Re: Judge declares zoom of iPhone image unacceptable in court

[Kaz Kylheku (gnu-misc-discuss)]

On 2021-11-15 13:36, Akira Urushibata wrote:
> I work on image processing software.  I may, one day, be summoned to
> court to testify on the technology in use here.  Software that
> enlarges images add pixels, but generally this is done by
> interpolation.  A new pixel between an existing red pixel and a yellow
> pixel is colored orange.  By this method, high-contrast details do not
> emerge from the void.

There is the rub. Since high-contrast details do not emerge from the
void, there is no point in doing such an enlargement in the first place.

Anyone trying to convince a court of something using some enlarged 
digital
image can be assumed to be inventing some pixels. Without those, nothing
new will be seen that the naked eye cannot see by looking closely at the
original image: closely enough to see the original pixels individually.

This has nothing to do with closed source. Even if the implementation
of the enhancing algorithm is free, it has to be validated to be
correctly based on some reference mathematics, and that the algorithm
isn't being misused to show something that really isn't there.

All of that is entirely outside of the scope of a murder trial. You
are talking about bringing in a lot of signal processing expertise
into the case, all for a topic that doesn't directly pertain to the 
case,
but which rather pertains to one party's insistence on being allowed to
use some particular kind of image enhancement algorithm to manipulate
the evidence.

The judge is basically right to toss this out, without a solid
system in place for dealing with claims of that type.

We are probably going to see a lot of this going forward.

It seems that a potentially good solution would be for the justice
system itself to provide certain signal processing tools, such that
only those tools may be used, and only in certain ways.

For instance, if you want to draw the jury's attention to, say, the
high frequency content in some image, then you just use the software
provided by the court, in specific ways. The result is then admissible
as bona fide high frequency content in the original evidence, and
whatever it makes more easily perceived is ruled as being real.

That software could be free; anyone should be able to obtain it in
source form. But only a certain fork of the software, built in a certain
way, run on a certain system, would be approved for manipulating
evidence such that the results are admissible in the courtroom.

For that matter, specific forensic staff would do the manipulating,
not the lawyers, defendants or plaintiffs or agents hired by them.
The original material would be submitted to the court formally,
with instructions on what is to be cropped, enlarged, enhanced or
whatever else. The deliverables from that would then formally come back
and get attached to the case. Or some workflow of that nature.

It's no different from any other forensics. For instance why and how is
DNA evidence admissible? It must be based on following very specific 
procedures
in handling and processing the evidence. You can't come to the court 
room
and, "Oh, Your Honor, I analyzed this piece of fabric from the crime 
scene
with a DNA munging system I ordered off Alibaba ...". Surely, that must
be done professionally.