Whitehouse PDF v Savannah Guthrie – Edge Erase

The Whitehouse PDF shows a white layer around the birth certificate, not present on the photograph. But as Kevin Vickland has observed, the Xerox printer has an edge erase function which removes a default of 0.12” on all sides.


Obama JPEG

So I measured the extra information on the Guthrie JPEG and found it to be 12-14 pixels. The total width of the certificate was 497 and the document width was 8.5”. 14/497 = 0.20-24”  which would be the width of the erased edges at both sides.


Another myth bites the dust. There is a 0.12” erase edge found on the Embedded JPEG, which matches the expected erase edge.

So let’s double check:

On the Obama JPEG there are 32 pixels of white separating the birth certificate from the right edge. 32 pixels/2252*8.5=0.12” [NBC: Not correct… will have to revise]

6 thoughts on “Whitehouse PDF v Savannah Guthrie – Edge Erase

  1. It is a setting on Xerox WorkCentre scanners. It can be turned on and off, and the default settings can be changed. It can be set to individual user profiles. So yes, it’s automatic, but is adjustable on a case-by-case, individual user, or machine level.

  2. I am still struggling with the details…

    top and right – 32 pixels white band
    left – 24 pixels or perhaps 30/32 pixels as there is some weird band
    bottom – 39/40 pixels

    There is some careful analysis to be done here.

    Edge erase is still a bit of a mystery to me.

  3. Okay, looking at the jpeg in Paint, where I can count pixels and such (lacking a more powerful program), I think I have it figured out. The original image, at 300 dpi, was in landscape and was 3300×2550 (11″x8.5″) [dimensions in pixels unless otherwise noted]. An edge erase border of 36 pixels (.12″) was applied to all borders. Then, the jpeg was compressed. To compress the image via jpeg, the size first was normalized to an 8×8 grid. A 2 pixel border was added to the bottom (left in portrait) and a 4 pixel border was added to the right (bottom in portrait). The image was then compressed to 150 dpi by compressing 16×16 blocks into 8×8 blocks. This works in a top-down flow, so the first 32 pixels at the top (right in portrait) and left (top in portrait) are compressed to 16 pixels and remain white. The remaining four pixels at the top and left are in 8×8 blocks with actual image, so the compression blends them with the colors at the edge. The right (bottom in portrait) compresses from 40 pixels of white border to 20, with no need for blending. Finally, the bottom (left in portrait) 24 pixels remain a 12 pixel white border, while a 14 pixel strip gets blended with a two-pixel strip from the image, creating a weird-looking almost white band that’s 7+1 pixels wide. The bottom and right edges end up being half blocks (8×4 or 4×8), but because they are white, no weird effects are seen.*

    As a result, when the 150 jpeg is placed in portrait, there is a white border of 16 at the top and right, 12 at the left, and 20 at the bottom, with an additional blended border of 2 at the top and right, and 7 at the left (with no blending at the bottom).

    *A slightly different workflow is that the extra pixels were added at the time of compression from 300 dpi to 150 dpi. These would give 8×8 blocks at the edges that would compress to 4×4 blocks, no trimming needed and the result would look the same. You still need the 16×16 to 8×8 when the large blocks are available to explain the weird bands.

  4. So basic algorithm is scan image, edge erase, do (if (edge of 16×16 block is within image boundaries, compress to 8×8 block) else (add pixels to get 8×8 block, compress to 4×4 block)), next, until (end of image).

  5. Actually, there’s another possibility. The analysis software for breaking up the image into layers seems to work on 8×8 blocks, so maybe that’s when the extra pixels get added – it seems to happen while the image is at 300 dpi. Then you compress to 150 dpi, and then you compress the resulting 8×8 blocks (although these could be a combined step).

Comments are closed.