Home · Register · Search · View Winners · Software · Hosting · Software · Join Upload & Sell

Moderated by: Fred Miranda
Username   Password

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

1       2      
3
       end
  

Archive 2013 · Magic Lantern improves 5D3 dynamic range to 14 stops!
  
 
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.3 #1 · Magic Lantern improves 5D3 dynamic range to 14 stops!


kevindar wrote:
For video, canon already had the lead, even before ML, b/c of sharper video images.


Not necessarily for sharpness they didn't. The D800 video, native, is sharper than that from all but the 1DX and 1DC (and even those only in certain modes). Of course the D800 video does have some aliasing and moire which the 5D3 avoids at least, if not the other Canon bodies (other than largely the 1DX and 1DC). Some preferred the D800 native video, although I personally preferred even the 5D3 native video despite the softness (although not the video from the 5D2/6D/xxD/Rebels/7D compared to D800 video) so I would say that the 5D3 video was already better (although some did disagree) than the D800 video even before ML but it wasn't because it was sharper because that it was not.

The D4 video had about the best high ISO performance of the DSLR video cams natively (other than perhaps 1DX or 1DC) but it was a squishy mess and I personally gave it a big fail.

Due to various factors in how they make their firmware AND hardware limitations in their liveview system it is 99% sure that the D800 will never be able to get ML firmware to get RAW video.

But with ML now the 5D3 just utterly whips ANY Nikon body for video every which possible way.

with ML, its in a different league with shooting raw video, and all the focus functionality.

For sure, a million times better shooting 5D3 with ML RAW and all the extra usability features than shooting video with D800. Maybe a billion.


This new addition is further useful, as it avoids motion artifacts.
For stills, I is a compromise. The group of photographers that have most desired the extra dynamic range have been landscape photographers. Loss of resolution, and interpolation artifacts is not something that sits very well with them. additionally, this needs to be further modified to also take a straight 100 ISO image, without interpolation. many times you really don't know if you need that extra DR, and since it comes at a cost (processing, as well as image quality) I would always like to play with my traditional image first
...Show more

Yeah as I much as I cry for more low ISO DR, this actually doesn't have me all that excited (it seems to me that the artifacts and resolution loss shouldn't be as disastrous as they appear now though with it, it seems that a very crude method is being used so it probably will get improved a good deal, we'll see if it can be improved enough though, if they went for less stops of improvement it might help it too (as it is some of the increase is fake anyway since if you trade half the resolution away the shadows will clean up anyway even just from a normal image)).



Jul 19, 2013 at 07:04 PM
johnctharp
Online
• • • •
Upload & Sell: Off
p.3 #2 · Magic Lantern improves 5D3 dynamic range to 14 stops!


One thing to consider is that the 'loss of resolution' won't be so bad when the final image is rendered to JPEG.

As well, it reminds me a little of how RAW processors must de-mosaic the Bayer'd sensor output. You have to wonder if the tools to process these images can't be improved in a similar way to mitigate the loss of resolution, especially if the image is rendered to a lower output resolution, which it almost certainly will be by the time it gets on paper or on a webpage.



Jul 19, 2013 at 07:12 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.3 #3 · Magic Lantern improves 5D3 dynamic range to 14 stops!


johnctharp wrote:
Thanks for sharing that Kevin. I didn't realize how well the D800 was doing in good light performance, though I have to wonder if there wasn't an issue with his lens setup. That doesn't make sense for someone like him in my mind, but I'd honestly expect both the Canon and the Nikon to record video that is as sharp as the recording format they're using, since they are both using far more than the ~2MP required by the output videos. I know that aspect ratios reduce the total sensor resolution available for video, but it has to be a
...Show more

It's not a lens issue. Much DSLR video simply is soft. The D4 video is soft. Almost all the Canon DSLR video is soft. Part of the problems come from the fact that most of the DSLR bodies can't read the entire sensor out and downscale (actually basically only 1DX/1DC can do that, although some of the other types of stuff like 4/3rds Panny stuff seems to do something somewhat fancy) and they have to skip reading lines and pixels combined with the fact that the AA filter is tuned for say 21MP or 36MP not for 2MP so the scale for the AA filter is all wrong so they sometimes try to blur things a bit, I think, to make the aliasing become not totally out of control, in some cases.

