It’s almost surprising this hasn’t happened before, but it’s happened now. Tyrone Lamont Allen was a suspect in four bank robberies. Why he was a suspect in the first place is unclear, as the most that could be said was that random callers who saw the image of the bank robber on TV said it looked like Allen, which isn’t much to go on,

But what was clear that one unique aspect of Allen’s appearance was missing from the images of the perpetrator and the description by witnesses. Allen had facial tattoos.

When asked to ID the perp from his mugshot, witnesses failed to pick Allen. This was a problem for the government, since they had already decided Allen was guilty but lacked the identification to prove it. They came up with a plan.

They covered up every one of his tattoos using Photoshop.

“I basically painted over the tattoos,’’ police forensic criminalist Mark Weber testified. “Almost like applying electronic makeup.’’

With a little photoshop magic, they took an image of Allen and conformed it to the image they hoped would get dinged by the eyewitnesses.

Is this a pic of Allen without tats or an image of a person who doesn’t exist but conforms to a description of a bank robber? It worked.

Police then presented the altered image of Allen with photos of five similar-looking men to the tellers for identification. They didn’t tell anyone that they’d changed Allen’s photo.

Some of the tellers picked out Allen.

Naturally, AUSA Paul Maloney neglected to mention that the iD was made from a doctored photo, but Mark Ahlemeyer spotted it and challenged the identification.

Assistant U.S. Attorney Paul Maloney defended police actions.

“The whole idea was to make Mr. Allen blend in – so his photo wouldn’t stand out,’’ he said. “These procedures were prudent. They were appropriate.’’

Maloney said the altering of Allen’s photo was done to “look like the disguises that were on the robber,’’ who wore a baseball-style hat and glasses, with no tattoos visible.

The “blend in” argument is backward nonsense, relating to the need to use a photo array where the images were sufficiently similar that one, the person who the government seeks to have identified, doesn’t stand out from the fillers so that the array isn’t “suggestive.” In other words, if the government sought a non-suggestive photo array, they should have included only images of people with facial tattoos, not removed Allen’s tats.

But the government knew that wouldn’t to get them a hit.

None of the tellers had described seeing any facial tattoos on the man who had demanded money from their banks. Video surveillance images didn’t show any tattoos on the thief’s face either.

So it couldn’t have been Allen? They had a theory.

Allen could have used makeup to cover up his tattoos, Hawkinson said. One of his objectives, he said, was to rule Allen in or out as a suspect and determine whether the sources who contacted police were credible.

At first blush, the “makeup” theory would seem to make sense. If a guy wanted to rob a bank, knowing he had rather unique facial characteristics that would make him very easy to ID, he could “sanitize” his appearance with makeup. But upon closer scrutiny, the theory has a flaw: the amount of makeup necessary to cover up Allen’s facial tats would itself have become an identifying characteristic. Even assuming he was highly skilled in the application of makeup, it would require so much makeup to accomplish a complete cover-up that the description would have been “guy in baseball cap, glasses and face covered in makeup.”

At the suppression hearing, Ahlemeyer challenged the use of the doctored pic which, as far as we’re aware, is the first time the government got caught using a photoshopped image in a line up to get an ID.

Ahlemeyer, on cross-examination, asked if any of the protocols instruct police that they can alter someone’s photo.

Hawkinson said no, but quickly added nothing tells police that they can’t change an image either. He said he learned of the practice through “on the job training’’ and from police supervisors.

It’s “standard practice among investigators,’’ he said.

Standard practice, evidently, is make up new “rules” to phony up evidence when you can’t get an ID through the use of an undoctored image. Perhaps the reason there was “nothing” that told police they can’t alter a photo is because it never occurred to anyone that such a rule was necessary, since doctoring images to get IDs seemed a bit too obviously wrong to need saying.

It was better to remove Allen’s tattoos from the lineup photo, he said, because they could be too distracting for witnesses. “The purpose of any alteration is not to change the physical attributes but to mask things that would stand out,” he said.

“You don’t consider tattoos to be physical attributes?’’ the defense lawyer asked.

Hawkinson replied that tattoos could be added or removed.

Removing tattoos is a bit more troublesome than adding them, and adding them wasn’t at issue in this case.

Contrary to the common assumption that eyewitness identification is strong evidence, it’s long been known, and overwhelmingly empirically proven, that eyewitness identifications make for very strong testimony but very unreliable evidence*. This is particularly true for cross-racial identifications. When photo arrays, even double blind as here, are done properly, they still present grave problems of mis-identification. When done with altered pics, they’re just goofy.

While the issue facing the court at suppression hearings is the suggestibility of the photo array, this case introduces a problem that goes outside the parameters of the normal hearing. Current FBI guidelines dealing with unique facial features provide a mechanism to address these issues.

When a suspect has a unique feature such as a scar or tattoo, agents should find so-called “filler” photos of people with a similar feature. If they can’t duplicate a suspect’s unique feature, then they can black it out and place a similar black mark on the other “filler” photos.

Here, they manufactured an image of a non-existent person instead, and that person was ID’d as the bank robber. Unfortunately for Allen, that non-existent person won’t be the one to serve a sentence if he’s convicted.

*Except in cases involving sexual assaults by men on women, where all known problems with empirical evidence are reversed, according to the New York Times. But I digress.