Hermitian has posted 5 long comments, several of which were placed in a queue for approval because they contained too many links. Since I want to give Hermitian the same level of visibility as my own blog postings I will present his claims in this posting
Fair is fair. Enjoy. I will let be intermixing my comments over time in red
His task is to show that his ‘forgery hypothesis’ explains the observed artifacts better than my hypothesized work flow. If I can show that a simple workflow can do the same as what he claims his forger has been doing, then the conclusion is simple: The forger is purely algorithmic because of the work flow. Now that Hermitian finally has shared with us his preferred hypothesis we can do an actual comparison.
Hermitian’s hypothesis is competing with the following.
- The original long form birth certificate was scanned in, upside down, on a Xerox Work Centre, most likely 7655 because it has been shown that the White House owns such a device which produces the same finger prints as found in the WH LFBC PDF
- The Xerox Work Centre was instructed to use the standard profile for constructing a document and email itThe email was opened in Preview and it was found that the scan was upside down
- This results in:
- Initial scan of 600 DPI
- Separation into a jpeg background and multiple monochrome foregrounds
- Alignment of two boundaries with 8×8 bit offsets
- Alignment of two boundaries with internal object
- Downsampling the background to 150 DPI
- Downsampling the foregrounds to 300 DPI
- Saving the JPEG with a quality factor of 47.48% with specific Quantization Matrix
- Embedding a YCbCr comment into the generated JPEG
- JBIG2 compression
- Images are all in landscape direction
- The PDF was rotated in preview 180 degrees
- The PDF was printed to PDF (possibly with scale to fit set)
- This results in:
- PDF version is changed from 1.x to 1.3
- JBIG2 objects are now encoded as FlateDecode monochrome bitmaps
- ObjStm objects are decomposed into multiple XObjects
- /Rotate tag is replaced with cm level rotations
- Metadata shows Preview creator/producer information
- A clipping mask is added
- This PDF was made available electronically
Forgery Part 1
As previously reported on this thread, I have devised a method for simultaneously re-creating the nine bitmap images, each having the same page-size as the original nine images that initially existed on the forger’s Macintosh computer (before he placed them within the final White House LFCOLB PDF document). These nine recreated image files include one JPEG file for the Green background image and eight PNG files for the eight monochrome layers.
[NBC: This presumes that the forger created landscape oriented (sideways) PNG and JPEG files, embedded the right Comment String and matched the Quantization Matrix found in the Xerox generated PDF. Worse, it fails to understand the other observations: Why the bitmaps all align at 8×8 bit boundaries on two sides and with the internal objects on the two other sides. It also fails to explain why the JPEG was created at half the resolution of the Monochrome Bitmaps, or why they were given different color values. When creating a Xerox PDF, all these artifacts are found in the resulting PDF. Furthermore, it fails to explain why some text is in the JPEG and some text is in the foreground, something we do observe in the Xerox created PDF as part of a simple work flow.
And while Hermitian does a lot of hand waving, he has yet to show that he can follow in the footsteps of the forger and recreate all the details found in the PDF. I have shown how a simple workflow can create all these features. Hermitian? Not so much. A bit of hand waving but nothing to support his position.]
During the re-creation, each of the nine new images is automatically rotated 90 degrees counterclockwise relative to their final orientations within the LFCOLB PDF image. The page-size of the re-created JPEG image of the background layer was automatically increased by a scale ratio of 208.3333%. The image size of the eight monochrome layers was scaled up by a ratio of 416.6666%. The actual page-size of the background layer in the LFCOLB PDF file measures 8.5067 in. X 11.0133 in.
[NBC: Hermitian does not explain why the layer is slightly larger than the canvas, again this is something we also observe in the Xerox workflow. So again, we compare an ‘ad hoc’ explanation with the consistent outcome of a Xerox workflow. Why would the images not map exactly in the mediabox? ]
The larger page-size of the re-created JPEG background image is 35.4444 in. X 22.94444 in. The size of the major text image in the composite LFCOLB PDF image measures 6.0633 in. X 4.8467 in. The size of its larger re-created image is 20.1990 in. X 25.2696 in.
