ARRI CEO: "We want to have the worlds best digital camera this year".

[laguun]

your way will provide max 2k or lower in cinema!

Foy your information, mr. radeck, any 35mm projection was is and will be below 2k.

shoot in 35mm scan with 8k or 12 k (the next scanner generation!) finish in 4k and project in 4k would be much better and is the Vision of DCI!

mr radeck, i am pretty sure you never saw 65mm or 4k in 4k projection.

if you would have, then you would understand why 2K is the dominating standard.

i recommend that you stay with PAL and wait for the public state broadcasters to go 720p50.

arri lost the marketleadership. i hate that.

but after 8 straight years telling them every single tradeshow that they would have to build a red-style camera, they lost the market. for good reason.

I suppose by the end of the year, there will be more red 1s than 235, 435, studio (lts) etc. combined.

arris mistake.
wrong aquisition medium.

 

[Bruce Allen]

What the heck is that???

It's Digital Magic, that's what it is!

This one is funnier:
http://www.boacinema.com/misc/lucas.mov

...but the first one seemed more on-topic RE: Western and Eastern aesthetics (and their possible fusion) in the HD camera world

Bruce Allen
www.boacinema.com

 

[Finner]

Sometimes I can't believe the shit I get paid to shoot. That stuff takes the cake though Bruce. I guess he hopes the new indy will save him. I would guess his car would have to be in danger of repossesion to take part in that.

 

[Poi Boy]

wow that is really ugly.
-A

 

[Gavin Greenwalt]

What's he doing hawking Panasonic Cameras though? He's shot exclusively sony has he not?

Anyway... back on topic:

Hence forth whenever I need to provide feedback on some work that's just lacking that little-certain-something I'm going to ask for "more Joofa". This shot is lacking in Joofa. Gimme more Jooffa! There! Now that's Joofa!

OH wait.... that wasn't the topic either... what were we talking about?

 

[Jeff Kilgroe]

What happens in japan stays in japan.

Uh... That didn't stay in Japan. It should have, but well, it didn't. At least there weren't any Ewoks.

 

[Jeff Kilgroe]

what were we talking about?

I don't know... The clapping storm troopers made me forget.

 

[Tom Lowe]

8K, maybe, and definitely 12K are beyond the maximum resolution for 35mm. What's the point of oversampling?

Anything beyond 6K seems very unlikely to me with 35mm, and much 35mm will barely hold up at 4k.

 

[GlennChan]

you can rank the two sensors subjectively using the human response as the base line.

At a certain level, that makes a lot of sense. And I think for particular applications it would be the way to go (e.g. medical).

But when it comes to filmmaking, part of it is about art. Sometimes we want to shoot at a high frame rate to get those slowmo shots. We want to shoot at some really excessive framerate that is more than enough for human vision... like what people are doing with the Phantom right now.
In terms of color, it is extremely likely that the color will be manipulated in DI. So the ideal imaging system has to stand up to whatever is done in DI... it needs to be better than human vision to be "perfect".

 

[Dj Joofa]

At a certain level, that makes a lot of sense. And I think for particular applications it would be the way to go (e.g. medical).

But when it comes to filmmaking, part of it is about art. Sometimes we want to shoot at a high frame rate to get those slowmo shots. We want to shoot at some really excessive framerate that is more than enough for human vision... like what people are doing with the Phantom right now.
In terms of color, it is extremely likely that the color will be manipulated in DI. So the ideal imaging system has to stand up to whatever is done in DI... it needs to be better than human vision to be "perfect".

I agree with you on some levels. Human visual response may not be best under some circumstance that you describe. However, given a particular technology, we have to establish a baseline and many people have taken the path of choosing the human visual response as a baseline, as it is good one under many circumstances. But I see your point -- that it is not a definitive approach, is open to interpretation, and one can always establish a better baseline -- hence, I agree with you.

I think I should be a little more specific also. When I said baseline human response, I did not mean sticking a real human or human eye in there. I just meant statistical modeling based upon the perceptual tri-chromatic response of human visual system, that has been well-modeled, and also using a device independent space for distance measurement given a particular metric.

 

[GlennChan]

"Our Goal is to have the best digital camera of the world, within 12-14 months." (Nov.2007)

In my (speculative) opinion... I think they are up against some very tough competition.

--If they want to make a HD-SDI centric camera (known, fast, real-time workflow) then they're up against the F23.

