November 27, 2014, 01:16:26 PM

Author Topic: 5D Mark III & RAW Video, A Case Study  (Read 13471 times)

Canon Rumors

  • Administrator
  • Canon EF 300mm f/2.8L IS II
  • *****
  • Posts: 2767
    • View Profile
    • Canon Rumors
5D Mark III & RAW Video, A Case Study
« on: May 30, 2013, 10:09:59 AM »

Magic Lantern is onto something…

There have been a lot of posts lately about Magic Lantern’s breakthrough enabling RAW video recording on the Canon EOS 5D Mark III.


Below is an ISO test of the standard H.264 of the 5D3, and what the RAW video from Magic Lantern looks like. This was done by Dustin J Tolman.


The results are pretty remarkable and easy to see, this software is still in its early stages and will probably get a lot better over the next few months.



Workflow

An ISO test comparing Canons H264 compression vs Magic Lantern’s 14 Bit RAW hack.


• First I converted the RAW files with ML’s raw2dng.

• Next I dragged the files into Photoshop which opened up Adobe Camera Raw.

• I selected the first image and used the auto white balance setting which got rid of the heavy magenta tone.

• Then I synchronized the additional files, and exported them as 16bit tiff files.

• First I tried to use QuickTime 7 to create an image sequence, but it render the video with incorrect color.

• So instead of QT I ended up using Final Cut Pro X to create an image sequence to a ProRes 422 HQ output.


Canon 5dmk3 @ F8

Canon 70-200 2.8L IS II @ 200mm

Technicolor Cinestyle


Music – Soundgiant


Read more about this test at Planet5D.com | EOS 5D Mark III at B&H Photo


cr


 


« Last Edit: May 30, 2013, 10:15:34 AM by Canon Rumors »
canonrumors.com

canon rumors FORUM

5D Mark III & RAW Video, A Case Study
« on: May 30, 2013, 10:09:59 AM »

Mt Spokane Photography

  • EF 50mm F 0.7 IS
  • *********
  • Posts: 9182
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #1 on: May 30, 2013, 10:19:15 AM »
I noticed some distracting jagged edges of the magenta plant to the left of the flame on the ML video.  They were not visible on the Canon version, likely because of the compression.
I wonder if they were artifacts of the raw processing or if they are there on the Raw images.

bchernicoff

  • Canon 7D MK II
  • *****
  • Posts: 563
    • View Profile
    • My Photos
Re: 5D Mark III & RAW Video, A Case Study
« Reply #2 on: May 30, 2013, 10:45:04 AM »
I did a test of my fiance sitting on our balcony. She has red hair and was wearing a royal blue sweater. The difference in color and detail was amazing. It really is incredible what ML have unlocked. If you have After Effects, the workflow isn't even difficult. Once you have converted the .RAW file to DNG files using the raw2dng, AE can import the DNG files easily. It then opens Adobe Camera RAW for you to tweak the first frame which it then applies to the rest.
6D, Fuji X-T1, X-E1
Tamron 24-70 2.8 VC, 70-200 f/2.8L IS II, 400mm f/2.8L II, 100mm L IS Macro, Sigma 85mm, & 35mm f/1.4's, Rokinon 14mm f/2.8,
Fuji 23mm 1.4, 35mm 1.4, 56mm 1.2, 14mm 2.8, 18-55, 55-200

Fr3lncr

  • Power Shot G7X
  • **
  • Posts: 11
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #3 on: May 30, 2013, 11:09:35 AM »
Hmm... I prefer the Canon one over the Magic Lantern one.

cayenne

  • 1D X
  • *******
  • Posts: 1226
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #4 on: May 30, 2013, 11:14:27 AM »
I did a test of my fiance sitting on our balcony. She has red hair and was wearing a royal blue sweater. The difference in color and detail was amazing. It really is incredible what ML have unlocked. If you have After Effects, the workflow isn't even difficult. Once you have converted the .RAW file to DNG files using the raw2dng, AE can import the DNG files easily. It then opens Adobe Camera RAW for you to tweak the first frame which it then applies to the rest.

I hope that at some point, they can figure how to convert the DNG raw out of the camera, into a DNG format that be piped straight into Davinci Resolve, so you can use that to do the color correction/grading right out of the camera....Resolve is such a powerful tool, it would amazing to be able to come out of camera into Resolve to grade it, output to ProRes and then into any editor of your choice....

Niki

  • Rebel T5i
  • ****
  • Posts: 127
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #5 on: May 30, 2013, 11:28:39 AM »
Is this working on the 5d mll yet???  If so is it working as well?

bereninga

  • PowerShot G1 X II
  • ***
  • Posts: 63
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #6 on: May 30, 2013, 11:43:16 AM »
The raw looks too warm. But I didn't like the halo around the fire of the candle of the H.264 at 200%. Interesting stuff and I can't wait for ML to get to the 6D, if it ever does.

canon rumors FORUM

Re: 5D Mark III & RAW Video, A Case Study
« Reply #6 on: May 30, 2013, 11:43:16 AM »

Nishi Drew

  • Canon 70D
  • ****
  • Posts: 254
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #7 on: May 30, 2013, 11:51:47 AM »
The raw looks too warm. But I didn't like the halo around the fire of the candle of the H.264 at 200%. Interesting stuff and I can't wait for ML to get to the 6D, if it ever does.