The size of each of the re-created images for the eight monochrome layers was automatically scaled up by a factor 200% greater than the scale factor of the background layer.
[NBC: No explanation is given as to the ‘automatic’ scaling of the foreground layers. Again, this follows trivially from the workflow I outlined where the background is stored as 150 DPI and the foreground as 300 DPI bitmaps.]
This was confirmed by scaling the composite WH LFCOLB PDF image by a factor of 416.6666%. The links panel data for the enlarged image of the WH LFCOLB indicated 200% scaling for the background layer and 100% for each of the non-background layers. Hence at the scale factor that re-creates the original size of the non-background layers the background layer is twice its re-created size. Simply stated, 2 = 416.6666% / 208.3333%.
[NBC: Impressive math but it fails to explain the why of these numbers, Again, they are trivially explained by the workflow I identified and verified. The simple explanation is that the foreground layers were subsampled from 600 to 300 PPI and the background layer from 600 to 150 PPI. No need to invent excessively large canvases. ]
This 200% difference in scale factors between the background and the monochrome layers was intentionally chosen by the forger to achieve a final difference in pixel size and resolution between the background and monochrome layers.
[NBC: So it was done to mimic the output from a Xerox workflow? In other words, you have no reason and fail to explain why the forger wanted to achieve a difference in pixel size]
The final resolution of these layers in the LFCOLB PDF image is 150 PPI X 150 PPI for the background and 300 PPI X 300 PPI for the monochrome layers.
[NBC: Exactly as I stated and exactly as is found in the Xerox workflow.]
The differences in pixel size and resolution were entirely caused by the 200% difference in scale factor between the background and non-background images.
[NBC: Yes, an outcome of the MRC step which saves the background as 150 DPI and the foreground at 300 DPI. This has again been verified]
Based on my analysis of and comparison between the nine re-created images and their respective nine embedded images within the WH LFCOLB PDF file, I conclude that the final PDF file was entirely created by a human using a computer graphics program installed on a Macintosh computer. Based on the facts uncovered by my research the WH LFCOLB is a clear forgery. There is no evidence that this PDF image was created by scanning a single certified paper copy of the purported Obama original hospital generated Hawaii birth certificate. Moreover, the WH LFCOLB PDF image did not result from a digital scan of any single original Hawaii Certificate of Live Birth.
[NBC: And yet I can show a scan of a WH LFCOLB document which is very similar in nature and which, once run through a simple work flow, shows all these features, and explains all the scaling and rotations and artifacts in a trivial manner. I have to conclude that so far, Hermitian has managed to identify the forger to be indeed the Xerox WorkCenter and not a human. Since his hypothesis is competing with my hypothesis which recreates all the features he has described, and more, it clearly remains the winner. Hermitian has, unwittingly, described the exact steps that take place when placing a document on a Xerox Workcenter (upside down) and scanning it, emailing it and rotating the PDF in preview and exporting it as a PDF. His workflow fails to explain the following
- The embedded comment
- The Matrix Quantization being identical
- Thhe 8×8 bit alignment and the alignment with the internal object for the two remaining sides
So in addition to being non ad-hoc, my workflow explains more. Clearly, so far we have to conclude that Hermitian’s hypothesis has been beaten by the workflow I identified. Worse, he has failed to explain why a forger would separate the text into a foreground and background layer, why one layer is JPEG and the other Monochrome Bitmaps, and yet my workflow explains all this.
Verdict so far:
Forger 0 Workflow 1
Using his graphics program and Macintosh computer the forger assembled his nine created images into a single PDF document. As he placed each image (in turn) he applied a size reduction and a rotation. The background image was scaled by a factor of 48% and the eight non-background images were scaled by 24%.He rotated each image clockwise by 90 degrees. The forger manipulated the images to hide the fact that he had individually created each of the nine images (within a graphics program on his computer) and then placed them within a single PDF file.
[NBC: In other words, Hermitian admits that the forger recreates the same as would be expected from a Xerox workflow, with failures to explain some salient parts, which is why we have to conclude once again that his hypothesis fails.]
