The merits of photographing a latent at 1000 ppi

[antonroland]

At the risk of being considered a spammer to up my post count but here follows a (what I consider to be logical) flow of info as to how I did the tests.

The target chart contains (from left to right) inked impressions of my right index and right middel finger and my right thumb.

All I did in all images about to be uploaded was crop to 800px X 533px and Save for web at "very high" jpeg quality to stay below 256Kb.

I thought at first to show how small latent fingers appear at 1000 ppi

[antonroland]

Now, to be fair, it is quite clear that the image above shows slightly more than the 2" x 3" block which is a fault of mine from rushing it and forgetting to think about the fact that D-SLR viewfinders are not absolutely correct in terms of coverage, so please excuse the approximately 5% inaccuracy - unexcusable for case work of course.

Even though, admittedly this is only about a 955 - 960 ppi capture, I am sure it will get the point across.

I can also submit that the 1000 ppi guideline might have had it's origin (at least partially) in practicality? The image file size in pixels is very easy to ascertain and then convert to a target size to be photographed at 1000 ppi.

I do not feel that cameras are prone to interpolation at different resolutions as scanners would be and so filling the frame would get my vote every time. Subsequent reworking to a specific size might strictly amount to interpolation of sorts due to downsampling but personally, I would rather fit more into less space than stretch a little to become bigger...within reasonable limits of course. We know that printing a latent 1:1 is a rather silly idea given the amount of detail and the limitations of the printing process and paper.

I did a quick numbers crunch and came up with a ppi score of around 2250 - 2290

Again, all images in this thread are 800 x 533px crops of the complete 6Mp jpeg colour captures, resized in "Save for web" to "high quality" JPEG. Nothing else.

[antonroland]

Not being a LPE, let me ask a few questions. First, what size are the 'features' of interest that need to be captured? According to the standard, they will be adequately rendered if you use 1000 ppi. I'm sure that using a higher resolution would give images that more fully represent the original being imaged but to what end? When Steve advocates "using best practices to get the most out of your image" what does this mean? If the imaging being done is that which the standard addresses, then "the most" isn't really needed; it's over-kill. The standard says 'minimum' but, once that has been achieved, the point is whether or not 'more' actually means 'better'.

Just my 2 cents worth...

Hello there.

I trust that the images preceding this post will be self-explanatory. In my economy, the "bigger" the capture the better with no real upper limit.

Please bear in mind that these images I quickly slapped together show ideal latents and not the type of prints one would typically encounter in the field

The next interesting comparison, even if only performed for comic relief, would be comparing 35mm film with digital.

The closest I would be able to compare would be to use the same macro lens on both the * ist DL body and also on one of my old K-1000 cameras with a roll of Ilford FP-4 maybe?

The only real difference would be difference of perspective between the systems...true 50mm on film vs 75mm focal length perpective on digital.

I think I will be limited to a flatbed scanner for the film as I am not aware of a decent drum or other dedicated film scanner very near me. Could mail it off somewhere I suppose.

Let me know if anyone thinks this is worth seeing.

[George Reis]

When I worked for a police department, and began scanning fingerprints, my goal was to scan at the lowest resolution that provided the level of detail necessary for the specific print. This made the scans faster for good quality prints (because they were lower resolution); made the files smaller for storage and transfer; and made image processing faster. I had plenty of scans that were 600 PPI and under that worked just fine (I think I scanned good quality tenprints at 300 PPI, good quality latents at 600 PPI and marginal prints at 1200 PPI). I found that scans under 150 PPI could create some moire patterns and were completely unacceptable in most, if not all, cases.

For photography, I always filled the frame - this is such an easy approach to photography, and whether the frame is filled or not, the file is the same size. I think the goal of those who calibrated their camera to photograph latents at a specific resolution was to make sizing the image 1:1 automatic. I always found sizing an image to a specific size quite easy, so never personally saw an advantage to that (I also used my latent camera in the field, so recalibrating it would be more work than filling the frame). When I first started doing this I had a 1.7 MP Kodak DCS 200 camera, and got great images. Later my 3 MP Nikon 990 did a great job as well. Now I see agencies wanting 16 MP cameras to photograph latent fingerprints and I wonder if they really think they will get something that they couldn't get with a 3 MP camera. Obviously, if they are photographing larger objects - palms prints, foot prints, etc. - there is an advantage to the higher resolution.

