Film Analogy

[Dj Joofa]

I do believe either Jim or Graeme has said something along the lines of "the best 2k comes from 4k."

That is when we are just talking about the spatial detail of an image; and may not have a direct relation with the dynamic range of each pixel, if the dynamic range is high enough. Remember, dynamic range is a "per-pixel" quantity and detail is a "pixel-neighborhood" quantity -- they are two different measures. Detail resolution and dynamic range are often independently quoted, without proper understanding of the relation between the two.

The best response is the *joint optimization* of dynamic range and detail (resolution). I wrote a response on what is the best approach to simultaneously optimize both for the best visual experience, and is available at the following link:

http://www.reduser.net/forum/showthread.php?p=156192#post156192

 

[William Robinette]

Of course, because no matter what format you are shooting on RED, the pixels are always the same size and therefore have the same photosensitive properties. Just in 2k/3k you are using less of them when compared to 4k.

Basically your image would be exactly the same except for spatial detail when comparing a native 2k image to a 4k down sampled to 2k image right? I am ignoring the different compression ratios RED has implemented along with the discrepancies in FoV.

 

[Dj Joofa]

Basically your image would be exactly the same except for spatial detail when comparing a native 2k image to a 4k down sampled to 2k image right? I am ignoring the different compression ratios RED has implemented along with the discrepancies in FoV.

Yes, you are right, given a particular camera/sensor, say Red, in theory you should get better 2k images from 4k. My sole point was that today you have Red 4k, and you may not think that its image is always going to be better than a 2k camera from a different manufacturer, if in the joint space of dynamic range and detail, the visual response of the other manufacturer's 2k image is better.

 

[Andrew M.]

Of course, because no matter what format you are shooting on RED, the pixels are always the same size and therefore have the same photosensitive properties. Just in 2k/3k you are using less of them when compared to 4k.

Basically your image would be exactly the same except for spatial detail when comparing a native 2k image to a 4k down sampled to 2k image right? I am ignoring the different compression ratios RED has implemented along with the discrepancies in FoV.

Not quite right, the dynamic range depends from the surface of the pixels and pixels when downscaling are combined so the surface of the 4 pixels forming 2K output from RED is bigger then 1 pixel forming one pixel at 4K format.

Now, if downresing will be done directly on the sensor then we can combain 4 pixels current and feed it in to A/D converter having 4 times bigger current to sample. 80% equivalent to replacing the 4K sensor with 2K sensor for particular shot. (imagine it would be possible to insert and lock Misterium sensor as a user changeable module)

Now, playing with 4K RED RAW converted to 2K footage I discovered that grading colors and manipulating noise and sharpness yields much better results then I could have using windowed 2K RAW. Here the results are highly dependent from the algorithm I was using. Whole industry is optimized for processing 1:1 RAW versus final output. Looks like times are changing and we will be downresing a lot in the near future.

I see the need for grading tools that will take advantage of oversampling.

Just to start with, using standard grading tools I see that noise reduction and sharpening works way better while working on oversampled RAW data. (experiment with all algorithms, big difference of results here)

Still waiting for the grading tools that will reduce CA.
BTW does SRATCH has CA correction tool?

 

[William Robinette]

I understand that if you scaled 2k from RED's 4k Mysterium right off the chip you could have 4 pixels acting as one and gain said benefits.

But in it's current implementation RED does not do that. My entire post is based on the capabilities of the camera right now. It windows 2k/3k, so the area per pixel is the same regardless of what resolution you are shooting in.

All I'm saying is right now the benefits of 2k/3k are over cranking, storage, less processing required, etc... Items along those lines. It's not like shooting in 2k/3k gets you more latitude or sensitivity in it's current implementation. And 4k downsampled to 2k should be sharper then native 2k with the added benefits of 35mm DoF/FoV.

Correct?

 

[danbrazda]

Mr. Robinette... my gut tells me you and I are on the same page. However... there is always a danger (and I have fallen into this trap MANY times in the past when comparing film to video) of losing sight of the BOTTOM LINE, which is- storytelling. In the past, when electronic cinematography's bar was set by units like the Ikegami EC-35, I have fought to the death over film and its' aesthetic advantage. However...

I think RED is the first time we have seen a crossover from simplistic resolution/film-look debates to aesthetic debates (God, I wish I had a better word than "aesthetic" to describe what I really mean) and the first realistic possibility of something new, different and (dare I say) better than what we have known as the standard.