After the forger placed, rotated, and re-sized each of the nine images within the WH LFCOLB PDF file, he then embedded all nine images. Only embedded images are found listed in the links panel when the WH LFCOLB PDF file is opened in Adobe Illustrator. By embedding the images, the forger eliminated the links to his nine externally created image files that initially resided on his Macintosh computer. This was a deliberate action by the forger to erase the links data from the links panel. This act prevents any investigator from using the file links data to trace the nine image files to the forger’s computer.
[NBC: And yet, the same is created by our Xerox forger, which has been verified using an actual Xerox scanned file that was saved in Preview. Again, Hermetian has shown that our forger is a Xerox Workcenter workflow, saved by Preview.]
Most of the typed text and signatures are contained within the largest monochrome (mostly text) layer image. All of the text contained within this image is binary and monochrome. The text within the Green background image is Grayscale. The monochrome binary text is aliased and the Grayscale text is anti-aliased.
[NBC: This again is explained by a simple Xerox Workcenter workflow. All these ‘artifacts’ are trivially explained yet Hermitian fails to explain why a forger would recreate a document exactly like a WorkCenter would. In other words, Hermitian fully embraces the forger to be the workflow I identified]
The pixel size of each of the nine re-created images is 1/72 in. X 1/72 in. Thus the pixel resolution of each re-created image is 72 PPI X 72 PPI. This resolution is the same as the native screen resolution for the original Macintosh computer.
[NBC: So how come he created 300 and 150 PPI images 🙂 Worse, 72 PPI is the standard settings for PDF on any machine.]
Based on these preliminary findings, I checked the other PDF images of the Obama LFCOLB that have been produced in several court filings and by the Associated Press. In every case the final LFCOLB image had been reduced from a larger size in which the pixel size was 1/72 in. X 1/72 in. (resolution = 72 PPI X 72 PPI). One advantage of this resolution for the original Macintosh was that the graphical image was replicated in actual size if printed to a Postscript printer.
The total number of pixels within each of the nine images is the same between the two different page-sizes for each of the nine images. The differences between each PDF image and its re-created, enlarged image is simply a result of the difference in page-size between the two images. The relationship between the different resolutions of the two background layer images is 150 PPI = 72 PPI / 0.48 PPI. Likewise, the resolutions for the eight monochrome layers are related by 300 PPI = 72 PPI / 0.24.
[NBC: Yes, these trivially flow from the Xerox Workflow I identified, and yet they do not flow rationally by claiming a human forger. Again, Hermitian is embracing the workflow as the forger]
Importantly, for typewriting text, each character of a particular font is assigned a size measured in points. The size of each character is equal to its width expressed as a specific number of points. One point is defined as 1/72 inches. Thus there are 72 points per inch of text containing a single character. Thus each character of a 72 point font is one inch wide.
[NBC: Wow, imagine that… These are some math skills. Okay a bit snarky but why was the text created in 300 DPI and partially in 150 DPI? Hermitian does not tell, and yet it follows trivially from the work center work flow I identified.]
The forger set the screen resolution (within his graphics program as it resided on his Macintosh computer) to 72 PPI. X 72 PPI. He set this resolution for a reason. His reason was to obtain a one-to-one correspondence between points per inch (ppi) and screen resolution in pixels per inch (PPI) for the eight non-background layers. There are 72 points per inch and also 72 pixels occupying the same inch on the forger’s screen. Thus for each non-background image displayed within his graphics program there is exactly one point per pixel.
[NBC and yet he created 300 PPI bitmaps and 150 PPI JPEGs… Weird…]
[Conclusions: Hermitian’s hypothesis has failed to explain several salient points which are trivially explained by the workflow hypothesis, in fact, he has adapted the fact that the forger made it look exactly like the Xerox workflow, and while this is always a logical possibility, he effectively has identified the forger to be the workflow I identified. It is worse since he has failed to explain various scaling factors, and several features.
