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 :)

Curious Case of Benjamin Buttons

Hi Greg

It's to bad you got poor information.

By the way is that a new website you have Greg? It looks good.
 
claudio said:
The files the Viper and F23 spits out are dpx files. We are recording to S2 hard drives.
Click to expand...

I'm not nitpicking, but the Viper and the F23 output video streams - in your case, 4:4:4, RGB dual link video. The S.two takes these streams and writes them to a disk array as a DPX file sequence. I just don't want people to be mislead and think that the cameras themselves are file based and output file sequences when that is not the case.

Claudio, I think perhaps you are experiencing something very similar to what I did prior to last NAB. I went in believing that compression at the beginning of the storage/processing chain was unacceptable, and that numbers don't lie. What I found myself thinking after seeing a number of tests from both Red and Cineform (via the Silicon Imaging camera) was that perhaps I was overstating the issue, and that given a higher original image resolution and properly done compression it was possible to achieve what I did not think possible, and with some serious workflow advantages to boot. The real lesson I took away from that show - and I wasn't alone in this - was that numbers sometimes do lie, and that the result is what matters, not what it took to get there. Certainly your eyes don't lie, and any dynamic range issues are things that should concern you, but I wouldn't look at the numbers - particularly regarding compression ratios - and draw conclusions directly from that. Particularly in the case of what Red is doing. Wavelet compression from a 4K Bayer source, done on the Raw data prior to debayering is very different than DCT, JPEG, or MPEG compression done on RGB image sources.

Of course, as with the Dalsa, the Silicon Imaging, or any digital still camera, the Red images require a full quality debayer pass to really yield the best results. And that is not quick - and many of the images already made available were not processed that way. Debayering artifacts are often mistaken for compression artifacts, even though they're not really part of the captured image and not the result of any compression - and are eliminated with a properly done, full debayer. So make sure you're comparing apples to apples, and that those apples are fully ripe.
 
Research has shown that if you're going to compress a digital image, the "best" place to apply the most compression is as near to the start as you can. At that point, least has been done to the image that can cause the compression to be less efficient.

Graeme
 
Graeme Nattress said:
Research has shown that if you're going to compress a digital image, the "best" place to apply the most compression is as near to the start as you can. At that point, least has been done to the image that can cause the compression to be less efficient.

Graeme
Click to expand...

huh?

exactly what research are you referring to? it certainly goes against my experiences, and what I'd assumed would be common sense, and common practice, at least in my experience.

Why would compression become less efficient if "work" has been done to it?

(like... say... noise reduction, for example...)

R.
 
mmost said:
I'm not nitpicking, but the Viper and the F23 output video streams - in your case, 4:4:4, RGB dual link video. .
Click to expand...

You are correct. I miswrote. I should have said the system. But what I am trying to get to the heart of the matter is why I am unimpressed with the images i have been presented. Maybe the presentation was not up to par. For me it seems the compression is a place to look. I really do not care about the math as much as the result.

I do need to reevaluate the RED camera. There has been many updates. Since my posting on this site, it looks like I am getting my tests. I am very open and looking forward to these tests. After these tests, I will have a more concrete opinion about the camera.
 
Hi claudio

I sure see nothing wrong with those Blacks :
152_1196012584.jpg


Here is the link to the original uncorected 4k picture:
http://digitalfx.tv/clients/red/offhollywood/suspicion_03.tiff

But I'd like to know what someone of your experience would be looking for in those black.

Yours

Emmanuel
 
Yes, NR can make an image more compressable. What the research was, and I think I heard about it from a BBC engineering contact, was that doing the heaviest compression at the end, for delivery, was, theoretically, the worse place to apply it. Applying strong compression at the head, with the cleanest signal, was preferential. Of course, this is a "general" thing, and some post production can make an image more compressable, whereas a lot of things can make it worse.

We're in a slightly different game as we're dealing with RAW sensor information, and I think we have shown that a reasonable amount of compression at the head of the chain can be done very well.

Graeme
 
Graeme Nattress said:
Yes, NR can make an image more compressable. What the research was, and I think I heard about it from a BBC engineering contact, was that doing the heaviest compression at the end, for delivery, was, theoretically, the worse place to apply it. Applying strong compression at the head, with the cleanest signal, was preferential. Of course, this is a "general" thing, and some post production can make an image more compressable, whereas a lot of things can make it worse.

We're in a slightly different game as we're dealing with RAW sensor information, and I think we have shown that a reasonable amount of compression at the head of the chain can be done very well.

Graeme
Click to expand...

Well, it's pretty easy to test this. I knocked up a grad in photoshop in 16 bit colour, and did an extreme curves colour correction before and after saving as REALLY heavily compressed (quality 1) jpg.

here's the uncorrected grad
http://www.ruairirobinson.com/red/16bit_grad.tif
and the curve I'm using: http://www.ruairirobinson.com/red/colour_correction.acv

and the colour correction done in 16 bits
http://www.ruairirobinson.com/red/16bit_cc_grad.tif

here's the colour correction done AFTER compressing in jpg quality 1
grad_compressionfirst_ccafter.jpg


and the correction done BEFORE saving as jpg quality 1.
grad_ccfirst_compressionafter.jpg


As you can clearly see, the version with the correction done after compression is way way crappier looking.

