They change pixel size?
Because it's probably not clear enough to the scanner that it can threshold it. What you see there is an effective conversion from gray to opaque black if outlines can be recognized. Probably some type of adaptive thresholding like the Otsu's method.
That changes pixel size?
What do you mean 'change pixel size'?
Look at the first of two images I just posted. The black and white pixels are smaller than the grayscale pixels.
Give me some evidence that the background around the alleged doctored areas was changed too to match and then we can start talking about forgery. If you can't show me that, then how the heck you can claim the actual text was changed?
That really is the bottom line.
You can't change the 'pixel size'. The image is 300 dpi. Each pixel is 1/300 of an inch. The guy is probably looking at it on a 72 dpi display, scaled to 72 dpi.
What he calls a 'big' pixel is simply a cluster of smaller pixels with the same color. There's nothing wrong with that.
So, everyone of those large pixels is a cluster of 4 smaller pixels of the same color? Not one group of four pixels (all, by the way, in a perfect square) has a different color?
Dude, the pixels are larger. They're also a different image than the one with smaller pixels.
There's no such thing as a digital image with 'different pixel sizes' within the same image. This is basic stuff. If you're telling me you did per-pixel touchups before, you should know this by now.
I agree, if they're the same image.
So, explain why every one of the "larger" pixels happens to be a perfect square grouping of like colored pixels?
And, if you know the difference between vector and raster based imaging, you know, in vector based imaging, two different images can be scanned at different resolutions and, in the final product, appear to be constructed of different sized pixels.
The green background pattern certainly isn't part of the original do ent.
Now sure what you want me to explain. Gray scanning is only 256 variations. As I said earlier, even that's reduced to less in order to remove the background. The actual removed pixels are converted to the same color (white) with varying alpha. There's plenty of perfect 'larger' pixels on a perfect square (or even bigger areas) with the same color all around the do ent. Look no further than the white shades around the text.
Vector based imaging doesn't even have the concept of pixels. With vectors it's points, not pixels, since they're scalable to any resolution when rasterized. There's no such thing as vector based scanning.
But I think I know what you're pointing at. Scanning is strictly a raster process, with the resolution being your selection of analog to digital sample rate. Works exactly the same as digital audio conversion. What you're thinking of is scanning or rendering two raster images at different resolutions. If you want to combine them and not lose detail, then the destination image has to have the same resolution as the highest resolution source image. Then the image with the lesser resolution will be scaled to fit. Now if you use nearest-neighbor scaling and the resolution difference between the two images is exactly a power of two, then the lower resolution image will be scaled up to use 2x2 same color pixels to make a single pixel of the lower res image.
But, you would also have the same lumps on the white surrounding area, since that area is strictly based on the source text. And you can see the 'smaller' pixels also exist in that area in this image.
There's also a rather simple explanation for that phenomenon, that follows the same concept. The scanner ADC might max out to 150 dpi, but the scanner digitally upsamples and process the scan at 300 dpi (much like what's known as 'digital zoom' in digital cameras).
That would actually make 'not processed' pixels as 2x2 same color, and processed pixels look 'smaller'. Since the background removal is a threshold function from the native sensor resolution (upsampling doesn't change the actual pixel value), it make sense to do that as part of reading in from the ADC (speculation, not done my own scanner design yet, only written drivers for high end Fujitsu scanners). That's pretty consistent with what you see in that image too.
Again, actual text tampering would entail also doctoring the outline of the text as applied in the background layer. If you want a smoking gun, that's where you would start looking.
This thread belongs in the hall of fame of weak sauce and fail.
@ talking about vector and then immediately going to pixel size. You can't make that up.
You know why they call it vector, Yoni?
Do you have any idea how raster files actually work? You cannot have a raster file made up of pixels with different pixel sizes.
Anyone who has used MS Paint has done pixel by pixel editing. I'm not even sure why anyone would bring that up as some kind of credential. Its obvious there's a severe lack of understanding here of how OCR works and what raster or vector file types are.
I'e never worked directly in image processing by any means but my GIS studies have gone into great detail about raster and vector file types because of their use in cartography and GIS. I literally LOLed at the idea of a scan producing a vector file with pixels.
2 x 6 resolution?
There are technologies that compose vector and raster graphics (PDF, svg, Silverlight, Flash, etc), but I understand what you are saying.
Why so many even gave this bogus thread so many replies is insane.
ElNono, it's obvious you know quite a bit about scanned images. Still, the layers had two different scaling factors. How do you explain that.
I can tell you other concerns, but i know you will find a reason. Just like conspiracy theorists can always find something to grab on to, you will find some excuse and be an apologist for their inept showing of the birth certificate.
It comes back to this. Why not do something like link a real 600 DPI scan? they went to the effort to show he actually had a birth certificate and utterly failed. All they did was add to the conspiracy ideas.
You do realize you just admitted to being a conspiracy theorist, don't you?
The scale on the layers on that video all show 24% by 24%.
Let me guess... the video is lying!
crofl
I could swear that one layer was at 48%.
As far as your 'conspiracy theorists' or not, you'll have those even if he pulls a printed copy. Already stated that from a digital forgery standpoint, it's easier to forge a single layer do ent than it is to forge multi-layered ones.
At the end of the day, you or Yoni or whoever will believe whatever they want to believe, tbh.
Maybe it was a different video?
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
Reply With Quote