Forger 0 Work Flow 1
Forgery Part 2
[NBC: The font issue is incidental to my work flow, however Hermitian does raise good questions that need asnwers
The answer appears to be straightforward:
1. The font best matches the pica font
Which means that 13.9 letters per inch actually were 10 letters per inch or pica font.
Kalanianaole is 12 characters and should at 10 letters per inch be 1.2” but the actually measured length is 0.85 or about 71%.
So how do I know that the document was reduced by ~70-72%. Some investigative work is necessary.
I compared it to the picture provided here
I count 337 pixels for the width of the box “signature of local registrar’
And 510 pixels in the WH PDF @150ppi or about 245 pixels
This means that the WH PDF is scaled by 72.7% according to this calculation
So in the end the WH PDF is also 10 characters/inch
That the WH LFBC showed the original LFBC scaled should have been expected since it takes up ~75% of the full width.
The problem is that Hermitian did not take into account that the text had been scaled during the copy process.
The calculations so far are somewhat inexact but I guess the point should come across that when measuring characters per inch, you need to know if the copy had been scaled and there are signs that it was.
This is also true for the re-created background image because it is also displayed at 72 PPI resolution on the forger’s screen. However, because of the 200% difference in scale factor between the background and non-background images, a 20 point character typed on a 72 PPI non-background layer would be equal in size to a 10 point character typed at 72 PPI on the re-created background image after the two images were placed within the LFCOLB PDF image (and then reduced in size by 24% and 48% respectively).
Consequently, for each of the nine re-created images, there are 72 pixels and 72 points filling each inch of the image, (provided that the inch is occupied by one text character).
All standard manual key-stroke typewriters (of 1961 vintage) always typed equally spaced characters on the page. Thus, only monospaced typewriter fonts can emulate the equal spacing of typewritten text.
[NBC: Actually, it is well known that typewriters, certainly old ones, do not necessarily create monospaced letters because of the wear and tear in the mechanisms. The only font which one would expect to consistently be monospaced would indeed be ‘fixed fonts’ found on computers. When you compare texts created on older typewriters, you will invariably see characters shifted both horizontally and vertically]
Monospaced fonts are often referred to as “fixed” fonts. The most popular character pitches available on 1961 vintage manual typewriters were 10 and 12 characters / inch. The points versus pitch battle has been most frequently fought in the application of the typewriter to screenwriting. The Courier New 12 font (10 point – 12 cpi) developed by IBM in 1955 soon became the standard font for typing screenplays. Courier New 12 was also the U.S. State Department’s standard typeface until January 2004, when it was replaced with 14 point Times New Roman. The Courier Standard font developed in the mid 1900′s was (12 point – 10 cpi).
The standard pitch for the Prestige Elite typeface of 1961 vintage manual typewriters was 12 characters per inch. These Elite typeface were described as 10 points wide. Obviously, this has to be an “assigned” point size because the “effective” point size for a 12 pitch fixed font is 6 points. Likewise the other standard typewriter pitch of 10 characters per inch was assigned to be a 12 point type face. However, the “effective point size is only 7.2 points = (1/10 inches per character) / 1/72 inches per point) = 72/10. The confusion probably arises from the different character size designations from typography for the same type face.
However, we can rule out the Prestige Elite typewriter font because it doesn’t match the font of the typed text on the LFCOLB PDF composite image. This leaves the Courier or Courier New monospaced fonts as strong candidates for the forger’s 1961 vintage typewriter font. The pitch of the Courier New monospaced font is also 12 characters to the inch. The New Courier font has evolved into the standard Courier font in currently available word processing programs.
The original Courier Standard type was created by IBM and released as a non-proprietary font in the 1950′s. It is still available from Adobe, Linotype, Monotype Imaging and others.
The classification of the size of the typeface for typewriters has been a subject of much confusion. Much of the confusion arises from the different conventions used in typesetting (typography) and typewriting. However, the differences between monospaced (fixed spacing) type versus variable space type also is a source of confusion. The kerning used by all modern word processing software was not possible on 1961 vintage fixed space manual typewriters.
