method to reveal photoshop manipulations

Discussion in 'Digital Photography' started by sobriquet, Sep 25, 2007.

  1. sobriquet

    bugbear Guest

    In general, no. Given sufficient care,
    it is (clearly?) possible to perform a
    perfect edit (or synthesis).

    bugbear, Sep 26, 2007
    1. Advertisements

  2. sobriquet

    JL Guest

    I didn't read the following messages when I posted mine.
    Here is the areas of the manipulated photo which are suspect to me:
    The ear is the more suspect because it has a darker edge although the
    backgorund is light (car).

    As I'm doing such manipulations myself (see, I
    know where to look. The transition between contrasted areas is a problem
    difficult to solve when doing such manipulations.

    Jean-Luc Ernst
    JL, Sep 26, 2007
    1. Advertisements

  3. sobriquet

    sobriquet Guest

    Yeah... sort of like how many governments and corporations steal or
    violate human rights. Ideas, thoughts, numbers, letters, pixels, bits
    and even your dreams can be stolen.
    Intellectual property, where it violates basic human rights like being
    able to exchange information freely, is a crime against humanity.
    sobriquet, Sep 26, 2007
  4. sobriquet

    sobriquet Guest

    But couldn't the dark edge be the result from sharpening in the
    Just like the dark line at the left side of the blue plants of the
    cyclist and the road that is part of the original picture.
    If you look at the other side of the brown jacket, where the van has
    been edited out, there seems to be a dark edge missing which is
    visible between the jacket and the road behind it.
    But if you look at the pants on the right side, there is no line
    between it and the road, but a smooth transition, so I guess it's all
    dependent on the way the light strikes the subject and makes it blend
    in with or stand out against its background.
    I thought about subtracting one image from the other and using the
    difference somehow as a layermask to get a kind of generated boundary
    instead of a manually painted one. Perhaps after this, an adjustment
    layer can be applied to ensure the edge looks natural like other
    transitions between elements in the picture against their background.
    I reckon that using such 'automatic' techniques as much as possible is
    a better method compared to manual manipulations of painting edge
    sobriquet, Sep 26, 2007
  5. sobriquet

    Eric Miller Guest

    That is a truism and it is logical. If a=b and b=c then a=c. The problem
    with stating that truism here is that it has nothing to do with the
    conclusion that you intend it support. The question is not whether the
    cyclist was part of both images; the question is which of the two images was
    Ah, finally, the logical fallacy that is hanging you up here: begging the
    question. This begs the question about what has changed between the images,
    which is the ultimate question. I take your word that the changes are as you
    describe, but logic does not dictate that result. The cyclist was in the
    position portrayed in both your images at one time and an image was
    recorded. That cyclist, in that position, could have been added to either
    photo. That modification could have been done to make it appear that the
    cyclist had not moved when, in fact, he had. In fact, you could have
    modified the second image to keep the van and cyclist in the same position.
    That fact partly addresses your question regarding the ability to
    objectively determine if changes have been made. If a person could not
    modify the exif data, but could modify the images, he could modify a
    sequence of images taken minutes apart to appear as if nothing happened
    when, in fact, a bank robbery (or any other event) took place.
    Infallible arguments are few and far between, even when they support a
    proposition that is later established as fact.
    That would be a silly assumption. But many, if not most, assumptions are

    But that is a different question than the one that you originally posed and
    resolves, as a given, that the cyclist has not been moved.