--If they want to make a data-based camera, they are up against the Red, whatever Sony will come out with, Dalsa, maybe Panavision, etc. etc.
In terms of image quality, of what I've seen, it looks like Red's images are looking the best. The very low noise is huge... it affects how far you can go in DI.

I think there's a gap right now in that not all people know how to use their color tools that well... I'm seeing footage where the skies are clipping out... what I think people should be doing is applying curves-based adjustments so you have highlight handling like film and to avoid that blown-out look. You can even go more advanced and do something like shadow/highlight in AE/Photoshop to bring out shadow detail. Once you start to do these things, noise becomes much more visible so you want the camera to be low-noise to begin with.

--Workflow-wise, I think Red has a leg up on the competition among data-based cameras. (Cineform too, but not in a 4k camera currently AFAIK.) The artifacts from good compression are very negligible, yet they can make the workflow a lot more practical.

So for data-centric digital cinema cameras, I think it will be very hard to compete with Red.
Red has a headstart in data-based workflow, debayering, image processing, etc.
And the Mysterium sensor is very good.
And the price point is extremely good. (Though less of an issue for the very high-end.)

I don't see Arri competing on price or workflow, so they have to really excel in image quality.

 

[Zack Birlew]

Well, I think the next step is realtime HDRI capture. That automatically takes care of exposure drawbacks and captures a very high resolution image at the same time. All that would need to be done is to have processors fast enough to capture and convert every frame at three different exposures on the fly.

I'm not sure if that's even possible, but I know that idea has been brought up again and again. Anything beyond 4K is just too much I think. Honestly, how much resolution do you need? Already, people are freaking out over 1080p making them look bad. Now the new thing, which I saw for the first time today, is 120hz televisions. The difference really surprised me. So much for 24p! Everything looked like 60i video! There was a Bluray example of "Pirates of the Carribean: Dead Man's Chest" and the visual effects really stood out in a bad way. Things were clearer as far as movement was concerned, no more motion blur. But it made everything more noticeable, such as the makeup and other small details. Yeesh! Granted, 8K could add a lot for certain things, particularly large epic films, but other than that I just can't see it being all that useful.

It's funny how nobody was really talking about "K" anything until we got a 4K camera making the news and breaking the industry norm (in a good way). Now I think people are just going to get "K crazy" like how they've gotten "HD crazy", "Progressive crazy", "Digital crazy", "Tape crazy", ect. in the past. We've got a great camera here folks, enjoy it! 4K not enough? There's always IMAX!

Now, the only real question is "What about the 3D camera?".

 

[Dj Joofa]

Well, I think the next step is realtime HDRI capture. That automatically takes care of exposure drawbacks and captures a very high resolution image at the same time. All that would need to be done is to have processors fast enough to capture and convert every frame at three different exposures on the fly.

You are right about realtime HDRI being in demand. There is work going on currently using two approaches:

(1) "Real HDRI" capture where the sensor is able to see a wide range of at least over 20 stops (I believe)

(2) "Simulated HDRI" where using a fast frame rate camera multiple images are derived at several exposures, and then they are collapsed into a single image using some approaches. The drawback in this case is if motion is fast enough, then it gets difficult to collapse different exposures as there is movement between different exposure images.

 

[Dj Joofa]

Oh, and one more comment on real-time HDRI capture. You do not have to increase the range of your digital sensor ADC to increase its dynamic range. While keeping the fixed range, you have to reduce the noise. Many times the standard deviation of the temporal noise is taken to mean the physical interpretation of bit 1 on a linear digital sensor. Hence, you may keep your Vmax on an ADC fixed, but you increase the resolution of measurement of the minimum signal above noise (i.e., having a sensor with less noise) that you can measure and you have a higher dynamic range for the same fixed range.

An e.g, say your range is 3 volts and your minimum physical measurement is only 1 volt then your dynamic range is 3 / 1 = 3 steps. However, if you are able to decrease the size of your minimum step to 0.5 volts then your dynamic range is 3 / 0.5 = 6 steps, while your range is still 3.

 

[GlennChan]

When I said baseline human response, I did not mean sticking a real human or human eye in there.

Hmm, I think we really do need to use human eyes for testing.

A lot of the theoretical models about human vision don't work 100%. e.g. I believe that some of the sampling theory out there leads us down the wrong path. Some of the theory considers (approximations of) sinc filters (AKA brickwall) to be the best, but I don't think they are ideal for chroma resampling or upscaling.

