The 1000 ppi "standard"??

[antonroland]

OK, folks, I have been lying low for a while but saw Boyd's opinions on other threads I seek your inputs please...

Now, let me immediately insert my plethora of objectivity disclaimers here but, for one I have a head cold causing me to spend a beautiful weekend indoors so I might not be my regular sunny self...but here goes.

I would like to know if it would be worth commenting or do I simply ignore it as I do with many local (did I say bird-brained? Did I? Really?? ) SOP's?

Surely the whole 1000 ppi / dpi thing, valid as it MIGHT have been belongs in a museum next to 1 Megapixel D-SLR'S??????

Further to this I sense what I perceive to be a shocking lack of practical knowledge and / or totally impractical assumptions in these draft papers and other already "adopted" papers. We need look no further than the "A NOTE ON 1000 ppi STANDARD" topic of the current Section 22 draft.

Now, on this whole 1000 ppi thing...it is my understanding that hand-held viewing sized photos are printed at a nominal 300 dpi. Yes, I know it can vary based on optimal sharpness or economy settings on some printers but "Photo quality" is generally 300 dpi. Even the good people at Adobe and a few other places seem to understand this...

I would like to believe that scanners have evolved around multiples of 300 dpi and for this very reason a typical current high end A4 flatbed like, say, Epson's V-750 has a maximum native optical resolution of 9600 dpi...divisible by 300, big surprise there!

Now I am fully aware of real life limitations on flatbed scanners due to my experience as a part time commercial photographer experienced in both film and digital media.

Now here is my main gripe: If L3D is what we want, surely we need to work with scanning resolutions that carry the least possible if any interpolation? Why then are we so hell-bent on working a 1000 ppi/dpi "standard" that is neither reconcilable with the whole commercial and professional imaging industry out there nor with the nice hardware they develop for us?

What am I missing here?

[Steve Everist]

Now here is my main gripe: If L3D is what we want, surely we need to work with scanning resolutions that carry the least possible if any interpolation? Why then are we so hell-bent on working a 1000 ppi/dpi "standard" that is neither reconcilable with the whole commercial and professional imaging industry out there nor with the nice hardware they develop for us?

What am I missing here?

It's a round number.

When I teach my charting workshop, I ask the question of what is the minimum resolution necessary to conduct a latent print comparison. It's multiple choice with the choices of:
1200 ppi
1000 ppi
600 ppi
None of the above

1000 ppi tends to be the number most often chosen - thank you SWG's, NIST and AFIS vendors. I remind people who include SWGXXX standards in their SOP's (if they have any) that they probably should know what they are, because they may be challenged by an accrediting agency or in court on their procedures which could refer back to these documents. And if they include one of the documents that mandates a minimum of 1000 ppi, then they'd better have it. Then I go on to show a few images of prints that I shot with my old Nikon 995 camera (3.1 megapixel) and I ask if anyone would feel uncomfortable with the quality of the image for comparison purposes. Then I calibrate it... under 1000 ppi. If their SOP's mandated 1000 ppi, then they may not be able to use it.

Unfortunately not many people are aware of the Frye challenge in State of WA vs Eric Hayden (http://caselaw.findlaw.com/wa-court-of- ... 89825.html). But I explain that the camera used in this case couldn't resolve near a 1000 ppi capture, especially of a palm. This wasn't the challenge in this particular case, but it's case law regarding the use of digital imaging/enhancement for latent print purposes, and included prints that wouldn't meet today's "standard".

Needless to say, I'm not a fan of a minimum resolution standard, because as with so many things in latent print comparison... "it depends." However, although I'm against a standard such as this, I tend to try to get as much resolution as possible in my images. I usually flatbed scan at 2400 ppi (largest native resolution our scanners will support), and I fill the frame as much as possible with my latent prints when I capture them digitally. I'd rather have too much than not enough. And with most of the cameras today - that shouldn't be too difficult.

[antonroland]

So we use 1000 ppi / dpi because it is an easy number to remember? Honestly?

IIRC, around about 2001 the FBI published an approved hardware list for AFIS / IAFIS hardware and this listed a number of scanners capable of 500 dpi / ppi. As time and technology moved on it would seem that this number was possibly simply doubled?

Also (and again if I remember correctly) Sirchie and possibly other vendors had a product that attached to the front of a standard lens enabling a 35mm camera to shoot a single latent finger at 1:1, very likely due to an optical element in the product allowing focus at approximately 2 inches or so as opposed to the regular 45-50 cm required by most standard lenses.

When palm prints had to be photographed in the field and a 35mm camera was what you had a prime macro lens was useless unless it was used from a distance thereby coming nowhere near 1:1...somewhere between 1:4 and 1:8 possibly. It was good enough because it was all we had.

Interesting link on that case law BTW, thanks! It is a few years old though, any chance of more recent case law or that case having been applied in other matters? In any event that case helps the State much more than the appellant and so that is fine right there!