    Eric Miller
    Eric Miller, Sep 26, 2007
  6. Read up on masks and binary operations with masks. That should show
    differences in areas between the two (especially a binary subtraction).
    Don Stauffer in Minnesota, Sep 26, 2007
  7. Sometimes the difficult part is impeccably posing the question you want
    answered. F'rinstance, I thought the OP was looking for a
    technical--i.e., software or pixel-by-pivel examination, etc.--means to
    differentiate a modified from an unmodified image). Apparently I read
    something different into his word "objectively" while it seems you
    zeroed in on a different element/elements in the two images from those
    which jumped out at me.
    sheepdog 2007, Sep 26, 2007
  8. sobriquet

    Eric Miller Guest

    That is what I thought too.
    I guess I thought he might have been a bit more clever (or tricky) than he

    Eric Miller
    Eric Miller, Sep 26, 2007
  9. sobriquet

    sobriquet Guest

    The question which of the two images was manipulated translates to the
    question on which side of the cyclist the white van was in the
    original picture. The white van (and it's associated shadows,
    reflections and headlights shining on the road) is the only element in
    the pictures that is relevant to the question which of the two
    pictures has been manipulated and the cyclist isn't (because the
    cyclist is part of the original image).
    No. You can subtract x.jpg from y.jpg in photoshop and that will show
    you exactly what has been changed. But it doesn't answer which of the
    two pictures is the original and which is the modified picture.
    But to answer the question which of the pictures has been modified,
    you must examine the changing elements and ignore the identical
    elements in both pictures.
    No, because adding something that is already there doesn't change
    anything. The cyclist is part of the original picture, adding it once
    more to the modified picture doesn't change anything about that fact.
    It would be impossible for the cyclist to move without changing. If
    the cyclist had moved and was pasted in to make it seem it hadn't
    moved, you could still use subtraction and levels in photoshop to show
    the differences between the moved cyclist and the cyclist in the
    original position.
    But the exif data and the content of the image can both be edited
    independently. I might as well have copied the exif info of the
    original picture to the modified picture instead of stripping the exif
    data from both the original and the modified picture.
    So basically, the exif data has nothing to do with the question of
    whether or not the actual content of the image has been manipulated.
    Well ok, infallible is perhaps too much to ask for, but a plausible
    argument maybe. :)
    No, because an answer to that question is exactly an answer to the
    question which of the two picture has been modified.
    That the cyclist hasn't been moved or modified can be seen visually.
    Just subtract x.jpg from y.jpg in photoshop and you will find that the
    cyclist is completely black and even the levels command will not be
    able to reveal any differences between the cyclist in both pictures,
    so that demonstrates it's exactly identical and hence part of the
    original picture, regardless of whether it has been added again later
    on which wouldn't have changed anything.
    It's a bit like having a picture filled with black and filling the
    picture with black once again and asking whether or not the picture
    has been modified. Since it is still identical to the first picture,
    it has not been modified.
    sobriquet, Sep 26, 2007
  10. sobriquet

    Eric Miller Guest

    That translation assumes that the location of the cyclist is not a
    manipulation of one of the pictures. You are therefore starting with an
    answer (that the cyclist is not a manipulation) and modifying your question
    to fit it. Your original question did not offer the condition that the
    appearance of the cyclist has not been changed.

    But, alas, I see that this is a brain problem and I will trouble you no

    Eric Miller
    Eric Miller, Sep 26, 2007
  11. It will reveal differences, but not which is original. If I have any
    image file, edit it in Photoshop, convert that to a Tiff or other
    straight color bitmap, then convert that to the original format, there
    will be nothing to reveal. The new file will be based on the modified
    image without any modifications to reveal.

    An example. If the original image was entirely a montage of many images
    and the modified file was the montage with all shadows made to line up
    the same direction, most analysis would pick the wrong file as original.
    If it was laundered by converting to a bit map and then back to the
    original format, a digital analysis would which was the original.
    Stephen Henning, Sep 26, 2007
  12. sobriquet

    sobriquet Guest

    That the cyclist (including his location) is not a manipulation of
    either one of the pictures folows from the fact that it is identical
    in both pictures and hence part of the original picture.
    The question was which of two pictures had been manipulated, given
    that one of the two pictures has not been manipulated.
    Yes, it did. Anyone with basic image manipulation abilities can verify
    that the cyclist as well as most of the context in which the cyclist
    appears is identical in both pictures.

    I'll show you the difference between the two pictures x.jpg and y.jpg:

    With the levels command applied to it:
    identical in both pictures and the cyclist is part of the black area
    (except at the boundaries).