I don't think most agencies are even aware of SWGIT or SWGFAST. It's great that we have volunteers from various agencies who work hard at trying to help those in our field. But, I think that some of these can backfire when it comes to court - I can imagine an attorney on cross examination bringing up these "standards" and the latent print examiner not knowing what a SWG is; or not being familiar with the "standard;" or knowing that they did everything right, but don't know how to explain why their image is only 800 PPI and therefore falls short of the standard and therefore shouldn't even be considered of comparison quality!

The bottom line really is knowledge, training and experience. With these, we can develop a good workflow, choose the file formats to use, choose the resolutions we prefer, explain our methods in court, etc. Thanks to Kasey for providing this forum to make it easier for all of us to share information!

George

[antonroland]

Thanks to all the contributions in this thread I think we can safely summarize by saying that 1000 ppi is a safe minimum guideline for capturing latent prints but that filling the frame will always yield best results.

I must admit never to have considered the data movement limitations fully as I am sure these vary greatly from place to place.

George, is it not sad that more folks are not aware of these bodies and guidelines? I realise all too well,looking around at some of my colleagues that every individual is not the type of person who would keep on digging and never rest before they found the answer to the last question.

I feel that it is the responsibility of managers and supervisors to inculcate this culture of perpetual learning.

Possibly more sad, in my opinion, is the fact that guidelines such as these do not keep up with the times.

Looking forward to any further comments in this.

[George Reis]

George, is it not sad that more folks are not aware of these bodies and guidelines?

If these were "guidelines" and not "standards" and if they included enough information to prevent their misinterpretation, then perhaps so. But, I think that the way they have been written, the way they are currently written, and the move to make these into standards, make it preferable that fewer people know about them.

If I were writing a guideline for fingerprint capture, I would suggest that one scan or photograph at a resolution that shows the detail of significance for the specific print. I would then list scanning rates that generally will capture first, second, and third level detail. This would be followed by information about native resolutions, bit depth, imager quality, etc. - to indicate that image quality and scans/photos at identical resolutions will vary because of many factors. I would not state that there is a minimum resolution that anyone should use - instead I would give "guidelines" so that individual agencies can develop policies based on this information, and their own equipment, workflow, knowledge, experience, and training.

So, as much as I'd like to agree with your statement Anton, I cannot do so currently.

As to your comment earlier in that same post:

Thanks to all the contributions in this thread I think we can safely summarize by saying that 1000 ppi is a safe minimum guideline for capturing latent prints but that filling the frame will always yield best results.

I understand why you say this, and I guess the word "safe" brings the right tone to this. But, I don't think 1000 PPI is needed in many cases; if it's interpolated, then it may not be as good as a lower, native resolution; and it really doesn't speak to the main issue of what resolutions with which specific equipment, captures what level of detail. None of this is rocket science, and I think all of us can photograph or scan a latent, view it, and determine if we have the detail we need. If we do, then move on to the next print; if not, then we need to consider the resolution, bit depth, interpolation, file format, internal image processing (by camera or scanner), imager quality, etc. We can run some simple tests, come up with basic internal guidelines, and there is really no need to debate whether 1000 PPI is right or not because we will know what works with our own equipment, at our own agency, with our own workflow, etc.

George

[antonroland]

George

Surely the camera and lens combination is not subject to a maximum / native resolution as say, a flatbed scanner would be, is it?

I read Gerald's discussion on line pairs of resolution and it makes sense in broad strokes but it primarily tells me that I have LOADS to learn still - not that I ever doubted this but the quantities I do not yet know become painfully clear.

[George Reis]

Surely the camera and lens combination is not subject to a maximum / native resolution as say, a flatbed scanner would be, is it?

Well, that depends. Suppose I photograph a latent and find that calibrating it to 1:1 without resampling gives me 1200 PPI or 800 PPI, and then I decide to resample to a 1000 PPI "standard." By doing so, I am resampling my digital photograph in the same way a flatbed scanner may resample when scanning at 1000 PPI. I would likely be much better off not to resample. And, if I were purchasing a camera to use for photographing latents and was only photographing single latent fingerprints, I'd probably choose a lower resolution camera with a good lens, good quality CCD, higher bit-depth over a higher number of pixels but lower quality optics or less bit-depth, etc.

But, if I have a camera that I wanted to test, I'd do the same thing as with a scanner - I'd ignore the "standards" and shoot test images to see if they are of adequate quality to meet my needs. If so, I'd be set. If not, I'd need to consider several variables - file format, sensor size, pixel density, bit-depth, in-camera image processing, filling-the-frame vs. a set calibration, optics, etc.

Our imaging systems and workflow is a cohesive whole. We need to think of it as such, and not focus on a single value - like PPI.

