Raw conversion ..

Discussion in 'Digital SLR' started by imbsysop, Sep 28, 2009.

  1. imbsysop

    imbsysop Guest

    I have been wondering for some time..

    Is there any info to what type of picture "format" raw is converted in its
    intermediate state ie before the user makes a decision to save it under the
    form of a known picture format? (jpg, tiff etc)
    What I mean is .. raw => converted(=demosaic)=> picture in PP program
    "memory" .. so what picture "format" is this picture in?
    imbsysop, Sep 28, 2009
    1. Advertisements

  2. imbsysop

    Ofnuts Guest

    At that point there is no picture "format" since the format is an
    encoding used to store the picture in a file.

    It's quite likely that each program does it differently and has its own
    optimizations to manage efficiently a fair amount of storage (especially
    when it has to support a long undo history).

    The Gimp and its RAW-handling extensions are open source software, so
    you can have a look :)
    Ofnuts, Sep 28, 2009
    1. Advertisements

  3. Why?

    Wolfgang Weisselberg, Sep 29, 2009

  4. Yes, TIFF is tops if you need layers, but keeping the RAW file is the
    most complete archival source of a digital image. And it's way smaller
    than a TIFF.
    John McWilliams, Sep 29, 2009
  5. imbsysop

    Guest Guest

    the photoshop format itself does all that and more.

    plus, many apps these days use a non-destructive workflow directly on
    the raw file so there isn't a tif or psd to save.
    Guest, Sep 29, 2009

  6. TIFF is tops if you need layers, or have already created one for
    whatever reason, but keeping the RAW file is the most complete archival
    source of a digital image, and it's way smaller than a TIFF, even a
    compressed one.
    John McWilliams, Sep 29, 2009
  7. imbsysop

    Guest Guest

    in fact, adobe provides profiles to match that of the manufacturer's
    Guest, Sep 29, 2009
  8. imbsysop

    Eric Stevens Guest

    But a RAW format does not have and cannot preserve layers.

    Eric Stevens
    Eric Stevens, Sep 29, 2009
  9. Which is one reason I wrote what I did. There's not even an inference
    that RAW can have layers.
    John McWilliams, Sep 29, 2009
  10. imbsysop

    imbsysop Guest

    OK this is more or less what I expect but there are a couple of conventions
    that need to be fullfilled ... ie de demosaic of the raw needs to generate
    "something" that is compatible with the normal representation standard in
    de computer hardware ergo IMHO there is at least some need for an RGB pixel
    representation in memory. I do not think that it is mapped to video memory
    for the simple reason that this is done by the video driver and at this
    level the driver needs some prechewed data that it can understand ...
    imbsysop, Sep 29, 2009
  11. imbsysop

    imbsysop Guest

    Thank you very much that sheds quite some light on the "raw" picture data
    "as seen" !
    imbsysop, Sep 29, 2009
  12. imbsysop

    imbsysop Guest

    yep indeed ... are you sure about the BMP format? This seems the most
    restrictive in terms of color tonalities and format flexibility ?
    Well I think the 12/14 bits raw matter is not relevant at this level
    anymore in so far that the demosaic/conversion has been done before it is
    even possible to "see" anything that looks like a real picture .. but is
    the conversion outcome a 16 bit format aka 16bit x3 color depth? I guess

    So does it make sense to suppose that initially this intermediate format
    is "proprietary" to the used program and highest resolution, color depth?

    Thanks everyone for your input. The Q was raised in some other (foreign)
    forum where people claimed they applied noise reduction to the raw
    file/data ..
    imbsysop, Sep 29, 2009
  13. imbsysop

    Eric Stevens Guest

    TIFFs have somehow got into the conversation. I was just making an
    essential difference clear for those who might think its a toss-up as
    to whether to use TIFF of RAW.

    Eric Stevens
    Eric Stevens, Sep 30, 2009
  14. imbsysop

    OldBoy Guest

    (To me this is a foreign forum :) )
    DxO (www.dxo.com) claims to apply noise reduction to the raw data.
    OldBoy, Sep 30, 2009
  15. imbsysop

    Guest Guest

    although it's true that lightroom and similar apps can't do everything
    and other apps might need to be used, you still don't need tiff. it can
    round-trip to photoshop in either tiff or psd, and in many cases, no
    intermediate file is needed at all.
    Guest, Sep 30, 2009
  16. imbsysop

    imbsysop Guest

    Thanks .. I was indeed thinking of having a peek at dcraw .. :) (but being
    lazy in the first place I though I'd give asking a go :))
    imbsysop, Sep 30, 2009
  17. imbsysop

    imbsysop Guest

    That is a statement that goes a long way .. ever seen/read how many
    people claim they do all their PP adjustments on their "raw" files? :-D
    imbsysop, Sep 30, 2009
  18. imbsysop

    imbsysop Guest

    Yeah .. and given the intrinsic nature of a "raw" file I for one would very
    much like to read some extended tech procedure how they may be doing such
    thing .. :-D
    imbsysop, Sep 30, 2009
  19. imbsysop

    Eric Stevens Guest

    Depends what you mean by 'PP' but many Nikon NX2 users might claim
    they do.

    Eric Stevens
    Eric Stevens, Sep 30, 2009
  20. "Re-mosaicing" is, for all intents and purposes, impossible. For the sake
    of utter simplicity (because demosaicing the values from a Bayer sensor
    depends on the values of many more than just four adjoining photosites) if
    you have the values of 9, 13, 5, and 8 for the RGGB photosites in one 2x2
    photosite section, and then demosaic it into an RGB value of 11, 9, 10
    (chosen algorithm dependent), how are you going to reconstruct the original
    four values from just those three? In the sake of true demosaicing
    algorithms; which use a large area of photosites (many more than four) to
    determine the final RGB values of each "pixel"; in 10, 12, or 14 bit
    depths; there's an astronomically high number of permutations when going
    from RGB values back to the RGGB values. Even if you choose just one of
    those sets of re-mosaiced values from one pixel, how do you determine they
    are the correct values without the original reference source, let alone
    choosing the correct values for all of them.

    "Re-mosaicing" is impossible. A good example of the problems and errors of
    extrapolation taken to the extreme.
    Bayer Aspririn, Oct 1, 2009
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.