Photography and forensic science

Discussion in 'Digital SLR' started by Charles, Feb 7, 2010.

  1. Charles

    John A. Guest

    Shoot a few dark frames at the exposure/ISO/etc. of the target image.
    John A., Feb 12, 2010
    1. Advertisements

  2. Charles

    John A. Guest

    See also:
    John A., Feb 12, 2010
    1. Advertisements

  3. Given that the author doesn't know the difference between a
    grand jury and a petit jury, and thought that Perry Mason
    "always got a conviction," I can only wonder at the quality
    of the rest of the research.
    Michael Benveniste, Feb 12, 2010
  4. That only shows the residual hot pixels. Thousands of hot and dead
    pixels are present on every camera sensor and are mapped out at the
    factory so you never see them, but their effects are there, hidden in
    every image the camera takes, just like a fingerprint. You can read the
    factory defect map out of the camera with the correct manufacturer's
    disassembly tools but, as I asked, how do you know where they are
    without disassembly?
    Kennedy McEwen, Feb 12, 2010
  5. Charles

    Charles Guest

    Thanks SD. That added a lot to the discussion.
    Charles, Feb 12, 2010
  6. Charles

    Charles Guest

    Once again, thanks. I have learned some interesting things here and am
    delighted that this thread has been constructive and has remained civil.
    There is still hope for Usenet!
    Charles, Feb 12, 2010
  7. Charles

    Peter Guest

    A good judge will weigh evidentiary value vs. prejudicial potential.
    Unfortunately, sometimes judges do bow to political influences.
    Peter, Feb 13, 2010
  8. Not so. Whether it's a professional photographer or a cell-phone
    snap, the same rule applies. All you have to do is call the third
    party and lay the foundation that I outlined earlier. Jurors will
    often give more credibility to the bystander's photograph than one
    taken by the police, because they assume bystanders are neutral

    Nor is a photograph a legal statement, so it can not be hearsay.
    Under the FRCP, they are covered under Article X, not article VIII.
    The point that you are both missing is that in general photographs
    are considered documentary and/or illustrative evidence and _not_
    physical evidence.

    With blood spatter or other physical samples, you need the actual
    physical sample that was collected. Not so with documentary
    evidence. At worst, all you need to do is produce an "original
    document," and for electronically stored data the legal definition
    of an original has evolved to the point where it's almost
    They are free to challenge it as long as they have a good faith basis
    for doing so. But the evidence is in, along with the testimony that
    the photo, _as it exists in court_ is a fair and accurate representation.
    So effectively, they have to convince the jury not just that the
    photograph was altered, but that the witness was lying on the stand.
    Tough job.
    If I was trying to impeach a photograph introduced as evidence, I'd
    _love_ for you to try to establish a chain of custody for just this
    reason. Each witness you call gives me another chance to plant
    seeds of doubt through repetition in cross-examination.

    Q: Isn't it true that you weren't there when this photo was taken?
    A: Yes.
    Q: So if was inaccurate or unfair you wouldn't be able to tell,
    A: Yes.
    .... etc.
    Michael Benveniste, Feb 13, 2010
  9. Charles

    John A. Guest

    If you can't get a reading on it from a dark frame (and/or similar
    image sampling), it's not going to be detectable in an image and is
    therefore moot in regards to image authentication via raw image
    John A., Feb 13, 2010
  10. Wrong. Just because you can't see them doesn't mean they aren't
    detectable. The point of a factory defective pixel map is to activate
    algorithms which reduce the visibility of defective pixels, by nearest
    neighbour cloning, in the final image. If you know where the defective
    pixels are, it is very easy to detect the application of these cloning
    algorithms - especially in areas of fine detail. However it is much
    more difficult, almost impossible, to do this without the defect map to
    begin with.
    Kennedy McEwen, Feb 13, 2010
  11. Charles

    John A. Guest

    Did I say "see"?

    This is digital. We have numbers to work with.
    John A., Feb 13, 2010
  12. I think we'd better agree to disagree on this one. Police photographers
    are still police, making it easier for the defense to assert bias in
    favor of the prosecution or motive for altering the shot. The question is
    one of believability, not one of detail.