The 5D3 sensor is able to do some sort of fancy block binning so it can make use of the entire sensor without skipping (or at worst with only the most modest bit) although it's different than a regular full read and then downscale. What is so curious though is how ML found that the internal liveview/movie buffer stores images with such bit depth and resolution and how that all suddenly goes away during some later stage in the native movie process with the 5D3. It is NOT the compression engine since using a Ninja2 over the clean HDMI basically sees all but zero increase in resolution and the DR is basically the same too. It's almost like they apply a gaussian blur filter in their pipeline?!?? and then Digic's poor sharpening and toning engine massacres the rest of the signal?? Interestingly they don't use Digic in the C100-C300 series, but a much older video camera image processing chip they had laying around it might not massacre thing as badly as the Digic JPG engine does (although this still doesn't explain the almost seeming gaussian blur filter thing)



Jul 19, 2013 at 07:13 PM
johnctharp
Online
• • • •
Upload & Sell: Off
p.3 #4 · Magic Lantern improves 5D3 dynamic range to 14 stops!


Yeah, it's appalling how bad the video on any DSLR really is, compared to what it should be on paper. It makes a whole lot of sense from a marketing perspective, though, for Canon to ensure that their full-frame cameras do not compete fully with their their cinema line.

I mean, seriously, every Canon APS-C camera with video used with any currently produced lens outresolves the 2MP needed for 1080p by a factor of at least two; the better ones by a factor of five to eight. The D800 should do far, far better.

It feels like no one is actually taking the job seriously. None of this stuff is hard; imagine if the Blackmagic cameras had Canon or Sony (or Toshiba) sensors in them with the full capability of the processors and the firmware that goes along with them, recording at full 1080p or 4k to an SSD. That's what we should be able to do now, with these DSLRs, within the constraints of the SD and CF card interfaces.



Jul 19, 2013 at 07:21 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.3 #5 · Magic Lantern improves 5D3 dynamic range to 14 stops!


johnctharp wrote:
Yeah, it's appalling how bad the video on any DSLR really is, compared to what it should be on paper. It makes a whole lot of sense from a marketing perspective, though, for Canon to ensure that their full-frame cameras do not compete fully with their their cinema line.

I mean, seriously, every Canon APS-C camera with video used with any currently produced lens outresolves the 2MP needed for 1080p by a factor of at least two; the better ones by a factor of five to eight. The D800 should do far, far better.

It feels like no one is actually taking
...Show more

The fact that they have so many MP makes it harder. It's hard to read all of them and the native AA filter is wrong scale.



Jul 19, 2013 at 08:40 PM
johnctharp
Online
• • • •
Upload & Sell: Off
p.3 #6 · Magic Lantern improves 5D3 dynamic range to 14 stops!


skibum5 wrote:
The fact that they have so many MP makes it harder. It's hard to read all of them and the native AA filter is wrong scale.


Just need more memory/faster processors, neither of which are actually expensive when considering the cost of the camera to produce. But the point I'm making is not that they can or can't do it, rather that they choose not to.

As an aside, I do agree that the AA filter is the 'wrong size' for video, but I don't think that argument holds much weight- they should be taking the full size of the sensor that the aspect ratio being shot allows and down-sampling that to the appropriate output resolution. The AA filter is the perfect size for the sensor's native resolution, so it shouldn't be a factor, if they were in fact taking full advantage of the sensor for video.

I mean, 1080p video shouldn't be 'soft' on any DSLR. That's just silly, and points to an intentionally engineered limitation for the sake of product segmentation and market stability.



Jul 19, 2013 at 09:03 PM
kevindar
Offline
• • • •
Upload & Sell: Off
p.3 #7 · Magic Lantern improves 5D3 dynamic range to 14 stops!


thanks for clarification on video. I could swear I saw a review on you tube, which showed how poor the video of both (5d3 and d800) are, but that 5d3 still held more detail. guess not.


Jul 19, 2013 at 10:04 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.3 #8 · Magic Lantern improves 5D3 dynamic range to 14 stops!


kevindar wrote:
thanks for clarification on video. I could swear I saw a review on you tube, which showed how poor the video of both (5d3 and d800) are, but that 5d3 still held more detail. guess not.


