Which Raw Format?

Started Oct 6, 2012 | Discussions thread
Barry Pearson
Veteran MemberPosts: 5,594Gear list
Like?
Re: Which Raw Format?
In reply to alanr0, Oct 15, 2012

alanr0 wrote:

PhotoHawk wrote:

I still don't understand your position awaldram.  At worst DNG is no different to any other raw file format.  At best it has features that may be quite useful.  Originally we were not talking about JPEG or TIFF.  A DNG is really just a more defined TIFF.  And while we are discussing TIFF please note that TIFF is an Adobe owned format.

I think awaldram's objections are pragmatic - in that some software vendors offer only partial DNG support, combined with reservation's about Adobe's control of the specification.

In effect, Adobe now has less control of DNG than Pentax has with PEF, Nikon with NEF, Canon with CR2, etc. That is what happens when you make things so public and also make it available to ISO for standardisation.

My understanding from Barry's posts (and your previous post) is that DNG mandates sufficient information to render the image is included in the metadata.  It does not follow that software vendors will necessarily use this information.

I stopped checking this a couple of years ago, but my tests until then showed that a lot of software products  must have used this data, because the raws I tested them with were from cameras released after the software, and the software didn't support the native raws.

What I also found was that software that didn't do it properly tended to get updated so that it could, or cease to exist. In fact, the myth that software typically didn't do it right came largely from 2 products: Apple Aperture before 2.0 and Pixmantec Rawshooter. (I covered these cases below the above link).

My guess (open for correction if Andrew, Barry or anyone else has solid information) is that some vendors choose to extract the proprietary MakerNote data, rather than use the 'official' DNG fields.  If they already support proprietary raw for specific camera models, they can just pull in the data they already know how to handle.  Often this relies on model-specific information which is not included in the MakerNotes, so the approach fails when handling output from an unfamiliar camera.

If I understand you correctly, Bibble was the classic example of this to the end. They appeared to have an ideological block against supporting DNG correctly. But the trend has been towards supporting the DNG metadata, as with Aperture.

The DNG Converter stores the Exif Makernote unchanged in the DNGPrivateData tag in a documented way.

Where it gets more interesting is when someone (often Fuji) comes up with a different (non-Bayer) sensor layout.  Presumably DNG has a means to describe the layout.  Is there a default (perhaps highly inefficient) demosaic algorithm, or does a new a layout require custom software?

Such a radically new sensor layout requires extra DNG parameters. They happen rarely! Most camera models are refinements of existing principles, or changes that are nothing to do with raw conversion. Cameras with such "disruptive" sensor layouts result in lots of delays before third-party software products catch up with them; some never do catch up, and those cameras get reduced support. DNG certainly doesn't make the problem worse, and typically makes it better.

The "archival DNG" argument is that sufficient information is made available to render the image.

To many folk this is moot, unless software is available which can interpret and render the specific image you have in hand.  Personally, I like the idea that the information is present in a documented format - even if I have to write my own software to render the image.

I think a problem that some people have with the "archival DNG" concept is that they don't realise that with other formats the software developer not only needs to reverse-engineer the raw file format in the absence of specifications, (which is a skill that several people have mastered, so it isn't such a big handicap), but they need to test a camera to find how its numbers map onto real-world colours plus lots of other stuff.

Will cameras available in 2012, generating their PEFs and NEFs, etc, still be available and working in 2022 or 2032 so that people developing software in future can test these 2012 cameras and build in conversion for those 2012 PEFs and NEFs? Will there be a market case to do so? DNG is a "catch-all" format: instead of bulding in data (and often specialised code) for 100s of camera models, you build in code for each DNG specification, 5 of them in the 8 years so far.

 Barry Pearson's gear list:Barry Pearson's gear list
Pentax K-7 Panasonic Lumix DMC-GF1 Pentax K-5 Pentax Q Pentax K-5 IIs +22 more
Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow