UPDATE I may have been too hasty in claiming to have been in error. For the moment, disregard my mea culpa below and read my later comments.

This morning I started to look more into the font/kerning issue which techdude illustrated in a number of animated GIF files, and in doing so stumbled upon something, well… something rather interesting. The Michele COLB appears to have some sort of aspect ratio distortion.

I took a cue from elliewyatt and decided to start by matching the green basketweave pattern of the background.

I pulled in the Michele COLB, and then overlaid the Obama COLB and set it to about 50% transparency. I zoomed in on the center of the image and used that point to line up the pattern. However I noticed that as I scrolled up and down and left and right, the patterns didn’t match up nearly as well at all.

I found this strange as the Michele COLB I had was a 300 DPI image, same as the Obama image. So I decided to brute force it. I separately adjusted the image width and height until I had a good alignment of the pattern over the entire page.

Here’s an image of the alignment left to right:

borderwidth3.jpg

Then I lined up the lower left corner of the Obama border to the lower left corner of the Michele border.

And you know what?

Even though the width of each of the four legs that make up the whole border were the same, and the overall height of the border was the same, the overall width of the Michele border was significantly greater than than the Obama border.

borderwidth4.jpg

What’s up with that?

But of course! The Obama COLB is a horrible forgery. The dim-witted forger who couldn’t even get the border right also screwd up the aspect ratio. Right?

Well, I’d thought of that so I figured I’d check to see if the aspect ratio problem was in the Obama image or the Michele image.

The only other known COLB I’m aware of which still retains enough of the background pattern to be useful is the DeCosta COLB. And since the change in aspect ratio is pretty much entirely horizontal, the DeCosta COLB will do just fine (it was apparently scanned without putting the lid down on it so the folds are still very slightly folded which screws up the vertical image).

The background pattern of the DeCosta COLB is faint enough as it is, and after making it semi-transparent in order to see the underlying Obama image at the same time, well, you really can’t see much of anything. So I took a different approach.

I took the opaque DeCosta COLB and laid hash marks on each side of each of the left and right vertical borders as well as on each pair of vertical bits in the background pattern.

borderwidth.jpg

Next I converted all the lines into a single group element, copied them and overlaid them on the Michele COLB image, sizing it so that the pairs of vertical bits were lined up across the width of the page. Then I copied that and moved it up a bit and centered the left vertical border between the two left hashmarks.

borderwidth2.jpg

The results were the same as with the Obama image. The overall width of the border in the Michele image was significantly wider.

Ok. What else could explain this?

I believe someone made mention of Hawaii having changed their paper sometime between 2003 and 2006. Ok. Maybe the aspect ratio of the background patern had changed and that’s only making it seem as if there’s something wrong with the aspect ratio as all of this has been focused on the background pattern.

Thought of that too.

If the aspect ratio of the printing on the background had changed, then one would expect everything else to be copasetic.

To check for this, I drew a line running from the lower left corner of the border to the upper right corner of the border. In the lower left corner of the border, I adjusted the angle of the line to match the angle of one of the lines running from lower left to upper right. Once I had the angle properly adjusted, I copied the line I’d just set and copied it several times over the next several lines.

If all that had changed was the aspect ratio of the background pattern on the paper, then these lines should line up with their counterparts in the upper right corner.

Here is the lower left corner:

aspectcheck1a.jpg

And here is the upper right corner:

aspectcheck1b.jpg

As you can see, they do not line up, showing that it’s not any change in the aspect ratio of the background pattern, but that the aspect ratio of the image itself is out of whack. Compressing the image horizontally by the same amount of the offset shown in borderwidth4.jpg brings the upper right lines into proper alignment as shown here:

aspectcheck1c.jpg

And just for comparison, here is the same test applied to the unaltered Obama image:

aspectcheck2a.jpg

aspectcheck2b.jpg

They line up as they should.

By the way, I tried the same thing on the DeCosta COLB. There was some misalignment, but that was expected owing to the vertical aspect ratio being slightly shortened due to the document having not been pressed flat when it was scanned. However the relative misalignment wasn’t quite as bad as it is in the Michele image.

Ok, so now how did the aspect ratio of the Michele image get so messed up?