[NBC: Of course fixed space manual typewriters do not support kerning. However a diligent researcher would have found that IBM “announced proportional letter spacing for typewriters in 1941, but IBM’s World War II effort delayed the introduction of a typewriter model, the Executive, with this capability until 1944.”
“In English-speaking countries, the commonplace typewriters printing fixed-width characters were standardized to print six horizontal lines per vertical inch, and had either of two variants of character width, called “pica” for ten characters per horizontal inch and “elite” for twelve. This differs from the use of these terms in printing, where they refer to the height of the characters on the page (“pica” making for ten horizontal lines per vertical inch).”
The character height measurement relates to typographical standards which are different than typewriter typeface standards. Typewriter fonts are sized according to their width rather than by their height. Typographic fonts are sized by their height rather than by their width. To add to the confusion typographical fonts are most often sized in “Picas” which is another size measure applied to type. One Pica is defined to be equal to 1/72 of a foot. However, typographical fonts are also sometimes sized in points which are defined as 1/72 of an inch. Hence the confusion arises.
Surprisingly, the measured height of the binary, monochrome (purportedly) typewritten capital characters on the re-created primary text layer is 0.375 in. This is the nominal height of each character. Each character is aliased with numerous random edge pixels which give each character a “Porcupine” appearance. The characters are thus 27 pixels in height for the typed text. The character height is most easily measured on the capital “B” and “I” characters for the two mostly text layer images. The final height of the same character in the final LFCOLB PDF image is thus 0.09 in. = 0.24 X .375 = 27 pixels X 1/300 PPI. Thus the height of each character spans the same number of pixels (i.e.27) in the final PDF image as in the re-created image.
The point size of text characters is less important than character pitch for monospaced typewriter fonts. For example a 12 pitch monospaced typewriter font would have an “effective” point size of 6 points = (1/12 inches per character) / (1/72 inches per point) = 72/12.
The problem arises because the forger fixed the point size of characters for his larger image rather than the character pitch size in the final image. This indicates that the forger was familiar with digital word processing but not familiar with 1961 vintage manual typewriters.
The nominal measured width of the capital letters of the re-created image of the mostly text layer is 20 pixels. This corresponds to a 20 point font in the larger re-created mostly text layer image. Thus the nominal character width is 20 pixels for the forger’s enlarged image of the mostly text layer. The nominal width spans 20 pixels in both the enlarged and reduced LFCOLB PDF images of the mostly text layer. The inch dimensions corresponding to a character width of 20 pixels are 0.27778 in. = 20 pixels / 72 PPI for the larger re-created image and 0.06667 in. = 20 pixels / 300 PPI for the LFCOLB PDF images — both images of the mostly text layer image. However, this point size would equate to a pitch of 15 characters per inch for monospaced type instead of 12.
Again this shows that the forger was relatively young and had no knowledge or experience with typewriting on a manual typewriter.
Forgery Part 3
Facts on the courier font can be found at the following links:
“Everything You Ever Wanted to Know About Courier . . . And Then Some”
“By Roland Stroud”
For a comparison of the four most popular Courier Fonts see:
“Which Courier Is the Best Courier?”
Many varieties of Courier fonts are listed here:
The proof of my conjecture about the forger’s motive is the actual measured character pitch of the typed characters within the two-different-size-images of the mostly text layer. The 20 point font of the larger image does not translate to a 10 point font in the smaller PDF image.
If we assume that the forger intended that his typewritten text would emulate a 12 pitch typewriter font in the final LFCOLB PDF image, then we find that the 20 point font that the forger selected for the larger 72 PPI image does not translate to a 10 point font in the smaller PDF image. Instead we calculate 4.8 points = 20 points X .24, much smaller than 10.
However, a 12 pitch fixed font has an effective character width that is smaller than 10 points.
The effective character width for a 12 pitch typewriter is easily calculated to be 6 points = (1 in. /12 characters) X 72 points per inch. The forger’s character width of 4.8 points is thus smaller than the effective width.