    In fact, in at least one case I know of, the court stated that the
    poor quality of a photograph lends weight to its authenticity. US
    vs. Edward F. Nolan, Jr., 818 F.2d 1015.
    Again, not so. Hearsay is generally inadmissible since it can't
    be cross-examined. It can be extremely detailed and of extremely
    high quality and still be inadmissible. The passer-by snap may
    not show as much detail as a carefully planned after-the-fact
    crime scene photo, but it can be far more probative.

    The chain of custody of the Zapruder film is perfectly clear. It
    doesn't make what it shows any less questionable nor any less
    subject to interpretation.
    Read what you wrote, especially the paragraph beginning. "Maintaining
    a chain of evidence for all physical evidence..." If photographs
    are not physical evidence, then that entire section of your post was
    Cite please? They are not according to any legal theory I've been
    taught, and the law on this was established long ago with X-rays.
    UV and IR photos are used to illustrate the testimony of whoever is
    interpreting the test and result.

    What _is_ necessary for IR and UV photography is evidence of the
    _equipment_ used to gather them and it's capabilities and correct
    operation at the time the photograph was taken. Here, a chain
    of custody of the _gear_ is more important than that of the photo.
    Cite please? You continue to argue by assertion.
    "Late entry?" Introducing a fact not in evidence? Again, the
    question is one of believability, not of detail. I know that many
    cops and ex-cops hated for amateurs to photograph them while they
    are working, but there are literally hundreds of cases where amateur
    photography and video have been the strongest evidence in the case.
    You've already agreed that the trier of fact is the one who determines
    if a photograph has been altered or not. For them to do so, it must
    have been admitted into evidence.
    A fair amount, but I admit it's been quite a few years.
    Congratulations, and I honor and respect your service. I stopped
    practicing law quite a while ago, but the basics of evidence and
    admissibility haven't changed.

    I know of no cases where a court has refused to admit a digital
    photograph based on the lack of a chain of custody. I know of
    no cases where a "fakery" defense rested on a chain of custody.
    I _do_ know of some child pornography cases where the defense
    claimed that the contraband image was totally computer generated
    and therefore failed to meet the legal definition, but a chain
    of custody wouldn't have helped there either.

    If you know of such cases, please cite them here. I do note with
    irony that you are trying to bolster your believability based on a
    showing of ethos rather than on even documentary evidence.
    When the facts are against you, attack the presenter, right?

    See above. I freely admit that I don't have the extensive
    courtroom experience you have, but when one's opponent gives
    me the gift of multiple opportunities to reinforce my case, I
    know how to take advantage of it.
    Michael Benveniste, Feb 13, 2010
  13. Charles

    Ray Fischer Guest

    Claims by the defense that the police are corrupt seldom go over very
    well with juries.
    Ray Fischer, Feb 13, 2010
  14. The jury knows who pays crime scene photographers, and who pays them.
    And they watch TV and movies, so it's easier to create doubt no matter
    how professional and clinical the photographer actually is. After all,
    Gus Grissom hangs out with Jim Brass, not with the defense attorneys.

    As prosecutors are finding out with the "CSI effect," the impact of
    TV on juries is very real and exploitable.
    With every type of evidence imaginable, yes. The problem is the
    late entry itself.
    I agree that interfering with a crime scene is not only stupid but
    criminally stupid, but the antagonism towards "civilian" photography
    goes well beyond that. For example, how is someone filming an arrest,
    such as happened here, contaminating a crime scene?
    Again, I look forward to you citing such cases.
    There is undoubtedly an element of theater in any trial. If you're
    going to claim that digital watermarking and the like add gravitas
    to the prosecutions case, I'll happily defer to those people who
    spent their careers in the trenches.

    The reason there have been fairly few "faked photo cases," is that
    it's typically a lousy trial tactic for either side. It's easier
    to create doubt in the interpretation or witness bias than in the
    actual photograph.