    As a comparison, I'll also show you the difference between the
    pictures 2380 and 2381 (referred to elsewhere in this thread) used to
    create x.jpg (and y.jpg without any modifications apart form saving it
    as jpg and stripping exif info):

    And again with the levels command applied to it:

    Which shows there are subtle differences between the two pictures
    throughout the pictures and hence the pictures do not contain
    identical parts.
    I'm sure we can understand each other and come to some kind of
    agreement or at least recognize on which points our opinions differ
    exactly. I'm patient and I don't mind explaining myself carefully as
    long as I have the impression people are openminded and willing to
    sobriquet, Sep 26, 2007
  13. Absolutely correct. I have read accounts of "forensic " or
    photointerpretation" experts who state definitively that this or that
    digital image is a fake, and I maintain that there are at least three
    kinds of edited images~

    1. ham-fisted amateur composites, where anything from a conflict in
    shadows or other lighting effects to jarring fringe pixels are
    giveaways to even a casual observer

    2. pretty good fakes, where it takes someone experienced in working on
    + studying many images to detect the fakery

    3. excellent work, which will fool even an expert
    I have been following this thread with interest to see if anyone has an
    algorithm or other software with a high rate of success in detecting
    edits. Trying to deduce the fake from gut or logic isn't in the same
    sheepdog 2007, Sep 26, 2007
  14. sobriquet

    Frank ess Guest

    I'd already guessed you're a freak. This confirms it.

    Frank ess, Sep 27, 2007
  15. sobriquet

    Pat Guest

    Yes, it is possible to detect most fakes, but it isn't done 'by eye'.
    That's pretty unreliable. I think I read about it in Rangefinder or
    some other magazine but it could have been the NYT. Anyway, like many
    areas of photography, it isn't the photographers who are leading the
    way. In this case, it's the biologist. The Journal of Cell Biology
    subjects all photos to tests to determine if any are fakes.

    I am doing this from memory, but here we go.

    First off, if you remove something you need to replace it with
    something. Often that is from someplace else in the picture. They
    can detect such a duplicate pattern.

    Another thing is to check for enlargements, reductions, and
    rotations. They cause new patterns in the image as new information is
    added or removed.

    Shadow orientation can also be checked. If one is off by a couple of
    degrees, that can be detected.

    I think the program also looks for "hard patterns" like a line cut
    from an image Precise lines are more likely to be fakes than
    feathered ones.

    If you want more info, you might want to check and see if they provide
    any additional info.
    Pat, Sep 27, 2007
  16. sobriquet

    sobriquet Guest

    Mr. america, walk on by your schools that do not teach
    Mr. america, walk on by the minds that wont be reached
    Mr. america try to hide the emptiness thats you inside
    But once you find that the way you lied
    And all the corny tricks you tried
    Will not forestall the rising tide of hungry freaks daddy!

    They wont go on four no more
    Great mid-western hardware store
    Philosophy that turns away
    The left behinds of the great society

    Hungry freaks, daddy!

    Mr. america, walk on by your supermarket dream
    Mr. america, walk on by the liquor store supreme
    Mr. america try to hide the product of your savage pride
    The useful minds that it denied
    The day you shrugged and stepped aside
    You saw their clothes, and then you cried,
    Those hungry freaks, daddy!

    They wont go on four no more
    Great mid-western hardware store
    Philosophy that turns away
    The left behinds of the great society
    sobriquet, Sep 27, 2007
  17. Necessity or no, it's not often you see The Mothers quoted on _any_ NG.
    sheepdog 2007, Sep 27, 2007
  18. sobriquet

    Paul Furman Guest

    No, you can't tell from a standard formula, only from following logic.
    Any formula will just tell you the differences, not which is the original.

    What I saw is the cyclist's ear has a dark outline on the white bus
    which looks unnatural. You could fool someone by creating fake effects
    like this though so it's no guarantee.
    Paul Furman, Sep 28, 2007
    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.