Proof of this comes from a measurement of the character pitch of the mostly text layer of the WH LFCOLB. The character count for the typewritten word “Kalanianaole” (on the mostly text layer within the LFCOLB PDF image) is 12 characters. The actual measured length of this word is 0.85 in. rather than the expected 1.0 in. for a character pitch of 12 characters per inch. This length was measured on the WH LFCOLB PDF composite image in Adobe Illustrator CS6.
As an additional check, I printed two B&W copies of the WH LFCOLB PDF image and measured the length of the word “Kalanianaole” from the paper copies. Black and White printouts were produced from Adobe Illustrator CS6 and Adobe Reader XI. The measured length of the word “Kalanianaole” was 0.85 in. for both B&W printed copies.
The 0.85 word length is to be expected because 0.85 in. X 6./4.8 = 1.06.
Consequently, the forger either picked the wrong size computer font (for emulating a 12-pitch typewriter font) or else he intentionally set the size of the typed characters (on the mostly text layer PDF image) to be smaller than the typewriter font of the typed original hospital generated birth certificate (assuming that it was typed on a 12-pitch typewriter). The actual measured pitch of the word “Kalanianaole” is 12 characters / 0.85 in = 14.1 characters per inch.
Another possibility is that the forger selected a computer typewriter font which matches the character size of the original birth certificate but which has a pitch greater than 12 characters per inch. However, I’m not aware of any 1961 vintage typewriter fonts with a pitch of 14 characters per inch.
[NBC: In fact, there is at least one: The IBM Selectric whose Pica font also matches the shape of the characters. But that’s not really important here.]
This leaves only the one explanation that the forger intentionally set the page-size of the WH LFCOLB PDF image to be smaller than the original WH hospital-generated birth certificate. In that case, it cannot be a duplicate copy of the original hospital-generated birth certificate or a duplicate copy of one of Obama’s two certified copies of his LFCOLB. The certified copies presumably would be duplicate copies of the original LFCOLB because they were (purportedly) produced by Xerography (i.e. dry photocopy process) by the HDOH.
The width of the typed capital letters within the mostly text layer is 0.06667 in. = 20P / 300 PPI whereas the width of a 6 point character is 6 p / 72 ppi = 0.08333 in. Thus we find that the point-size of the typed text on the mostly text layer is less than the expected width for a pitch of 12. Thus the actual pitch must be greater than 12 characters per inch.
[NBC: That is of course not the only solution. Scaling also resolves the issue, and so much nicer… It starts with the observation that the DOH of Hawaii reduced the size of the original document when it copied it onto the security paper. Anyone can understand that if you scale, you will give the impression that the point size has changed. I find it somewhat surprising that Hermitian overlooked this.]
In order to match the size of typed characters between the background and monochrome text layers the forger used a 10 point font to type the few characters that appear on the background layer image. The 10 point fonts were typed onto the same resolution (72 PPI) Macintosh screen as were the typed 20 point characters of the eight monochrome images.
In order for the forger’s plan to work, he would have had to select a scalable typewriter font with both 10 point and 20 point sizes available. The typed text characters were most likely vector fonts typed onto the larger images at a screen resolution of 72 PPI. These nine larger images containing the typed scalable fonts were subsequently placed within the LFCOLB PDF composite image and then each of the nine placed images were scaled, rotated and then embedded as bitmap images.
[NBC: Show us how you can recreate the posted PDF using the hypothesized workflow of the forger. Remember, I have a workflow that already does all this]
OCR was not applied to the composite LFCOLB PDF image. None of the text of the composite LFCOLB PDF image is selectable, searchable or editable when the PDF file is opened in Adobe Reader.
Part 4 removed since it adds little to the forgery claim. I will present it in a separate posting
Forgery Part 5
There is no evidence that any type of file-size optimization (compression) was applied to the Obama LFCOLB PDF file except for its relatively small file-size of 380 KB compared to the file-size of bitmap images which are typically several MB.
[NBC: ROTFL… What do you call JPEG if not file-size optimization… My goodness sakes… That by itself explains a good chunk of the compression. We also see how the document is separated into foreground and background layers, the foreground layers compressed using ZLIB. Come on Hermitian, this is basic stuff.]]
