|
|
No. 18769
>>15942
>From having multiple images, you can piece together a more or less perfect text mask. From having a solid bright area on the picture where you can see the text, you can figure out how many bits darker the text is.
It's not quite that simple. In the bottom left corner of the pic above, the background is perfectly white (i.e. RGB(255,255,255)) and the mask takes it to RGB(248,248,248). In non-white places, though, the mask doesn't seem to subtract (7,7,7) from the RGB (though it's hard to tell precisely since there aren't any perfectly flat colored, non-white areas that I can find). Presumably what's being targeted directly is the hue or brightness or...(?), and the RGB values are adjusted in a certain way to make that happen.
>Then it's a matter of using the mask to add the exact same amount of bits back.
Though I only have OP's example to look at, I'm guessing the mask is the same size and in the same position on each picture (evidence: if you consider the vertical whitespace to belong to the text *below* it, the watermark starts basically flush with the top-left corner, and is chopped off on the right, suggesting it wasn't rescaled to fit this particular image). Once you figured out how exactly the watermark changes the RGB values and had a b&w mask with just the watermark in it, it'd be straightforward to write a program that takes the mask and de-applies it to a whole directory full of images automatically, thus dramatically reducing the amount of time spent dicking around in photoshop.
|