Well guess what, it's RAW and a simple adjustment in white balance and your gold! (or silver, for the cooler color??)
And I recall some early attempts/tests proving the 6D is capable as well, but alas, extreme card speeds are required and both the current limitations of SD card speeds and the major issue, the uber slow transfer from buffer to card in the 6D kills it. You might be able to run lower res sequences around 720P but still doesn't look great. Now the 50D looks to be capable of RAW video as well, and that camera has fast transfer speeds so with the fastes SD cards (95mb/s) one can achieve desirable results.

bluegreenturtle

  • Rebel SL1
  • ***
  • Posts: 91
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #8 on: May 30, 2013, 11:53:27 AM »
The raw looks too warm. But I didn't like the halo around the fire of the candle of the H.264 at 200%. Interesting stuff and I can't wait for ML to get to the 6D, if it ever does.


Remember in looking at all these raw samples that much of it is entirely the decisions that the person who took it made in color correction in post - 90% of which I've found to be questionable.  Usually they are either garish colors or not correcting the magenta slant that comes straight out of the camera. 

Tom W

  • PowerShot G1 X II
  • ***
  • Posts: 46
    • View Profile
Re: 5D Mark III & RAW Video, A Case Study
« Reply #9 on: May 30, 2013, 11:56:30 AM »
The difference is amazing, especially at high ISO.

What is the H.264 doing to the image to destroy the contrast and brilliance of the color?

bchernicoff

  • Canon 7D MK II
  • *****
  • Posts: 563
    • View Profile
    • My Photos
Re: 5D Mark III & RAW Video, A Case Study
« Reply #10 on: May 30, 2013, 12:10:50 PM »
The difference is amazing, especially at high ISO.

What is the H.264 doing to the image to destroy the contrast and brilliance of the color?

It's not necessarily the compression. Most Canon DSLR sites recommend you turn the in-camera saturation, contrast, and sharpening down all or most of the way then add those grading steps later in your editor. This gives you the ability to tune it to the scene. When you import the DNG video frames using Adobe Camera RAW using the defaults, a contrast curve, and sharpening are applied. When I did the comparison of my fiance, I used something like -51 contrast, -20 saturation, and no sharpening in ACR to try to have a fair comparison.

I would suspect that what we are looking at here is just that...the H.264 was shot with in-camera processing turned down, but the ACR defaults.

I believe when the ML guys were posting their early success of getting RAW frames they mentioned that internally Canon is down-rezing the video to something like 720p before upscaling it to 1080p and converting to H.264. This would explain the lack of detail in Canon's H.264 output. They had to work around this to get real 1080p output. In other words, H.264 itself isn't the problem...the Vimeo video above is likely H.264 compressed. Also so are Bluray movies.
6D, Fuji X-T1, X-E1
Tamron 24-70 2.8 VC, 70-200 f/2.8L IS II, 400mm f/2.8L II, 100mm L IS Macro, Sigma 85mm, & 35mm f/1.4's, Rokinon 14mm f/2.8,
Fuji 23mm 1.4, 35mm 1.4, 56mm 1.2, 14mm 2.8, 18-55, 55-200

Drizzt321

  • 1D X
  • *******
  • Posts: 1675
    • View Profile
    • Aaron Baff Photography
Re: 5D Mark III & RAW Video, A Case Study
« Reply #11 on: May 30, 2013, 12:23:06 PM »
The difference is amazing, especially at high ISO.

What is the H.264 doing to the image to destroy the contrast and brilliance of the color?

It's not necessarily the compression. Most Canon DSLR sites recommend you turn the in-camera saturation, contrast, and sharpening down all or most of the way then add those grading steps later in your editor. This gives you the ability to tune it to the scene. When you import the DNG video frames using Adobe Camera RAW using the defaults, a contrast curve, and sharpening are applied. When I did the comparison of my fiance, I used something like -51 contrast, -20 saturation, and no sharpening in ACR to try to have a fair comparison.

I would suspect that what we are looking at here is just that...the H.264 was shot with in-camera processing turned down, but the ACR defaults.

I believe when the ML guys were posting their early success of getting RAW frames they mentioned that internally Canon is down-rezing the video to something like 720p before upscaling it to 1080p and converting to H.264. This would explain the lack of detail in Canon's H.264 output. They had to work around this to get real 1080p output. In other words, H.264 itself isn't the problem...the Vimeo video above is likely H.264 compressed. Also so are Bluray movies.

They're down-rezzing to 720p before upscaling to 1080p for output? Really? Wow.

