Digital Video and Photo Evidence

The Rittenhouse trial has brought up some serious questions about to what degree a digital photo or video might be accurate or reliable enough to be evidence.

If you display a photo on a screen that is not the same resolution as the original photo, it is being scaled. Scaling a digital image means you are adding or removing pixels. What color those pixels end up being has no relation to the reality of the original scene. They are fabricated out of math. Sometimes they are interpolated values, sometimes software uses AI to determine what the image was of and guess at how to resolve the color of new pixels so that they are more likely to make logical sense.

No image adjustment is 100% reliable. Digital photo evidence must be taken at the native resolution of the file itself to be reliable. If you can’t see something in the original photo, it isn’t there at all.

Digital zoom is NOT the same as a magnifying glass. If it were, they would just hold a magnifying glass up to the screen of the iPad and get the same results.

Any photo manipulation used as evidence needs to be done by a digital forensics expert with a complete chain of tooling and custody so that every algorithm used and every setting can be scrutinized if needed. Each intermediate filter or algorithm applied needs to have the image saved and analyzed to make sure artifacts were not introduced by that filter or algorithm.

The scaling algorithms on phones and tablets are not reliable for anything more than aesthetics. Pinch and zoom should not ever be admissible as evidence. None of us have any way of even knowing which image scaling algorithm an iPad is using, check out this incomplete list of publicly disclosed algorithms on Wikipedia.

I am not a digital imaging expert, but I have a degree in computer science and understand how display technology and image encoding work. I am very happy that the judge did not allow the scaled image to be shown to the jury, but I also feel the digital image evidence has been grossly mishandled in this case.


I agree that algorithms are not reality, any more than a photograph is the object or a video is the event or eyewitness memory is a record. But photo interpretation and image enhancement are real things, so I would not outright disallow their use in presenting evidence. The merits and meaning of both original and altered images should be defended and attacked by subject matter expert witnesses, just like other evidence — not just laid out as “reality” with a frosting of lawyer jabber.

Certainly, any chain of image development needs to start with whatever capture is originally recorded — raw off the CCD or 1st gen compression. Same as a film photograph is originally the film, not the print — except Polaroid-type processes where the original capture is generally not available. But “pinch” and “zoom” are about screen display, not the image file — scaling a screen display on a specific device is often necessary in order to view the full actual resolution of the capture at 100%.

I would also agree that a random screen dump is among the least defensible representations of a captured image. There may be times when it is the only thing available. The merits of such an image should be debated, but may be stronger than uncontrolled (or intentionally manipulated) application of “filters” and other algorithmic artifacts. I definitely don’t know the ins and outs of hiding and detecting altered images.


I agree that outright disallowing them would be a disaster as they can be evidence of innocence or guilt and mostly they are reliable.

I think detecting altered images is another topic.

I’m suggesting that an image is collected as evidence, any alterations after that point including scaling should be done with forensic and scientific rigor and thoroughly logged. In the case of social media or cloud storage of any kind, transcoding is common where they automatically downsample or scale the image to be “good enough” or “optimized for the website”. Any such images need to be scrutinized for alterations before being considered good enough quality to be subject to processing as evidence before trying to enhance or interpret them.


I remember when I saw this original broadcast episode of Star Trek, and thought how impossible it seemed:

Not so impossible now, is it?