Well, the image I’d been working with was the image I pulled off Polarik’s photbucket blog. Maybe he boogered it up somehow.

So I grabbed techdude’s 2007 overlay image (crappy as it may be) and ran the same test.

aspectcheck3a.jpg

aspectcheck3b.jpg

No dice.

So I think I’ve pretty well established that the aspect ratio of the Michele COLB is messed up and this has absolutely nothing to do with Hawaii’s OHSM. I mean, what are the chances that the aspect ratio of the paper’s background pattern would change and that the aspect ratio of certificate information they laser print on that paper would change by exactly the same amount?

No, it’s definitely the aspect ratio of the image itself that’s messed up.

Some of you may be saying by now “What’s the big deal? Michele screwed up when she scanned the COLB. So what?”

I’m glad you asked.

Here’s the so what.

AND PLEASE PAY CLOSE ATTENTION TO WHAT I SAY NEXT BECAUSE IT IS THE MOST SALIENT POINT I HAVE TO MAKE IN THIS POST.

I’ll start by quoting from techdude’s report over on Atlas Shrugs:

Thanks to several individuals I managed to collect and review multiple images of certificates issued between 2001 through 2003 all of which bore an identical layout to the previous Decosta image which was issued during the same time frame.

Ok. These are the images that have been out there for a while now. These aren’t really used by techdude to prove the Obama COLB “a horrible forgery.”

Next there’s this:

Several more certificate images and physical certificates were also sent to me of certificates issued between 2006 through 2008 directly by their respective owners. All of the 2006 through 2008 certificates bore an identical layout to one another.

These are the certificates of interest. These are the certificates that are anonymous, and we only have techdude’s word that they came as both images and hardcopies from several different individuals. And it is these which techdude uses to prove the “horrible forgery.”

Ok, now pay close attention to this:

The post-2006 COLBs were then compared against one another for a direct 1:1 comparison. Using copies of the images I digitally enhanced and modified the scans by removing only the hatch pattern background and then removing the merged information fields leaving just the raw document templates and saving them as a series of digital overlay templates. When the 2006 overlay was placed on top of the 2007 image they matched from corner to corner with some minor variations on the minute angle of the images. The fonts were observed to be in the same locations and of the same size and kerning. The procedure was then used with the 2006 overlay on top of the 2008 image. Once again, they matched from corner to corner with some minor variations on the minute angle of the images. The fonts were observed to be in the same locations and of the same size, style, and kerning. The 2007 overlay was then applied on top of the 2008 image. The 2007 and 2008 also matched from corner to corner with some minor variations on the minute angle of the images. The fonts were observed to be in the same locations and of the same size, style, and kerning. Having verified that all of the examined post-2006 certificates were identical in form and substance…

You get that?

The 2006, 2007 and 2008 images were all overlaid on each other and according to techdude were essentially a perfect match. He makes no mention at all of even having to size them, let alone changing their aspect ratios. They all perfectly overlay each other with the exception of “some minor variations on the minute angle of the images.”

Now, the 2008 image is the Michele image, which I think I’ve amply shown has a messed up aspect ratio. So now how was this possible? How did these other two COLBs, supposedly from two completely different people, one of which apparently was in hardcopy form which techdude would have had to scan himself, all come to overlay perfectly on an image with a fucked up aspect ratio?

I mean think about that. Three different people, with three different scanners, all of which are so crappy that they screw up the aspect ratio by a not insignificant amount, and that amount is exactly the same with all three scanners.

What are the odds of that?

Now, either I’m going to end up falling flat on my face and looking stupid because I overlooked something simple and obvious, or I smell a rat.

I noted before that the Obama image, save for its funky borders, was quite consistent with all the previous COLBs whose provenance isn’t really questionable. Whereas the images used to prove the Obama image “a horrible forgery” had aspects to them which were consistent with each other, but not consistent with any of the other known COLBs.

And now apparently there’s yet another aspect about them which is consistent among them, but not consistent with any of the other known COLBs. A mess up aspect ratio.

I also noted before that all three of these COLBs are anonymous and sourced through either TexasDarlin or techdude.

Interesting that.

TexasDarlin.
TechDude.

Oh well, time for bed.

se

Advertisements