George

[Gerald Clough]

Our imaging systems and workflow is a cohesive whole. We need to think of it as such, and not focus on a single value - like PPI.
George

That's the great truth. And the draft standard tends to drive one back to that quite rational position. If I can show that the image was captured and that detail was not created or destroyed beyond that captured by the sensor and preserved in the native raw format or by any conversions later, and I can show that the detail is sufficient to rely upon, I can rely on it. Not unlike the conventional "scene as observed" qualification. That doesn't mean I have to work from the RAW file, any more than I would have to use a negative viewed by projection as photo-documentation of a crime scene. The new draft somewhat corrects the ambiguity about 1,000ppi by clarifying that to mean optical resolution at 1:1.

(I'm not at all sure optical resolution can be stated in ppi. Does this mean 250 line pairs per inch for the sensor/lens combination? It takes 1,000pixels to resolve 250 line pairs, but it doesn't mean your system will resolve 250 lpi. Optical resolution is defined as system resolution. Pixel resolution is a somewhat valid measure of display resolution, but I don't think it means anything in optical resolution. Seems to me that if you are talking about optical resolution, you could as well say pickles per inch as pixels per inch. You can at least resolve pickles, assuming there is a standard pickle defined. I suppose they figured a discussion of actual optical resolution would cause readers' heads to pop. I always stop reading before mine does, which would occur somewhere in the Fourier transforms. But I expect better usage from a scientific working group.)

Consider, from the draft:

Friction ridge impression digital images to be used for comparison purposes shall be stored and
transmitted without compression or with lossless compression. Capture in raw file format is
recommended.
4 Image Integrity
4.1 A primary image is the result of the first recording of an image onto media. An original image is an
accurate replica (bit-for-bit value) of the primary image. Each original image shall be stored in a
manner which permits authentication

I suppose this language means RAW is preferred, but a JPEG can be a "primary image." (Unless we want to get technical and consider that the camera RAW is recorded onto "media" before being converted.) If a JPEG qualifies, then certainly it can be replicated bit-by-bit as an "original image." I suppose that was necessary to admit a lot of JPEG-only equipment, but it feels self-contradictory. (Not with any practical detrimental effect. It just seems like waffling.) With a camera limited to JPEG or shooting JPEG by choice, there is an actual "primary image." It doesn't make it out of the camera, but it exists as an image file, if only briefly. I suppose the camera maker's engineers would have to speak to the issue of documenting the image processing. They seem to be playing loose and free with the compression issue, requiring uncompressed storage one one hand and tacitly admitting compressed camera files on the other. Or am I reading all this wrongly? Are they actually requiring RAW or TIFF or another uncompressed or lossless compressed format out of the camera, but not JPEG? Naw! (I'll bet they wanted to, though.)

I note, too, that it speak only to actually capturing the in-situ impression by photography. It does not at all speak to scanning or photographing lifts and gives no standard or advice for those things.

[Neville]

If my memory is still work at all, doesn't all this 1000dpi go back to how the capture stations for ten prints operated back in the 90's. Our old AFIS used a video camera that photographed each finger on the form one at a time; the resulting image was about a 100k file not much good for most things. Other systems used flatbed scanners at relatively low scanning rates, a little better. But with the thought of storing forms (cards) digitally and not hard copies, i.e. live scan images plus the move to 3rd level detail the thought that 1000dpi was required as a minimum requirement was born, this at a time when providers could sell 1000dpi scanners and at a price much lower than a 1000dpi scanning cameras. I think you will find good sales talk and a recognition that there was a need for better by our admin, and availability of cheap technologies, decisions were made for good reasons by people who may not have the knowledge we have today as general staff and that we are stuck with to day.

I think the 1000dpi figure related to ten prints and flat beds and not to SLRs and in the field of latents, but somehow the ten print decisions got included into the field work. I maybe completely wrong and I am happy to be corrected but that is the way I suspect it came about.

There was and still is (to a lesser degree now) the arguments about film v digital etc which may have had some influence also.

The thing is these things should only ever be taken as guide lines not the gospel truth; they should be living documents that can be applied to your working environment as required to suit your requirements. Not as hard and fast rules that must be complied with at all costs. Common sense is something that seems to going out the door at a very fast rate it's almost non-PC to use the term 'common sense'.

[Gerald Clough]

I have posted the following comments to SWGFAST


Comment on Draft
Standard for Friction Ridge Digital Imaging

Commenting on:

3.1 Friction ridge impressions to be used for comparison purposes shall be captured (color or
grayscale) at a minimum of 1000 ppi optical resolution when the image is sized 1:1. Justification
for any deviation from this standard shall be documented on a case-by-case basis.

