1 person found this helpful
1. First of all, the term "linear" as contrasted to "mosaic" is total misnomer, for the "mosaic" is linear as well.
2. The "linear" format is not raw any more, while the "mosaic" format is as raw as it is possible with the given camera (which is sometimes far from being raw).
3. The "mosaic" DNG format is a straightforward format conversion of the native raw data, the "linear" is partly processed. The "mosaic" format image can always be converted in "linear", but not the other way.
In cleartext: the "linear" data converted from the native raw file is far from the original.
4. There are devices, which generate non-mosaiced format, like scanners.
5. The "linear" format is created by some raw processors as alternative output to TIFF and JPEG. For example if you want to use DxO, for you like their camera and lens specific corrections, you can perform a "pre-conversion", and process the result (a "linear" DNG) in another raw processor, like ACR or Aperture.
Facit: there is no reason whatsoever to convert your native raw files in linear DNG. In fact, there are not many reasons to convert the native raw in DNG at all. The only sound reason is, that your camera is not supported by the version of ACR you have.
Thank you very much for your answer! I wasn't very clear: actually, I want to know if there is one format which more "future proof" between DNG linear and DNG mosaic. Or are they both just as likely to be compatible with future programms as JPEG for example?
As both formats are variations of the same standard, DNG, the question is moot, which one will live longer.
Following is only my personal opinion: there will always be programs around, which support the *raw* format; for me it is questionable if the same is true regrading the linear (derivative) format.
In fact, it would be possible *with specialized programs* to recreate the native raw files from DNGs, with some exceptions. However, it is *impossible* to create Bayer-type raw files from "linear" data.
Since I got an EOS 400D, I don't think it is threatened in the near future, but all this different RAW formats are less likely to be readable in say, 20 years than DNG or JPEG... Too much RAW formats.
FYI, although the "mosaic" vs "linear" terminology is not perfect, please note that the mosaic data in many cameras is not linear. In most cases it is reasonably straightforward to derive a linear signal from the mosaic data, but the actual original mosaic data certainly does not respect f (x) = k * x for some constant k.
So, there are many different "DNG mosaic" formats, depending on which RAW is has been imported from? This is bad news, it means it is no more future proof than RAW...
So, there are many different "DNG mosaic" formats, depending on which RAW is
has been imported from? This is bad news, it means it is no more future proof
These aren't differences in the DNG / TIFF-EP mosaic format. They are merely
differences in how camera makers choose to store their mosaic data. There
are different choices the designers make with regards to details such as
encoding, black subtraction, stitching, etc.
1 person found this helpful
I think you are overly concerned about multiple DNG formats. If you really want to know this level of detail, then the DNG spec is available. I don't really understand all of the conspiracy theories everyone wants to develop around the DNG format.
The answer is both yes and no: the metadata in a DNG file includes information on the layout of the Bayer pattern for the RAW data. The original RAW files seldom include this, but the DNG converter or writer provides the details on how to interpret the sensor data as long as it's fairly a fairly standard Bayer pattern sensor. Most major DSLRs use sensor technology that is very similar as far as how the RAW data is laid out and what it represents. Often camera makers get creative in how this is stored (I'm sure there's a reason CR2 is so whacked out, but it is what it is), or encrypt relevant information as Nikon does, but the technology is somewhat consistent when viewed from a distance.
Every now and then some researcher proposes a non-Bayer layout, and there is a lot of buzz around it, but I haven't seen the big players moving to adopt any of these. I'm sure that if there was some crazy layout that became mainstream the info in the DNG metadata could be further generalized to describe its layout, but the version of DNG might need to be bumped in this (unlikely) case.
Nice! Thank you, you answered all my questions about DNG. Actually, I noted that DNG is like 10 to 20% smaller than the RAW, so if the transformation is lossless and it has better legs in the future, why wouldn't I convert my RAWs in DNG before converting them? =)
Yes, I know the reasons for the raw data as it is stored not being linear. However, black level offsetting and non-linear lossy compression are side issues, the pixel values can and will be transformed immediately in as far linear format, as they are in the "linear" DNG file.
Just to clarify, I was responding to the statement "many different DNG mosaic formats," and I was describing the information in a RAW DNG (i.e., not linear) file. The process for describing the sensor layout in the metadata in the file was for converting RAW to RAW DNG.
I don't think you can call converting RAW to linear DNG an archival process.
I don't think you can call converting RAW to linear DNG an archival process
I don't think you can call converting the native raw file in DNG an archivation process, not matter in which DNG format, except if you embed the entire original raw file in the DNG file.