Pages: 1 2 [3] 4 5 6   Go Down

Author Topic: Canon 5D Mk4 announced  (Read 32258 times)

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #40 on: August 31, 2016, 12:09:08 am »

If one is to average both, he needs to normalize them first to prevent banding. Normalization makes subtraction to restore the "B" pixels values problematic, not to mention that such normalization is problematic in itself, and Canon already suffered a lot of badmouthing for noise and especially for banding. Another variable that we do not know yet is the non-linearity of the highest 1 stop in the aux raw. If it is highly non-linear, I can see why Canon decided on clipping of the sum. Migrating to 15-bit format would be logical solution, but all third-party raw converters will be off-line for half a year or so (unless, of course, Canon would publicly document the new format, fat chance, really). Having no alternative to DPP folks would start complaining, that lesson from Kodak failure is well-learnt by both Canon and Nikon, and I think SONY are also realizing that. IMHO another reason not to go to 15-bit format for the main subframe is that that would also cause serious size increase, with very marginal benefit. On a side note, nothing is cast in stone, Canon may be changing firmware as we speak ;)
Logged

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3369
    • http://www.paulholman.com
Re: Canon 5D Mk4 announced
« Reply #41 on: August 31, 2016, 04:06:14 am »

Now this may not be a joke (thanks to Illiah for brining this to my attention):
http://www.rawdigger.com/howtouse/Canon-dual-pixel-technology
Interesting, but just because the data is there doesn't mean we can easily(ever ?) utilise it.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Canon 5D Mk4 announced
« Reply #42 on: August 31, 2016, 08:35:34 am »

Interesting, but just because the data is there doesn't mean we can easily(ever ?) utilise it.

Why not? It may take a bit of time before software developers pick it up because it's not mainstream yet, but if there are advantages besides faster and more accurate focus, someone will exploit it.

I also disagree with those who say that it's a joke. Maybe they see no benefit for their limited use, and based on the minimal effect it has on refocusing a 35 mm (!) Wide angle lens, but that doesn't mean it cannot be useful for others.

Once the data is captured, it's up to the software engineers to exploit what's possible with that data. Many people are unaware of a booming development in R&D, called Computational Photography. We are now only seeing the first practical steps of what has been in the making for many years and what will additionally be possible several years from now.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3369
    • http://www.paulholman.com
Re: Canon 5D Mk4 announced
« Reply #43 on: August 31, 2016, 09:49:59 am »

Why not?
In the real world the software companies won't bother going to expensive lengths to support a minor feature on a single camera model, it would make no economic sense.

Not even Canon have bothered with their own software.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Canon 5D Mk4 announced
« Reply #44 on: August 31, 2016, 09:56:07 am »

In the real world the software companies won't bother going to expensive lengths to support a minor feature on a single camera model, it would make no economic sense.

Not even Canon have bothered with their own software.

Are we talking about the same thing???

Digital Photo Professional, Canon's free Raw converter utilizes the dual pixel technology in post-processing or those Raws that can also record the data. Who's to say other companies won't adopt similar (patents may apply) technology?

You say Never, I say never say Never.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #45 on: August 31, 2016, 09:57:30 am »

In the real world the software companies won't bother going to expensive lengths to support a minor feature on a single camera model, it would make no economic sense.

Not even Canon have bothered with their own software.

Not all is lost as of yet!  ;D
http://www.cnet.com/news/adobe-to-support-dual-pixel-raw-photo-format-in-canon-5d-mark-iv/
Logged

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3369
    • http://www.paulholman.com
Re: Canon 5D Mk4 announced
« Reply #46 on: August 31, 2016, 10:03:35 am »

Yes, the 'format' may be 'supported' but they don't say that they're supporting potential the extra stop of DR. They only mention the focus/Brokeh/flare issues.

There's no mention of extracting an extra stop of DR from the files. Given the ludicrous amount importance this aspect seems to command at the moment, I would have expected Canon to shout about it if it was available.
Logged

Zorki5

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 486
    • AOLib
Re: Canon 5D Mk4 announced
« Reply #47 on: August 31, 2016, 10:13:04 am »

Not all is lost as of yet!  ;D
http://www.cnet.com/news/adobe-to-support-dual-pixel-raw-photo-format-in-canon-5d-mark-iv/

Iliah, quick ( :P ) question: can this dual-pixel format be represented as a DNG?

In other words, imagine Canon had adopted DNG as their format some time ago. Would there be way to represent this data in a DNG in a meaningful way, so that not only use DNG as a container, but specify the "meaning" of the extra data? So that converters would "know" what to do with it?
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Canon 5D Mk4 announced
« Reply #48 on: August 31, 2016, 10:14:39 am »

Not all is lost as of yet!  ;D
http://www.cnet.com/news/adobe-to-support-dual-pixel-raw-photo-format-in-canon-5d-mark-iv/
No, it's not lost in our real world (the one where speculation and FUD is not accepted so easily)!  ;D
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #49 on: August 31, 2016, 10:41:09 am »

> can this dual-pixel format be represented as a DNG?

Same way it is with Fuji S5 and other dual-pixel Fuji cameras.

> imagine Canon had adopted DNG as their format

Well, that's somewhat hard to imagine.

> Would there be way to represent this data in a DNG in a meaningful way, so that not only use DNG as a container, but specify the "meaning" of the extra data?  So that converters would "know" what to do with it?


That would mean publishing the document that specifies what are the interpretations for Canon undocumented Makernote tags  ::)
Logged

Zorki5

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 486
    • AOLib
Re: Canon 5D Mk4 announced
« Reply #50 on: August 31, 2016, 10:57:24 am »

> can this dual-pixel format be represented as a DNG?