Commentary

Although I am far from being a physicist or even an optics tech, I have tried to accumulate some basic knowledge of issues that help me transition from film imaging to digital. I noted the sparse treatment of resolution in the current Guidelines and was pleased to see the clarification by adding the qualification of 1:1, which at least corrected the possibility that one might imagine a sensor density of 1,000ppi was sufficient in all a cases. However, from my understanding of resolution issues, the term "optical resolution" is not an appropriate standard and is in absolute conflict with the use of pixels per inch. I will address first the conflict and difficulty with optical resolution and then proceed to a possible meaningful standard.

In the scope of the various resolution issues we might want to address, pixels per inch is a valid unit of measurement ONLY when applied to displays or sensors. It is specifically not a measure of optical resolution, a term that is in common scientific usage and is specifically an effect of system resolution. System resolution measures the resolving power of a combination of lens and sensor and display. (We may safely ignore other components, such as atmospheric, for our purpose.) Optical resolution may be expressed in either of two ways, MTF (reallly CTF) being the most commonly encountered. Both involve complex concepts and considerable sophistication and are of little use in a standard to be interpreted by those for whom the study of resolving systems are not routine.

We are not, for the purpose of this Standard, dealing with the display component. Therefore, neither optical nor system resolution is appropriate. And, at any rate, ppi is not a measure of either and is therefore meaningless. This is complicated by discussions of "optical resolution" in scanners. Hardly a good thing for a Standard. So, we still seek something that expresses a meaningful standard that may be meaningfully applied by practitioners. I think we must abandon the pixel references. Pixel counts merely place upper limits on resolution. The problem is that optical or system resolution seems to suit what we are after, but they are too difficult to apply and cast practitioners into defending something they are not trained to handle. The bottom line is that "pixels per inch" is merely a sensor density and in isolation means absolutely nothing in terms of what detail can be captured. Adding the 1:1 qualifier does not change that.

What we are most interested in is what we could define as the ability of a system to capture an image and resolve some minimum detail in the target. While it is true that it requires four pixels to discriminate two line pairs, the pixel number alone, even when specified that the reproduction ratio is 1:1 is insufficient. The sensor is merely one component. We may, however, use a lines per inch standard. What is nice about a line pairs per inch standard is that it may be reasonably tested by any practitioner. The ISO-12233 target is widely available, including adaptations designed for digital still camera systems.

http://www.aig-imaging.com/mm5/PDF/ISO- ... _v1-01.pdf

The user may read resolution by direct examination of the recorded image and be assured that their system meets any minimum standard specified in lines or line pairs per inch. And the practitioner can measure the resolving power of a system at any reproduction ratio. Such a standard would be both meaningful and something that can be locally evaluated by any practitioner, with no additional aids but the chart, and can do so under exactly the conditions under which the work will be done. It automatically takes into account the whole system, sensor and lens, and the testing even provides a means of checking the ability of the display or printer (and their associated drivers) to resolve captured detail. As a bonus, the users gets something that evaluates the system for some artifacts, such as aliasing, a useful thing when comparisons are being done on printed or displayed images that have been compressed or otherwise processed. There are other charts available that can be used for similar effect. At any rate, even anyone who is confident that their system exceed the minimum standard but is concerned about the possibility of being challenged on their compliance with a standard can easily demonstrate compliance.


Recommendation

It is quite easy to restate the kind of resolution we want in these terms. We know that modern lenses can generally resolve more finely than what I presume was intended in the draft. In that case, it seems to me the draft presumed the image projected onto the sensor by that a finely resolving lens would be adequately captured at a density of 1000ppi. This converts to a spacial resolution of 250 line pairs per inch (lpi) or 500 lines per inch. The standard may be restated as " a minimum of 500 lines per inch spacial resolution" or "250 line pairs per inch." No reference to 1:1 is needed, since spacial resolution refers to the ability to resolve lines on the target. Such a standard eliminates any presumptions about lens resolution, since the actual system of lens and camera may e tested. This is a more appropriate standard, given that latent print photography situations are not always addressable by 1:1 capture. It matters not, so long as the system can meet the standard of resolving the target. While this discussion with, which you have been patient enough to read, has been a bit long, restating the standard in these terms is no more elaborate than the draft and has the great virtue of being easily understood in very concrete terms of target and image. I urge the inclusion of "spacial resolution," because an Internet search on that term will find for a practitioner an explanation of this rather simple concept. (Try searching "optical resolution" for comparison.) I think the best reason for making this change is that we do want practitioners to understand and apply the standard. If they cannot, they simply retreat to a "my camera captures images I can work with" position. I would also consider perhaps appending a brief discussion of spacial resolution in those concrete terms, making reference to the availability of standard targets. Considering the change from Guidelines to Standard, a subject I know is the product of considerable discussion, we had better make it something we can all really apply.

