Some thoughts on linear and non-linear RAW images

Discussion in 'Photography' started by Malte, Sep 6, 2004.

  1. Malte

    Malte Guest

    Hi guys,

    I have a bit of trouble getting my head around linear and non-linear
    images being generated by Canon's RAW conversion utilities.

    Canon provides a "TIFF 16bit/channel linear" mode in their RAW
    conversion software. Is that linear image data what is produced by the
    image sensor in the camera? Out of curiosity I converted a RAW image
    to "TIFF 16bit/channel linear" mode and opened it up in Photoshop. It
    looks really dark. I increased the gamma until it looked pretty much
    like the same image in "TIFF 16bit/channel" (non-linear) mode.
    Surprise, surprise... the gamma is about 2.5. I deduce from that that
    - the RAW conversion software applies a gamma of 2.5 to every image by
    default in non-linear mode (likely), or
    - the nature of the image sensor is such that it applies a gamma of
    2.5 and the RAW conversion software applies the inverse gamma to
    produce the linear mode image (less likely).

    If the first case is true, doesn't that mean that my great 16bit RAW
    image is already degraded by the gamma application? Doesn't that mean
    that certain values in the histogram do simply not occur anymore
    because they have been "gammaed" together with brightness values in
    the neighbourhood?

    One thing may be worth mentioning in this context: When I opened the
    image in Photoshop I applied the Adobe RGB (1998) profile to it (I
    _applied_ it, I did not convert the image to it) since that is the
    color space I usually shoot in and I have my EOS 10D configured as
    such. However, since color spaces describe the "meaning" of "color
    numbers", so to speak, that color space can only be the correct color
    space for either the linear _or_ the non-linear image. I strongly
    suspect it is correct for the non-linear image since it works fine on
    JPEGs as well and JPEGs look pretty similar to the TIFFs produced in
    non-linear mode.

    I may get a bit disillusioned here about RAW images that I always
    thought are "a direct image of what the image sensor sees".

    To complicate the matter even more: Adobe RGB (I always thought,
    please correct me if I am mistaken) as a working color space is
    linear. Hence, every image I have on my screen in Photoshop is in a
    linear color space. By the use of my monitor profile an inverse 2.5
    gamma (this is all roughly of course, color profiles contain much more
    information than that) is applied to the image _only for display_. My
    monitor then, by its nature applies a gamma of 2.5 to the image when
    displaying the image, hence the inverse gamma applied by the color
    management system for display purposes. That is why images in
    Photoshop just look right on the screen. However, when I _apply_ a
    linear color space such as Adobe RGB (1998) to an image, I basically
    tell Photoshop that the image data is linear. Hence, if I apply Adobe
    RGB (1998) to the linear RAW image it should just look right (but as
    previously mentioned it looks far too dark). If I apply Adobe RGB
    (1998) to non-linear RAW image it should look somehow odd depending on
    the (unknown) nature of the non-linearity (but it just looks fine).

    It almost seems that the linear image is not linear at all but has a
    gamma less than 1 (the reciprocal of 2.5) applied to it. But that is
    probably not the solution, or is it?

    Sorry for so many words. I thought I better explain the issue in
    detail. I am curious what others think about all this. Maybe someone
    knows the answer to all this.

    Thank you anyway for listening.
    Malte, Sep 6, 2004
    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.