PS-

So as not to disappoint my fellow techno-nerds that were attracted to this thread in the first place- I'm still shootin' nothin' but 4k, MoFo's! :wink:

 

[Mohammed El Sharqawy]

my idea there wasn't about resolution, getting better 2K is obviously from downconverted 4K, but, here when shoot 2K instead you are going to loose the beautiful shallow depth of field provided by the 35 mm lenses,,, also using larger sensor area will give you more chances to process the footage in post and adding graininess or any picture effects without destructing the final image.. so, go go 4K with RED .. if you want more film look or any look control..

all the best,

 

[William Robinette]

the BOTTOM LINE, which is- storytelling.

Something too often and too easily forgotten.

 

[Andrew M.]

I understand that if you scaled 2k from RED's 4k Mysterium right off the chip you could have 4 pixels acting as one and gain said benefits.

But in it's current implementation RED does not do that. My entire post is based on the capabilities of the camera right now. It windows 2k/3k, so the area per pixel is the same regardless of what resolution you are shooting in.

All I'm saying is right now the benefits of 2k/3k are over cranking, storage, less processing required, etc... Items along those lines. It's not like shooting in 2k/3k gets you more latitude or sensitivity in it's current implementation. And 4k downsampled to 2k should be sharper then native 2k with the added benefits of 35mm DoF/FoV.

Correct?

Misterium sensor has ability to address single photocell like we can address RAM in our computer.

So I will disagree that we can’t do downresing directly on the sensor.

The question is the firmware that controls the Misterium and this will be updated again and again in the future. The limitation is 12 bit A/D converter that can’t handle more than 12 stops if used in liner mode.

However if you read my post carefully I mentioned that I had good result with noise reduction when processing oversampled data. It means that we can reduce noise, thus increase dynamic range.

You will tell me that getting more then 12 stops out of 12 bit A/D based sensor is impossible.

Well, wait a bit.

I just got such comment from the ARRI guy “you will never get as good picture from $2K lenses as from $25K Master Prime. Well he is up to big surprise, using Premiere Pro and CS3 I got perfectly CA clean clip shot on $2K lenses.

As to the story telling factor mentioned by Dan, well in Hollywood movies, yes the story is important, but in the nature, National Geographic, Science, underwater, etc. material, the storytelling is secondary item, the quality of presented material, and the craftsmanship of the DP is the most important factor.

 

[Stuart English]

4K v's 1080p

4K v's 1080p

Nope, dynamic range is determined by the bit depth of your sensor and its associated ADC and not by output resolution. I think you are mixing up detail with dynamic range. Higher detail should be present in 4k vs. 1080.

Well O.K, yes you are correct. Maybe its better to say that if you shoot RAW the mapping of the dynamic range available from the sensor can be delayed to post production, so that you have more control over how that dynamic range is shoe-horned into 1080p video, rather than letting the camera make that decision for you.

The sensor area captured when we enable 1080p will be 3840 x 2160. That compares to 4K RAW at 4096 x 2304 - so another advantage of 4K is the ability to pan / scan within the 4K area to find the best 3840 x 2160 section to performa a 2:1 scale from to generate a 1080p signal when in post.

Bottom line is - yes there are image quality & production flexibility reasons to shoot RED 4K v's any other 1080p camera and also RED 1080p when available.

 

[Dj Joofa]

Well O.K, yes you are correct. Maybe its better to say that if you shoot RAW the mapping of the dynamic range available from the sensor can be delayed to post production, so that you have more control over how that dynamic range is shoe-horned into 1080p video, rather than letting the camera make that decision for you.
The sensor area captured when we enable 1080p will be 3840 x 2160. That compares to 4K RAW at 4096 x 2304

Agreed. You are right. However, my basic point is that is I am considering optimizing the joint product of dynamic range and resolution and it is not always clear that what should be the trade-off of dynamic range vis-a-vis resolution.

... so another advantage of 4K is the ability to pan / scan within the 4K area to find the best 3840 x 2160 section to performa a 2:1 scale from to generate a 1080p signal when in post.

I agree with you. Pan/scan is a powerful option. However, I am not sure how users will be able to take advantage of this feature, and may not be more confused by it. Will there be framing issues to consider when shooting, as in post they will be extracting a certain smaller frame from the one originally shot?