Obviously this is an extreme example, but it the point remains. Compression is MUCH better at the very end of the chain if you want to retain image quality.

R.
 
Emmanuel Cambier said:
Hi claudio

I sure see nothing wrong with those Blacks :

But I'd like to know what someone of your experience would be looking for in those blacks.
Click to expand...

It is not about blacks hitting zero. A jpeg does a good job with that. The files I get from the Viper/F23 look pretty flat so the image is very malleable. The more dynamic range I have, the more chances I have of bringing down a blown out sky or lifting something out of the blacks.

I did open your file. I discovered no new information in the blacks.

Here is an example:

http://www.claudiomiranda.com/heinekenttl.html
 
Ruairi Robinson said:
here's the colour correction done AFTER compressing in jpg quality 1
Click to expand...

You're making a pretty theoretical argument here.

The proper way to test something like this is compare an uncompressed frame from RED with a frame compressed by RedCode Raw and then uncompressed again.

If you want to capture uncompressed RED footage it'll be coming when that capture solution is added to the camera. Until then the image is going to get compressed and it's either going to be good enough for you or not.

What I do know is I've seen at least one compositer who has a lot of experience with film posting that the keys he got out of the RED footage he tested were the best keys he'd ever seen. Better than film. Clearly many more people will continue to test that.
 
Here is a DPX file that is about as full range as I can find. It will appear very flat, it has almost no correction applied, only a small bump in brightness and little bit of highlight retention.
SamPullUp.jpg

Ctrl Click to download. (32MB)
http://www.redninjas.net/SamPullUp.dpx
 
I think what Claudio needs is several frames directly from the camera, Raw. And let him have at it with RedCine. The frames should have extremes, in both over and underexposure, as well as some flat images. That way he can discover what lies within much like converting a Raw image from a still camera. Raw files have a ton of dynamic range that converted images almost never have.

Also when I got my D200 a while ago I did tests on both compressed and uncompressed Raw files and discovered that the compressed Raw files looked almost the same as the uncompressed Raw files after conversion, debayering. I shot a lot of subjects over a week or so since I started a 2 month QTVR panorama project soon thereafter. I was amazed with the results. Until then I had been a Uncompressed all the way kinda guy. Modern compression done really well early in the image chain doesn't seem degrade the image too much and saves a huge amount of bandwidth and storage.
 
Umm, what Claudio needs is the camera to test. His Viper tests were far more inclusive then a few random ones, including enviro tests (which red could use right now) :wink:. Claudio that Heineken commercial is one of my favorites. :wink:

There were a few names on a list I had who would do a great job of testing the camera while knowing the digital and film equivalents very well. Claudio was at the head of the list, Mr.Mullen, Dion Beebe, Mathew Libatique, and a few others.

Claudio, I would LOVE if you post any test results here, with you usual mix of perspicacity and honesty; it would be very much appreciated ...

:beer:

Obrigado
 
Evin Grant said:
Here is a DPX file that is about as full range as I can find. It will appear very flat, it has almost no correction applied, only a small bump in brightness and little bit of highlight retention.
Click to expand...

Hi Evin.
Do you generally see this level of compression noise in the blue channel?
 
All I want in this world is for Claudio to test the Red the way he wants and to post his results.

I would also love to know what exactly him and David saw, was it the wedge tests, milkgirls, or what...?
 
What about Claudio participating in the upcoming LART (L.A. RED Test). I think it would be great to have another unbiased view based on experience with other digital cameras.

Gibby, Ken, what do you think?
 
claudio said:
It is not about blacks hitting zero. A jpeg does a good job with that. The files I get from the Viper/F23 look pretty flat so the image is very malleable. The more dynamic range I have, the more chances I have of bringing down a blown out sky or lifting something out of the blacks.

I did open your file. I discovered no new information in the blacks.

Here is an example:

http://www.claudiomiranda.com/heinekenttl.html
Click to expand...


Well I think I start to see what you mean, it is true that the original file does not contain more informations in the black zones but just some kind of grey, wich is not very helpfull, I agree.
The thing is, and this the meaning of my first question, this grey zone is not suposed to retain any relevant informations, as the Jpeg shows, it's just meant to be solid black, so here's my question again, what are you expecting to find in those blacks ?
I mean the picture doesn't seem under or overexposed, at least not the woman nor the background, it's just the black car's roof (bottom right corner) that doesn't seem to have readable informations, but does it matter so much since you'd probably want to have those black really black anyway.

Poor Claudio, it must feel like hiting a hornet's nest, but you'll survive, don't worry.

Yours

Emmanuel
 
You must log in or register to reply here.
Back
Top