However it is important here to report the sum total file-size for the nine re-created images from my devised method of image file re-creation. As previously stated the nine re-created files consist of a JPEG file for the background layer and eight PNG files for the other non-background files. All of these nine re-created image files are bitmap files. The size of the JPEG background file is only 271 KB.
[NBC: That is in its compressed form. You do understand the JPEG concept?]
The summed total size of the other eight PNG files is only 113 KB for a total summed file size of 384 KB. Thus the sum total of the file-sizes of the nine re-created images is only 4 KB larger than the final size of the WH LFCOLB PDF file with the nine linked files embedded as bitmap files.
Consequently, we must conclude that any file-size compression would have been applied to the nine separate image files and not to the composite WH LFCOLB PDF file as posted on the WH website. This would have been most easily accomplished if the nine image files were first placed in individual PDF wrappers.
[NBC: Huh? The easiest way to do this is to take the document and scan it using a Xerox Work Center. It explains the major reduction in file size, the compression of the JPEG as well as of the individual layers. This is getting ridiculous, there is no evidence to support Hermetian hypothesis that a forger was involved and I have shown how a simple workflow explains in all.
Verdict so far:
Forger 0 Workflow 2
Support for this conclusion is the fact that the forger ultimately embedded the nine image files as bitmap files within the LFCOLB PDF file. It makes no sense for him to have embedded the linked files in the PDF file if he had already optimized the PDF file-size before embedding the nine linked images. Embedding the linked files often doubles the PDF file size. Thus the only reason for the forger to embed the nine image layers would be to cover his tracks. This is the case if the file optimization was applied to the LFCOLB PDF file before the linked files were embedded.. Thus it would be nonsensical for the forger to re-optimize the final WH LFCOLB after he embedded the nine linked files because he could have achieved a smaller PDF file-size by just not embedding the nine image layers. Consequently, I conclude that the forger embedded the linked image files simply to erase the link data for these files.
As evidence for this conclusion, I flattened the WH LFCOLB PDF file in Adobe Photoshop CS6. Attempts to flatten the file in Adobe Illustrator CS6 and Adobe Acrobat XI Pro were unsuccessful. The flattened file was saved as a Photoshop CS6 PDF file. The resulting image was created at 300 PPI X 300 PPI resolution. The image quality was excellent. However, the file size increased from 380 KB to 17.86 MB. The flattened file was then opened in Adobe Acrobat XI Pro and the file-size optimizer was applied. At the “Smallest Size” setting of the slide lever, the file size was reduced from 17.86 MB to 116 KB. This file size is smaller than the WH LFCOLB file size of 380 KB. The resolution of the image was reduced from 300 PPI X 300 PPI to 100 PPI X 100 PPI. The image quality was degraded but useable. Another attempt was made to optimize the flattened WH LFCOLB PDF in Adobe Acrobat XI Pro. For this second trial the slide lever was set to “High Quality”. The resulting image was created at a resolution of 300 PPI X 300 PPI. The file size was reduced from 17.86 MB to 810 KB. Thus there is no reason for the forger not to flatten the image to reduce file size.
[NBC: Again you are undermining your own hypothesis here]
Additionally, it makes no sense that a file-size optimization algorithm would segment a scanned bitmap image from a 8.5 in. X 11. in. sheet and then enlarge each segment by a different scale factor just to change the resolution of each segment to 72 PPI X 72 PPI and then rotate each segment 90 degrees counterclockwise and then import each segment into a common PDF file and then reduce their size back to the size on the scanned paper image and then rotate them 90 degrees clockwise before placing them. And after all that, then the optimizer would embed the nine images but not flatten the PDF file.
[NBC: You really do not understand how this all works now do you. This is getting pretty ridiculous especially since I have shown how a real workflow does exactly this. Don’t let your ignorance as to why or how stop you from following where the facts lead you.]
So much for using file-size optimization to explain all the image anomalies that have been identified by the many experts (and researchers) who have examined the composite LFCOLB image contained within the WH LFCOLB PDF file.