I'm thinking I might put this on once it's more of a beta/official release. I have a few video friends, and I want to record a bit of something, hand them the file and ask them to see what they think :)
5D mark 2, 5D mark 3, EF 17-40mm f/4L,  EF 24-105mm f/4L IS USM, EF 135mm f/2L, EF 85mm f/1.8
Film Cameras: Mamiya RB67, RB-50, RB-180-C, RB-90-C, RB-50, Perkeo I folder, Mamiya Six Folder (Pre-WWII model)
http://www.aaronbaff.com

bchernicoff

  • Canon 7D MK II
  • *****
  • Posts: 563
    • View Profile
    • My Photos
Re: 5D Mark III & RAW Video, A Case Study
« Reply #12 on: May 30, 2013, 12:53:44 PM »
They're down-rezzing to 720p before upscaling to 1080p for output? Really? Wow.

I'm thinking I might put this on once it's more of a beta/official release. I have a few video friends, and I want to record a bit of something, hand them the file and ask them to see what they think :)

I'm trying to find a more concrete source for my assertion than my memory ;-)

So far, I find this article which I think is the original one I read and then followed the link to their facebook page. Facebook is blocked at my work, so that is as far as I can go to track it down right now: http://www.eoshd.com/content/10221/magic-lantern-discover-2k-raw-dng-function-on-5d-mark-ii-and-iii
6D, Fuji X-T1, X-E1
Tamron 24-70 2.8 VC, 70-200 f/2.8L IS II, 400mm f/2.8L II, 100mm L IS Macro, Sigma 85mm, & 35mm f/1.4's, Rokinon 14mm f/2.8,
Fuji 23mm 1.4, 35mm 1.4, 56mm 1.2, 14mm 2.8, 18-55, 55-200

canon rumors FORUM

Re: 5D Mark III & RAW Video, A Case Study
« Reply #12 on: May 30, 2013, 12:53:44 PM »

Drizzt321

  • 1D X
  • *******
  • Posts: 1675
    • View Profile
    • Aaron Baff Photography
Re: 5D Mark III & RAW Video, A Case Study
« Reply #13 on: May 30, 2013, 12:59:52 PM »
They're down-rezzing to 720p before upscaling to 1080p for output? Really? Wow.

I'm thinking I might put this on once it's more of a beta/official release. I have a few video friends, and I want to record a bit of something, hand them the file and ask them to see what they think :)

I'm trying to find a more concrete source for my assertion than my memory ;-)

So far, I find this article which I think is the original one I read and then followed the link to their facebook page. Facebook is blocked at my work, so that is as far as I can go to track it down right now: http://www.eoshd.com/content/10221/magic-lantern-discover-2k-raw-dng-function-on-5d-mark-ii-and-iii

Hmm...don't see anything on that article or the front of the FB page. Ah well, I'll take it with a grain of salt then.
5D mark 2, 5D mark 3, EF 17-40mm f/4L,  EF 24-105mm f/4L IS USM, EF 135mm f/2L, EF 85mm f/1.8
Film Cameras: Mamiya RB67, RB-50, RB-180-C, RB-90-C, RB-50, Perkeo I folder, Mamiya Six Folder (Pre-WWII model)
http://www.aaronbaff.com

bchernicoff

  • Canon 7D MK II
  • *****
  • Posts: 563
    • View Profile
    • My Photos
Re: 5D Mark III & RAW Video, A Case Study
« Reply #14 on: May 30, 2013, 01:08:03 PM »

I'm trying to find a more concrete source for my assertion than my memory ;-)

So far, I find this article which I think is the original one I read and then followed the link to their facebook page. Facebook is blocked at my work, so that is as far as I can go to track it down right now: http://www.eoshd.com/content/10221/magic-lantern-discover-2k-raw-dng-function-on-5d-mark-ii-and-iii

Hmm...don't see anything on that article or the front of the FB page. Ah well, I'll take it with a grain of salt then.

okay, I found the section I was recalling. It's not quite as bad as remembered. The relevant paragraph is this (2k RAW frames refers to what the camera grabs for LiveView and video out):
Quote
The 14bit raw is turned into compressed 8bit 4:2:0 but rather than use all the native resolution of the 2K raw frames, the image processor is playing all kinds of tricks, at one point even downscaling to 1904 and then back up to 1920 again in the case of the 5D Mark III. Why does it do that?

Found on this page: http://www.eoshd.com/content/10250/canon-5d-2k-raw-feed-update-1920x720-possible-on-1000x-card
« Last Edit: May 30, 2013, 01:11:33 PM by bchernicoff »
6D, Fuji X-T1, X-E1
Tamron 24-70 2.8 VC, 70-200 f/2.8L IS II, 400mm f/2.8L II, 100mm L IS Macro, Sigma 85mm, & 35mm f/1.4's, Rokinon 14mm f/2.8,
Fuji 23mm 1.4, 35mm 1.4, 56mm 1.2, 14mm 2.8, 18-55, 55-200

canon rumors FORUM

Re: 5D Mark III & RAW Video, A Case Study
« Reply #14 on: May 30, 2013, 01:08:03 PM »