All the early posted video did make the 5D3 video look sharper as well as less aliased, but it turned out it was just that all the D800 video posts were either running it in a weird mode or using nasty uploading compression and such. I do recall being like what in the world are people going on about when they say the D800 video is sharper, it sure doesn't look sharper to me. But it turns out they were correct. That said I still prefer even the native 5D3 video for it's lesser issues with aliasing and moire (and easier usability, liveview and movie tracking on the D800 really is pretty poor), and again with ML no question 5D3 blows the D800 away for video every which way from quality to usability (using external recorders helps D800 somewhat since it apparently has a much worse internal codec than the 5D3 where external recorders make only very modest difference, but even D800 with uncompressed out is no match for 5D3 ML RAW quality although you might get some usability features back (zebras, peaking) with Ninjas and such (at many hundreds of cost extra)).

Anyway the 5D3 is a total video beast with ML, it makes it one of the greatest DSLR in history in my opinion. The quality of footage you can take is just amazing. A total revolution.



Jul 19, 2013 at 10:31 PM
jimmy462
Offline
• • •
Upload & Sell: Off
p.3 #9 · Magic Lantern improves 5D3 dynamic range to 14 stops!


Hi All,

looks like Canon's not going to take things laying down...

Canon Germany Teaser « Canon Rumors:
http://www.canonrumors.com/2013/07/canon-germany-teaser/

...a teaser for a July 31, 2013 release (German). Of note, the featured footage all shot bright midday sun, the highlights are preserved and the shadows are wide open. Hmm.

Interesting, Captain,
Jimmy G



Jul 22, 2013 at 01:33 AM
 

Search in Used Dept. 



skibum5
Offline
• • • • • •
Upload & Sell: Off
p.3 #10 · Magic Lantern improves 5D3 dynamic range to 14 stops!


jimmy462 wrote:
Hi All,

looks like Canon's not going to take things laying down...

Canon Germany Teaser « Canon Rumors:
http://www.canonrumors.com/2013/07/canon-germany-teaser/

...a teaser for a July 31, 2013 release (German). Of note, the featured footage all shot bright midday sun, the highlights are preserved and the shadows are wide open. Hmm.

Interesting, Captain,
Jimmy G


They used the 5D3 ML RAW video hack (or maybe a new Canon P&S with a Sony sensor ).



Jul 22, 2013 at 02:51 AM
johnctharp
Online
• • • •
Upload & Sell: Off
p.3 #11 · Magic Lantern improves 5D3 dynamic range to 14 stops!


jimmy462 wrote:
...a teaser for a July 31, 2013 release (German). Of note, the featured footage all shot bright midday sun, the highlights are preserved and the shadows are wide open. Hmm.

Interesting, Captain,
Jimmy G


You know, I didn't pay any heed to that video when I saw it on CR. Looks like I should have looked more closely .



Jul 22, 2013 at 08:31 AM
jimmy462
Offline
• • •
Upload & Sell: Off
p.3 #12 · Magic Lantern improves 5D3 dynamic range to 14 stops!


skibum5 wrote:
They used the 5D3 ML RAW video hack (or maybe a new Canon P&S with a Sony sensor ).


...or, maybe, Canon enabled their own code? As for the sensor, Canon being Canon, methinks they're using their own hardware but decided to incorporate film-like S-shaped gamma curves...

johnctharp wrote:
You know, I didn't pay any heed to that video when I saw it on CR. Looks like I should have looked more closely .


...some other intentionally-shown items in the video: it wasn't just that they continually showed lots of whites (shirts, towels cake icing, even) juxtaposed with both bright sun and deep shadows (even hand-held shots traveling between these lighting extremes), there were lots of shallow, razor-thin DOF shots (hinting at either larger sensor or DSLR?), and the demographic targets (and uses) as portrayed through characters (creatives, music, artists, food, culture, sports, architecture, nature, etc.)




Jul 22, 2013 at 01:17 PM
Gunzorro
Offline
• • • • •
Upload & Sell: On
p.3 #13 · Magic Lantern improves 5D3 dynamic range to 14 stops!


kevindar wrote:
thanks for clarification on video. I could swear I saw a review on you tube, which showed how poor the video of both (5d3 and d800) are, but that 5d3 still held more detail. guess not.


Kevin -- Perhaps that was the Philip Bloom video, where the D4 was also compared? Anyway, he has one where the Nikon looks soft and poor at high ISO and the Canon model shines.