    As an example, consider O.J.'s civil trial. In that trial, the
    defense tried to assert the shoe photos were faked. It failed
    miserably, even though the damning photos were taken by a civilian,
    there was no chain of custody, _and_ some of the photos were "late
    adds" to the plaintiff's case.
    Michael Benveniste, Feb 13, 2010
  15. Alleging corruption is a very different thing than implying bias. There
    are many ways to present a photograph so that it is still accurate and
    fair, but still generates the emotional response you are looking for.
    Compare, for example, a mug shot to a good professional portrait.

    We expect our judges and juries to be impartial, but we shouldn't
    expect the same thing from the police or state criminalists.
    Michael Benveniste, Feb 13, 2010
  16. No, you said "detectable" - as did I.
    Numbers don't help, especially when most of the precision has been
    thrown away in creating that final 8bpc jpg. With no way to even
    reproduce the original fully functioning pixels with the same precision
    as an original raw, you have even less chance of finding the defective
    pixels in a bit reduced jpg let alone recreating the same relationship
    between defective pixels, even if you could find them, and their
    neighbours. Its a one way street - without the key (the defective pixel
    map) there is no way to unlock the door.
    Kennedy McEwen, Feb 14, 2010
  17. Charles

    John A. Guest

    Numbers is all that there is.
    You don't need to recreate the original raw, just create a plausible
    raw file with realistic sensor noise. Assuming you have access to the
    camera you want the synthesized raw image to *appear* to have come
    from, you can get that from dark frame RAWs.

    In fact, the more loss there is in the JPEG, via rounding or
    compression, the less the original sensor's noise will be detectible
    and the easier it would be to fake a raw.
    John A., Feb 14, 2010
  18. Charles

    Ray Fischer Guest

    That has almost nothing to do with police work and everything to do
    with lawyers in courts. They select the evidence to be presented, and
    if a photo is believed to be biased then it's the job of the attorney
    to show how and why.
    Now all you have to do is show any bias. Granted, forensic photos of
    victims are often emotionally disturbing, but that does not make them
    Ray Fischer, Feb 14, 2010
  19. When the numbers are the same they don't tell you anything.
    I didn't say you did need to recreate the original raw but, as explained
    previously, you can't simulate realistic sensor noise in the raw since
    so much of the data is thrown away to get to jpg in the first place. For
    example, going back to the linear space used in raw from jpg leaves many
    missing codes, especially in the mid tones and highlights. If you add
    enough noise to conceal these missing codes then you are adding much
    more noise than would be in true raw file, and that is simple to detect.
    It might work in the shadows, but that's about it.
    No - you can't get the factory defect map from a dark frame!
    Rubbish - compression artefacts of jpg are completely different from
    noise. Apart from anything else, jpg compression aretefacts have a
    significant spatial component due to the 8x8 dft core, whilst the noise
    on raw has no spatial component with low correlation between pixels.
    Adding sufficient random noise to conceal these spatial jpg artefacts
    makes it very easy to differentiate the fake raw since, as above, the
    noise levels as a function of image brightness would be very different
    from a genuine raw.
    Kennedy McEwen, Feb 14, 2010
  20. Charles

    John A. Guest

    Which let's you tell the story you want to tell, without the JPEG
    contradicting you.
    Who said anything about getting it from the JPEG?

    Get the noise profile, via dark frame exposure(s), from the camera to
    which you want to attribute the picture.

    It's actually better if the JPEG obscures all the real source camera
    sensor noise. Covers your tracks.
    Don't need it. If it's being filtered out of the data going into the
    RAW files produced by the camera you want the picture to appear to
    have come from, it doesn't matter. To fake the RAW for that camera you
    need to add the noise that *does* get through to the RAW. Otherwise
    you're saying "Here's the RAW for that shot, which clearly came from
    my camera. Ummm... not sure why the dark & cold pixels got through to
    this one and not my other RAW shots from this camera, but that's no
    reason to be suspicious. Honest!"
    So smooth them out at a higher bit-depth, convert to the curves &
    bit-depth the RAW gives, and apply the sampled sensor noise
    accordingly. IIRC there are existing programs that will do that kind
    of de-noising. (It's not like there isn't a demand for it.)

    Come to think of it, instead of a dark frame it might be better to
    expose a grey card then normalize to find the up *and* down
    adjustments from the mode for each channel.
    John A., Feb 14, 2010
    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.