I appreciate that they have a noble cause but I really wish that bodies such as SWGIT would simply go that extra 5 - 10% and issue more detailed real-life guidelines rather than the one-size-fits-all-agencies documents currently out there.

[RL Tavernaro]

The 500 ppi & 1000 ppi 'standards' pre-date your 2001 date by a number of years. Although I don't recall the original source, my recollection is that 500 ppi was a standard used in relatively early AFIS , prior to the cost-effective availability of higher resolution scanners, when fixed, special purpose digital cameras were used for image capture as part of the system. The cameras were calibrated for capture at 500 ppi, in accordance with what was then considered the minimum resolution required for reliable capture of 'level 2' detail in digitally captured inked fingerprints.

The FBI, among others did tests demonstrating that 500 ppi was not sufficient in many cases for poorer quality latent prints, however 1000 ppi could eliminate much of the artifacting (creation of false features) and misrepresentation of legitimate features that occurred at 500 ppi. The costs associated with both processor speed and memory at that time made those standards the most cost effective for the relative database sizes, and the latent print image info could easily be downsized to be compatible with the tenprints.

Like Steve, I photographed many latent prints with a Nikon 995 3.1MP camera. Many of those images were of sufficient quality to resize to the 500 ppi image resolution required by our AFIS, & resulted in hits, and many more (including palmprints) could be identified to known suspects. So I agree with Steve that 'it depends'.

While it is true that off-the-shelf hardware is now available to significantly improve the image resolution & maximize image quality, there are huge AFIS databases of tenprints at 500 ppi that current exist. What would be the cost of converting those databases, assuming that they all could be converted without unacceptable image degradation?
Regards, RLT

[antonroland]

I don't think I ever implied transforming existing databases as this would indeed be costly if not near impossible.

I do feel that our future captures (and guidelines about them) should keep up with the times and tools we work with today. In this regard I am HUGELY ENDEBTED to GEORGE REIS.

If you do not have his book, get it now, don't even read further, order it! NOW!

I also feel that, for the sake of this discussion at least, we need to differentiate clearly between the photographic capture aspect and the scan capture aspect.

Now please allow me to explain my sentiments at the hand of an example or two as I have had more than one PM from members who admit to having difficulty with this stuff...

Our national policy is that we go to court with a comparison chart "enlarged at least 3x" but 3x has pretty much become the standard.

If I now scan my tenprint card to prepare, say, the right middle finger I would need to scan it at 900dpi in order to reproduce it 3x lifesize. 300dpi being photo quality for hand-held / arm length viewing distance.

The only problem with this is that I am scanning at an odd resolution. Here's why:

The Epson V-750 A4 flatbed scanner I am using has a maximum native optic resolution of 4800 dpi. This is the best resolving power number that the optical qualities of the scanner's lens assembly will allow.

Two other points (at least...) on the Epson V-750 need to be mentioned here...1) Yes, it has a dual lens system which gives it 9600 dpi capabilities and with interpolation it can be pushed to 12 XXX dpi...

2) Beyond 2400 dpi there is rarely any increase in quality when scanned at larger resolutions so a drum scanner for negs and transparencies are still better...but that is irrelevant to this discussion...

Getting back to my comparison chart, I have now scanned the #3 at 900dpi BUT I introduced unwanted interpolation into this image...what would have been better would be to scan it at 1200 dpi which is a half or half of a half factor of native optical in the scanner. Resizing it down in Photoshop from 1200 to 900 would be by far the better way to go about things.

My latent is slightly more complicated as it is a UV capture on digital camera.

My scanner knew what scale the image is, the camera does not and so I have to assign a scale to it and then resize the image to print at an exactly predictable magnification factor. This really is very easy and entails about 6 or 8 clicks.

It is admitted that many of these steps containing interpolation but my point is that we should know the tools we work with and minimise the gremlins we introduce to our images.

I undertake to do some tests and post them up here if anyone would be interested. That way we can have some verifiable material proving that either (a) I am onto something or (b) I wasted your time all along...

Hoping on the former...

Have a great day all!

[RL Tavernaro]

Anton,

I am not in opposition to your views, and the subject has perplexed me as well (& I am familiar with George Reis & his book). I merely sought to offer a little historical context and speculate on at least one reason there could be resistance to change. Although not clear in my previous post, I did not mean to advocate for that position.
Regards, RLT

[antonroland]

No worries, R

I think I understand the way you meant it. Here's to hoping we can improve it from here for the future...

Will post examples on here soon!

[Steve Everist]

Like Steve, I photographed many latent prints with a Nikon 995 3.1MP camera. Many of those images were of sufficient quality to resize to the 500 ppi image resolution required by our AFIS, & resulted in hits, and many more (including palmprints) could be identified to known suspects. So I agree with Steve that 'it depends'.

Even at 3.1MP, I had to back off from my usual "fill the frame" mentality to purposely get these images to calibrate at <1000ppi. Only with a palm or extremely large finger would a camera of this resolution not be able to make that number.

