Barack Obama Birth Certificate Image Tampering Analysis, Redux

The thread that would not die. Apparently "OpenDNA" has now made a template birth certificate you can download and make your own Hawaiian Certificate of Live Birth for fun and profit. It's somewhat fashionable now to question the ancestry of all of the images -- is the Obama certificate a fake made from the OpenDNA one?

Update 11/02/08: Obama's Birth Certificate Verified By State of Hawaii

Well, I don't think anyone really believes that, or if they do, they're watching too much Fox News. But here's an easy, step by step way to reveal to yourself the family tree of the documents.

  1. Because the 800x781 OpenDNA image is larger than the 585x575BarackObama one, go get the largest, the 2427x2369 KOS Certificate.
  2. Load the KOS image into Photoshop, any recent version. Resize to 800 wide (which will naturally make it exactly 781 high).
  3. Load the OpenDNA image. Use Image/Adjustments/Invert (CTRL+I on Windows) to make OpenDNA a negative. Select/All (CTRL+A). Edit/Copy (CTRL+C).
  4. Switch to the KOS image. Edit/Paste (CTRL+V).
  5. On the Layer Palette (Window/Layers (F7) if not already open), click the top layer (which is the OpenDNA negative) and change the layer opacity (not fill!) from 100% to 50%.
  6. Profit!!

You now have a difference between the two. Anything that isn't identical is no longer canceled by the 50% blend of the positive and negative versions. Before we interpret, we'll boost it a little more for visibility.

  1. Layer/Flatten Image (necessary so we adjust the contrast of the combined whole, not just one layer)
  2. Image/Adjustments/Brightness / Contrast. Adjust Contrast to 95 and hit ok.

    Several things stand out in order of significance. First, the obvious alterations -- the removal of the various pieces of text and the black censor block -- these show up as high-difference white -- these are the noticable-to-the-eye differences. You can mostly ignore these, we knew about them.

    Second, there is a general amount of difference all over, anywhere there is detail, especially higher-contrast detail. There is a slight "emboss" look that suggests maybe our alignment of the images isn't perfect -- the OpenDNA image may have been cropped slightly before resizing down, or it was resized with a different algorithm that what we used. It's not really important.

    Thirdly, and the part we're looking for, is that everywhere we know text was removed, there is a smudgy gunk. This is where OpenDNA copied over some material from elsewhere, and blended it into the surroundings while covering the text. It looks good to the naked eye, but since we have the digital files to compares, we can pinpoint the changes. It's interesting to note that in places he had to extend the coverup quite a ways to make the patterns match up ok. Witness how far to the right of "CAUCASIAN" the gunk goes, and how far below "HONOLULU".

    If OpenDNA had been the parent, then only the text additions would show up. The patterns would otherwise be unchanged. But, this comparison reveals the underlying fingerprints of the copy & paste coverup.

    There is another method for detecting copy & paste detection using DCT matching:
    http://www.hackerfactor.com/blog/index.php?/authors/1-Dr-Neal-Krawetz/P3...

    I don't have that algorithm handy to test, but it might trip up a bit on the repetitive pattern in this document.

    AttachmentSize
    HI_birthcert_delta.png1.09 MB