Jul 22, 2013 at 02:14 PM
kevindar
Offline
• • • •
Upload & Sell: Off
p.3 #14 · Magic Lantern improves 5D3 dynamic range to 14 stops!


could be. it was early after release of both


Jul 22, 2013 at 02:37 PM
jimmy462
Offline
• • •
Upload & Sell: Off
p.3 #15 · Magic Lantern improves 5D3 dynamic range to 14 stops!


Maybe this is the video announcement for the 31st?

Canon 44.7MP DSLR Due at End of August?:
http://www.photographybay.com/2013/07/24/canon-44-7mp-dslr-due-at-end-of-august/

...4K UHDTV...

Ultra high definition television - Wikipedia, the free encyclopedia:
http://en.wikipedia.org/wiki/Ultra_high_definition_television

...which means Rec. 2020 color space and gamma versus the Rec. 709 color space of HDTV...

Rec. 2020 - Wikipedia, the free encyclopedia:
http://en.wikipedia.org/wiki/Rec._2020


Jimmy G




Jul 25, 2013 at 02:54 AM
jimmy462
Offline
• • •
Upload & Sell: Off
p.3 #16 · Magic Lantern improves 5D3 dynamic range to 14 stops!


Hi All,

For those looking for to see the actual ML HDR workflow involved...from downloading the software, installing the firmware, shooting a video sequence, uploading the video for processing, how the software compiles the 2 different ISO shots and creates the HDR frames, and, finally compiles the sequence into a video clip...there's this video walk-through (I have no affiliation with this producer, etc.)...

Magic Lantern HDR Workflow Tutorial - YouTube:
http://www.youtube.com/watch?v=VuhVZWaOVOY

...be mindful that the brief segment in the demo where they demonstrate downloading the ML firmware, they show the older 5D2 firmware download page not the newer 5D3 firmware. (Oopsie!)

Looks like this whole HDR hack is a bit of a kludge, currently, with some serious limitations for shooting fast motion (as demonstrated in the video). This is contrary information to that found in the OP about "fast moving subjects" and "no motion artifacts".

Seems to me that the workable the wider-DR solution will be found down the RAW video path.





Jul 26, 2013 at 02:40 AM
super35
Offline
• •
Upload & Sell: Off
p.3 #17 · Magic Lantern improves 5D3 dynamic range to 14 stops!


jimmy462, Magic Lantern HDR video is not the same thing as Magic Lantern Dual ISO video or still images. The workflow you posted above is for HDR video, not the new Dual ISO feature.

The idea is similar, but Dual ISO is alternating the ISO within a raw video frame or still image, whereas HDR video is alternating the ISO from frame to frame in a standard h.264 (not raw) video recording. Obviously, the workflows and results are completely different.

The big difference in use is Dual ISO video can be used with moving subjects where HDR video cannot because of severe ghosting.



Jul 26, 2013 at 06:03 AM
jimmy462
Offline
• • •
Upload & Sell: Off
p.3 #18 · Magic Lantern improves 5D3 dynamic range to 14 stops!


Hi Super35,

My apologies for the confusion...I had gone to the ML online User Guide and could not find an entry for the Dual ISO sub-menu, and I erroneously confused the HDR Video entry...

userguide [Magic Lantern wiki]:
http://wiki.magiclantern.fm/userguide#hdr_video

...for my search for a step-by-step video demonstration of this latest technique. Oopsie, on my part!

Hopefully, ML will issue an updated UG which both shows both the layout of the Dual-ISO sub-menu, and explains whatever functionalities are to be found there. (Some of us, apparently, confuse easily!)

On an updated note, after viewing this video demonstration of results obtained using Dual-ISO I will have to add a "kludge" rating (IMHO) to this new technique, as well...

Test dual iso magic lantern - YouTube:
http://www.youtube.com/watch?v=ZXqwequSBgE

...6-lines of skipped vertical resolution makes for dancing, mosquito moiré?! Well, at least the music matched. Oofah, I'll wait on further RAW developments.

Anyhoo, thanks again for setting me straight, I do appreciate it...not everyone will tell you when you've got parsley in your teeth!

Best to you,
Jimmy G



Jul 26, 2013 at 02:22 PM
1       2      
3
       end




FM Forums | Canon Forum | Join Upload & Sell

1       2      
3
       end
    
 

You are not logged in. Login or Register

Username   Password    Retrive password