED uses quite complex image reconstruction algorithms and AFAIK, the Genesis uses none. The optical low pass filter in the RED is set appropriately for it's 4k output resolution. I have no idea on the Genesis - I need to point it at a chart to see what it's doing. I'd assume the OLPF was set more for 1080p output, but I'd not stake my reputation on that without measuring actually.
Click to expand...

Of course, it should also be noted that the Red and the Genesis had considerably different design targets. The Red was designed to record RAW files that would later be reconstructed. The Genesis was designed to provide a final quality live output that could be recorded on a video based recording device. The reasons for those design targets are irrelevant to the fact that they are, to be sure, two very different approaches.

The Arri D21, however does generally downsample - from the 3k demosaic to HD output. In the RED has a nice 2k from 4k high quality downsample mode in post that is similar in concept in that you reconstruct then downsample for best quality. The image quality doing that is visibly superior to a simple extraction (which is similar to how basic reconstruction of an RGB stripe works).
Click to expand...

Yes, which is why I didn't use it as an example for this discussion. It should also be said, though, that the D21 also provides RAW output, which allows a software based reconstruction that can be either 2K or 3K, if desired. In live video mode, however, the output is HD (1920x1080), just as you stated.
 
Yup, you design the whole camera "as a system" from the light, through the lens, all the way through to how it gets displayed. You try to take "everything" into account and balance the variables to make your product unique and to suit you and your customers needs.

And that's why each of the cameras we've mentioned are as they are.

Graeme
 
why the name?
this should be known as reduser's canon forum
 
You must log in or register to reply here.
Back
Top