Same way it is with Fuji S5 and other dual-pixel Fuji cameras.

> imagine Canon had adopted DNG as their format

Well, that's somewhat hard to imagine.

> Would there be way to represent this data in a DNG in a meaningful way, so that not only use DNG as a container, but specify the "meaning" of the extra data?  So that converters would "know" what to do with it?


That would mean publishing the document that specifies what are the interpretations for Canon undocumented Makernote tags  ::)

OK... There are indeed questions that can't be answered with simple "Yes" or "No" (like, say, "Did you quit drinking cognac every morning?"), but I'll still take the liberty of interpreting your answer as a "No".

You see, every now and then there's outcry on these forums: "Just adopt DNG already! The only reason manufacturers don't do that is the sizes of their egos!", whereas my answer had always been it's far more complicated than that. Semantics/interpretation of new data (e.g. this dual-pixel layout) is one complication, competitive advantage (e.g. Sony's compression, which I personally consider brilliant) is another, and so on and so forth...
Logged

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #51 on: August 31, 2016, 10:59:34 am »

> There's no mention of extracting an extra stop of DR from the files.

Why, to get answers one needs to meet the right people, have the right reputation, and ask the right questions.

> I would have expected Canon to shout about it if it was available.

Given the level of comprehension, they should not, or they will be asked why they trimmed it before  ;D
Logged

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #52 on: August 31, 2016, 11:02:23 am »

> "Did you quit drinking cognac every morning?"

That's an easy one. Replace cognac with a single malt, and the answer is "No". I have it in my morning coffee.

> You see, every now and then there's outcry on these forums: "Just adopt DNG already!

Right. But just adopting DNG does not solve much. Opening the format solves more. Question of meta-metadata ;)
« Last Edit: August 31, 2016, 11:09:34 am by Iliah »
Logged

scyth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Canon 5D Mk4 announced
« Reply #53 on: August 31, 2016, 12:52:51 pm »

> can this dual-pixel format be represented as a DNG?
Same way it is with Fuji S5 and other dual-pixel Fuji cameras.

except it is still different (given the possible usage for "focus adjustment" - you can't do this with Fuji data), so while you can purely technically physically write it it is not exactly documented (in DNG standard that is) for software to interpret it and the whole point is in that.
Logged

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #54 on: August 31, 2016, 12:57:41 pm »

except it is still different (given the possible usage for "focus adjustment" - you can't do this with Fuji data), so while you can purely technically physically write it it is not exactly documented for software to interpret it and the whole point is in that.
I was a bit sarcastic. Parallax between those 2 sub-pixels, by the way, is not enough for meaningful re-focusing. It is useful, however, for selective sharpening and blurring in conversion. Fuji indeed is a different case from this point of view, literally.
Logged

scyth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Canon 5D Mk4 announced
« Reply #55 on: August 31, 2016, 12:58:19 pm »

Digital Photo Professional, Canon's free Raw converter utilizes the dual pixel technology in post-processing

but (unless I am grossly mistaken) DPP does not use half-sensel data for clipped /in combined frame/ data "restoration", does it ?
Logged

scyth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Canon 5D Mk4 announced
« Reply #56 on: August 31, 2016, 01:02:56 pm »

It is useful, however, for selective sharpening and blurring in conversion.
certainly if a pair of subsensels tells that their sensel belongs to the area which is sufficiently within DOF you might want to apply sharpening proportionally based on how much it is defocused (estimated)... if a pair of subsensels tells that their sensel belongs to the area which is sufficiently OOF you might want to blur that instead, do heavy denoising (and not sharpen at all) for a better background rendering... and attempts at LoCA correction as well can be made differently based on that
Logged

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #57 on: August 31, 2016, 01:04:56 pm »

certainly if a pair of subsensels tells that their sensel belongs to the area which is sufficiently within DOF you might want to apply sharpening proportionally based on how much it is defocused (estimated)... if a pair of subsensels tells that their sensel belongs to the area which is sufficiently OOF you might want to blur that instead, do heavy denoising (and not sharpen at all) for a better background rendering... and attempts at LoCA correction as well can be made differently based on that

True. 5D4 gives a taste of the things to come with more pixel density in regular-sized sensors.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Canon 5D Mk4 announced
« Reply #58 on: August 31, 2016, 02:52:16 pm »

but (unless I am grossly mistaken) DPP does not use half-sensel data for clipped /in combined frame/ data "restoration", does it ?

No, not as a separate function, but maybe they do something with it in regular exposure correction? DPP has never been a good tool for highlight recovery, maybe they assume that a photographer knows how to correctly expose an image?

There is a small issue, and that is that the dual pixels are sampling at a slightly (half a pixel) different position so strictly the signals do not share the same input signal, which they do after binning for the main Raw plane. How that would work out when one wants to use it for highlight recovery remains to be seen. It was not designed for that purpose.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Iliah

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 770
Re: Canon 5D Mk4 announced
« Reply #59 on: August 31, 2016, 04:02:26 pm »

> There is a small issue, and that is that the dual pixels are sampling at a slightly (half a pixel) different position so strictly the signals do not share the same input signal, which they do after binning for the main Raw plane. How that would work out when one wants to use it for highlight recovery remains to be seen.

Consider this raw, dual-pixel, ISO 400:
http://www.kamerabild.se/sites/kamerabild.se/files/_91A0045.CR2

The main frame:


The auxiliary frame:


IMHO well worth working to make that sky rendered not as a blob ;)
Logged
Pages: 1 2 [3] 4 5 6   Go Up