Thank you for your consideration.

[Steve Everist]

Gerald,

Did you just send your comments? The three drafts currently listed were to be discussed at the last meeting in September. The final versions, resulting from that meeting, should be coming out some time soon. The web page hasn't been updated since that meeting regarding the new drafts coming up for comment.

Back in June I posted these drafts: http://clpex.com/phpBB/viewtopic.php?f=2&t=919

[antonroland]

Gerald, just to clarify one thing...

The reference to 1000 ppi and 1:1 in close proximity to one another confuses me greatly. I cannot reconcile the two as 1:1 (to me) means the magnification of the subject matter on sensor as resolved by a proper macro lens. The image is thus "life-size" on the sensor.

Most D-SLR bodies in the market today have APS-C size sensors which give an effective field of view crop and perspective of 1.6 times the actual focal length of the lens

This gives a sensor size of 22mm x 14mm which is way too small to contain a single latent fingerprint, let alone a part of a palm.

The so-called "full frame" 35mm D-SLR cameras can reproduce a single latent fingerprint 1:1 but what about palms?

Am I misreading something or what? I cannot see the correlation between 1:1 and 1000 ppi

[Gerald Clough]

I figure they were trying to clarify that it wasn't just a matter of a sensor with a density of 1000ppi, but that they meant that the image of one linear inch of the target should be represented by at least 1,000 linear pixels at the sensor. Referencing 1:1 doesn't mean the sensor must be large enough to capture the entire target area of interest at optical 1:1. In fact, as you say, the sensor may well be smaller than the actual size of the target object. That, to me, means that it's true that a sensor with exactly 1000ppi would indeed have to be large enough to capture the entire target to both bring all the target into the recorded image and to meet the minimum. It just tries to set a minimum recorded pixel density in terms of actual target size. It does not imply that latent photography system must always be able to cause a 1:1 projection of the target image on the sensor. If the physical dimensions of the sensor are too small to take in a 1:1 image, a higher pixel count will be needed to meet the standard. And most anything we would likely use has enough pixel count to meet the minimum. To take an absurd case, if we had a sensor of 500ppi, we would have project a larger than 1:1 image upon it to meet the standard. Of course, if that 500ppi sensor was the physical size of typical equipment today, we would be able to capture only a small part of an impression, if we wished to meet the standard.

So, you're not misreading. It's incorrectly worded, mixing their idea with the wrong metric. They began just requiring at least 1,000ppi. That said nothing and technically admitted another absurd case in which I could photograph an impression of one finger from across the room, so long as I used that minimum sensor pixel density. By adding the 1:1, they hoped to explain what they meant. My comment addressed that by pointing out that (1) they don't mean "optical resolution," a cumbersome measure that cannot even be expressed in ppi, and (2) "spatial resolution," which can be expressed in terms of how many lines per inch a system can resolve and which both exactly conveys the idea and is something easy for a user to confirm that their system satisfies.

The confusion over 1:1 is that, while film produces a physical negative that can be said to record an image of the same physical size as the impression, a digital image exists as an image file which has no physical "size." (We could as well use a device with a single line of pixel units to scan the projected image, in which case we couldn't even pretend to say our sensor had "size.") For that reason, also, "1:1" has no real practical meaning. It's like a text file. The only "size" that applies is the number of bytes required to code the characters. It only has physical size when it's printed or displayed, and that depends on the device and how it displays the font specified in the metadata embedded in the file. We kind of got used to talking about 1:1 with film because pretty much any film we used would more than adequately handle the detail, and specifying 1:1 was at least one way to imply the kind of spatial resolution ballpark we were working in. But it doesn't mean anything in digital imaging where we don't worry in the same way about sensor size. (There are considerations, but they are different in nature from the film issues.)

In effect, you're checking for adequate resolution when you've been shooting your test shots and determining the limits of your system's ability to resolve adequately. All they're trying to do is set a minimum that can be well-defined as something other than "it's good enough to work with." It just looks to me like the way they expressed it can't do that.

Steve:

Yes. Just sent it. I was commenting on v1.0 and wasn't aware that a new draft was coming soon.

[Steve Everist]

Steve:

Yes. Just sent it. I was commenting on v1.0 and wasn't aware that a new draft was coming soon.

I think they have the final, approved, version to be posted soon. It is no longer in draft status. However there will be several new drafts for comments as many previous documents had come up for review. Hopefully they'll be posted soon.