Another example: all the Nyquist stuff doesn't work for human vision. Human vision has (vernier) resolution that far exceeds the Nyquist limit. Classic sampling theory only works if you assume that the sampling is not moving / the phase of the sampling is not shifting. Our eyes move... that's probably how our visual system can achieve very good vernier resolution.

Until our theoretical models are very good (they are not), the best thing to do would be to use humans during testing. Subjective testing has some issues (e.g. bias, learning, how good is the sample), but I think these issues are less problematic than objective testing.
And almost all of our theoretical models for how human vision works is derived from human observations anyways (e.g. CIE 1931).

2- The other thing to do is real world testing. How good does something look after it's gone through color grading and so forth?
And in real world viewing situations, surround illumination, ambient light, and the limitations of the display all affect how we perceive an image.

 

[Dj Joofa]

Hmm, I think we really do need to use human eyes for testing.

A lot of the theoretical models about human vision don't work 100%. e.g. I believe that some of the sampling theory out there leads us down the wrong path. Some of the theory considers (approximations of) sinc filters (AKA brickwall) to be the best, but I don't think they are ideal for chroma resampling or upscaling.

Another example: all the Nyquist stuff doesn't work for human vision. Human vision has (vernier) resolution that far exceeds the Nyquist limit. Classic sampling theory only works if you assume that the sampling is not moving / the phase of the sampling is not shifting. Our eyes move... that's probably how our visual system can achieve very good vernier resolution.

Until our theoretical models are very good (they are not), the best thing to do would be to use humans during testing. Subjective testing has some issues (e.g. bias, learning, how good is the sample), but I think these issues are less problematic than objective testing.
And almost all of our theoretical models for how human vision works is derived from human observations anyways (e.g. CIE 1931).

2- The other thing to do is real world testing. How good does something look after it's gone through color grading and so forth?
And in real world viewing situations, surround illumination, ambient light, and the limitations of the display all affect how we perceive an image.

Some very good points and I agree with them. One comment on sampling though. For a fixed static gaze the human vision at the center of fovea does follow Nyquist -- there are about 120 cone photoreceptors at the center of fovea per degree and that gives the maximum frequency as 60 cycles / degree. The rods are smaller in size than cones, however, several are joined together (are they doing pixel bining ;-))))) and hence, offer blurry image than cones. Though in practice the maximum frequency response seen by the human eye in the opponent color space (luminance, red-green, blue-yellow), is around at 10 cycles per degree, less for blue-yellow than red-green.

 

[Dj Joofa]

Hence forth whenever I need to provide feedback on some work that's just lacking that little-certain-something I'm going to ask for "more Joofa". This shot is lacking in Joofa. Gimme more Jooffa! There! Now that's Joofa!

I shall have to charge royalty from you for this usage ;-)))))))

 

[Deanan]

It's also worth noting that what is technically correct is often not desirable. For example, we could increase sharpness but would also impact aliasing, perceptual depth of field (making it really hard to pull focus like some other cameras with a weak low pass), and quality of image (ie. harsher vs more pleasing).

The same can be said for lenses. For example, one the lenses I have is 28mm Leica M mount and is superb from a technical standpoint but lacks beauty and life. Same could be said of Cookes vs UltraPrimes. Lens design is one of the fields where being imperfect in some areas is actually desired.

Another example is framerate or grain. 24fps could easily be considered less than technically advantageous but when compared to technically superior 60p , 24fps is often more desirable. Same for grain.

 

[GlennChan]

Some very good points and I agree with them. One comment on sampling though. For a fixed static gaze the human vision at the center of fovea does follow Nyquist -- there are about 120 cone photoreceptors at the center of fovea per degree and that gives the maximum frequency as 60 cycles / degree. The rods are smaller in size than cones, however, several are joined together (are they doing pixel bining ;-))))) and hence, offer blurry image than cones. Though in practice the maximum frequency response seen by the human eye in the opponent color space (luminance, red-green, blue-yellow), is around at 10 cycles per degree, less for blue-yellow than red-green.

If you really think about it... I think even there Nyquist has been violated.

At 60 cycles/degree, it doesn't quite seem possible to have zero ringing, little aliasing (*depends on what you're looking at), AND resolve 60 cycles/degree. Theoretically, you can't have all three at once.

In any case it's a moot point... human vision doesn't follow Nyquist.

 

[GlennChan]

It's also worth noting that what is technically correct is often not desirable.

Holga photography anybody?