While it is true that off-the-shelf hardware is now available to significantly improve the image resolution & maximize image quality, there are huge AFIS databases of tenprints at 500 ppi that current exist. What would be the cost of converting those databases, assuming that they all could be converted without unacceptable image degradation?

Regards, RLT

I'm glad you mentioned this. This is something that's come up in our office recently. Boyd Baumgartner found an interesting article from a couple of years ago relevant to the 500/1000 AFIS database issue: http://forensicbiometricidentifications ... 00ppi.html

I do feel that our future captures (and guidelines about them) should keep up with the times and tools we work with today. In this regard I am HUGELY ENDEBTED to GEORGE REIS.

Now we just need him to update it for when version CS6 comes out - I've show him the path to sefl-publishing, now I hope he'll step into those waters having tasted the major publishing route.

Getting back to my comparison chart, I have now scanned the #3 at 900dpi BUT I introduced unwanted interpolation into this image...what would have been better would be to scan it at 1200 dpi which is a half or half of a half factor of native optical in the scanner. Resizing it down in Photoshop from 1200 to 900 would be by far the better way to go about things.

I'm guessing this is a little mysterious to many - the idea of interpolation when scanning down at a fraction of the native resolution of your scanner. Of course 1200 would be a 4X enlargement of a 300ppi resolved printout (for those with longer than normal arm lengths…).

I never really thought too much about knowing specific magnification regarding resizing images for court. But I guess this depends on how they’d be presented – in your case the 300ppi for arm length viewing is a thought. But if they’re going to be placed on easel or be shown on a screen/wall as part of a visual presentation, there are so many factors that determining the exact ratio wouldn’t be that important.

But at the point of creating a chart, the comparison process would have been done already, and that’s what I assume the SWG’s are thinking about when creating their guidelines/standards. I’ve just never seen any justification for their numbers. They seem rather arbitrary – at least if it was 1200ppi, it could be justified by saying it’s a fraction of the native resolution of most scanners.

[RL Tavernaro]

Steve,

I'm glad you mentioned this. This is something that's come up in our office recently. Boyd Baumgartner found an interesting article from a couple of years ago relevant to the 500/1000 AFIS database issue: http://forensicbiometricidentifications ... 00ppi.html

Thanks for the link. The article touches on many of the issues that were going through my mind when I originally posted.
Regards, RLT

[Pat]

SWGFAST guidelines and standards, once written, are approved for five years, then come up for review. At the time the 1000 ppi "standard" was originally adopted, digital imaging for latent prints was still in its infancy. The idea was that anything less than 1000 ppi may fail to capture all relevant detail, especially at level 3. One problem was that memory was still expensive enough that many agencies were capturing and saving images at much lower resolution. SWGFAST did not mandate 1000 ppi, but issued a guideline of "at least" 1000 ppi. Individuals or agencies are certainly free to use higher resolution, if desired.

SWGFAST is always eager to receive input from the latent print community. Although originally convened as primarily having a US focus, members now hail from Europe, Canada, and newly elected from Australia. SWGFAST works to ensure that documents are appropriate for the science as a worldwide practice, not merely from an American perspective.

Input is always reviewed and seriously discussed. Therefore, if you express your concerns directly to SWGFAST, you can rest assured they will be read, discussed, and considered. Changes are made in both "draft for comment" documents and approved documents. All submitters of comments are contacted with a reply following the meeting at which their suggestions are addressed. The next meeting is tentatively scheduled for early March, 2012. Please send comments to the Executive Secretary via the "contact us" tab at www.swgfast.org. Your comments and suggestions on this topic would be most welcome.

[antonroland]

Mail was sent as per IAI site guidelines

Will also do some tests as promised and then post results on here and forward same to Mr Kriigel.

[Neville]

It is not about the resolution as much as about the lens. Many people think that by increasing the resolution you improve the photo, not true I can use a 10mp camera with plastic lens and get a a photo inferior to say and old SLR 5mp camera with excellent lens. As for scanners you are right and this thing about 1000ppi means very little if you are using a Epson V700 series scanner you could probably get a great scan at 600ppi where as if I was to use a 4800ppi crap scanner from China made by an unknown company the result would refect the price.

[Neville]

May I add that scanners have other issues the quality of the motor, the drive belt system, the position of the light bar and the software that stitches it all together. Then also the same issues that affect cameras the lens quality and the software that controls interpolation and enhancement.

[antonroland]

Very valid points there! For those very reasons I believe one should always go with upper level models of well-known and respected brands and never mark down cheap-n-nasties.

[Tazman]

. . . one should always go with upper level models of well-known and respected brands and never mark down cheap-n-nasties.

Right. I would like a new Nikon top of the line with the best lenses for crime scene and latent print photography, plus a new scanner that can make quality 2400 ppi scans. Plus a new computer with memory for TIFF format photos and the newest version of Adobe PhotoShop.

Sgt. Miller wants a new squad car.

Who wins?