[NBC: And thus, using his ignorance Hermitian rejects the simple facts that a Xerox Workcenter/Preview workflow explains all this, and we can explain the major reduction in file size… This is just getting plain silly. Hermetian claims that software would never do this because he does not understand why it would and when I show how software does indeed do this, he just ignores the facts. This is one of the strongest examples of confirmation bias I have ever seen.
Verdict so far:
Forger 0 Workflow 3
the Forger created the Obama LFCOLB by digital “cut and paste”.
The forger first acquired various examples of original Hawaii paper birth certificates and other vintage 1961 documents which contained the words and characters to be transferred to the Obama LFCOLB PDF image. He then digitally scanned each document to files residing on his Macintosh computer. From each digitally scanned document, he then used his graphics editing program to digitally “clip out” sections from each document which contained the form text, form lines signatures, and stamp impressions needed to construct the Obama LFCOLB PDF composite image.
The forger then placed the selected digital clippings taken from the scanned images of Hawaii birth certificates and other documents onto the screen of his computer (i.e. within the window of his computer graphics program). He then adjusted the angular position of each clipping to align the baseline of the lines of text parallel to the horizontal screen gridlines. Alternatively, he could have aligned the form lines from the clippings to be parallel with the X and Y gridlines.
Then, using the zoom tool, he sized each clipping to achieve a pixel size which corresponds to the screen resolution of 72 PPI. He then re-saved each clipping after each was aligned and sized to the screen within his graphics program.
Using a digital cut and paste methodology, the forger then separately assembled each of the nine images from selected form text, form lines, signatures and stamp impressions digitally excised from his screen compatible digital clippings. The typewriter text was typed directly onto both the background and non-background images. The completed nine images were then cut out (as described below) as rectangular images and then rotated 90 degrees counterclockwise.
Each of the nine images was then placed within the multilayer (composite) LFCOLB PDF document. As each separate image was placed within the LFCOLB PDF image, the image size was reduced and each image was rotated by an angle of 90 degrees clockwise. The background image was reduced in size by a scale factor of 48%. Each of the non-background layers was reduced in size by a scale factor of 24%. The pixel resolution of each image was automatically increased in proportion to the reciprocal of each scale factor. The pixel resolution of the background image was increased from 72 PPI to 150 PPI. The pixel resolution of each of the non-background images was increased from 72 PPI to 300 PPI.
Based on the actual location and dimensions of each of the placed nine rectangular images within the LFCOLB PDF composite image, I conclude that the forger set the grid for his computer graphics program as follows. First he set the screen resolution to 72 PPI X 72 PPI. Then he chose settings for the orthogonal grid so that the grid would be congruent with the screen pixels. The screen pixels were 1/72 in. X 1/72 in. For each of the nine original images. Thus, as displayed on the forger’s Macintosh computer the major grid lines were set at a spacing of 1/9 in. = 0.1111111111 in. The minor grid spacing was selected for 8 minor divisions between each pair of major grid lines. With this grid setup, the lines were exactly congruent to the four edges of each square pixel of each of the nine re-created images.
Other details of the method that the forger used to align and position each of the nine rectangular images within the composite PDF image were previously posted on this same thread.
[NBC: I would love to see Hermitian use the same tools to recreate the effects seen in the WH PDF, but we already know that a simple workflow with a scanner and preview, can create all these artifacts with no need to invoke an ad hoc forger. There is hand waving, lack of specifics, and no explanation of the JPEG comment and the Quantization matrix, no attempt to explain why the images are stored sideways in landscape form, no attempt to explain much of anything.
I conclude that Hermitian has totally failed in his endeavor. Given the wasted time on the Font size and Applewhite, I find his arguments to be begging the question, ignoring the facts as we know it.
In the introduction I explained that Hermitian has to explain at least as much as I can explain using my workflow hypothesis, Furthermore, he cannot just invoke a forger that acted in a manner indistinguishable from my work flow without adding to rationale as to why the forger did it the way he did it. Many of the steps hypothesized by Hermitian fail to capture the essentials such as the 150/300PPI images, which he claims were created in a 72 PPI setting. Totally hilarious.]