Robust PDF Watermarking against Print–Scan Attack

. 2023 Aug 23;23(17):7365. doi: 10.3390/s23177365

Algorithm 1 Embedding algorithm

Input: Original PDF file, watermark bit sequence

w_{1}, w_{2}, \dots, w_{N}

Output: Watermarked PDF file

1:
Convert the original PDF file into a series of images $I_{1}, \dots, I_{n}$
2:
for each image I do
3:
Map $I$ from the RGB color space to the YCbCr color space, decompose the $Y$ component, mark it as $I_{Y}$ , and convert it into a binary text image $I_{b}$ .
4:
Perform character segmentation and text line information extraction.
5:
for each text line do
6:
Confirm the left and right boundaries of each text line.
7:
Calculate the spacing between each two characters and select the maximum spacing to divide the current line into two sub-blocks.
8:
end for
9:
Obtain the set of sub-blocks ${B_{i} | i = 1, 2, \dots, K, K = 2 P}$ , P is the number of text lines.
10:
for each sub-block $B_{i}$ do
11:
Calculate the number of black pixels in each row of $B_{i}$ and find the row with the highest number of black pixels, denoted as $r_{l o n g}$ .
12:
Extract the columns with black pixels from $r_{l o n g}$ and mark them as $L = l_{1}, l_{2}, \dots, l_{n}$ , $n < c_{i}$ .
13:
for each column in L do
14:
Extract all pixels of that column, find the white pixel closest to $r_{l o n g}$ , and mark it as $p = I_{b} (x, y)$ .
15:
if $w_{i}$ is “1” then
16:
When p is above $r_{l o n g}$ , flip the pixel in $I_{Y}$ with position $(x, y + 1)$ to a white pixel; When p is located below $r_{l o n g}$ , flip the pixel in $I_{Y}$ with position $(x, y - 1)$ to a white pixel.
17:
end if
18:
if $w_{i}$ is “0” then
19:
Flip the pixel at position $(x, y)$ in $I_{Y}$ to black pixels.
20:
end if
21:
end for
22:
end for
23:
Get the watermarked image $I_{YW}$ .
24:
Replace the Y channel in the YCbCr color space with $I_{YW}$ , map it back to RGB space.
25:
Save the watermarked image $I_{W}$ .
26:
end for
27:
Combine all watermarked images in the original order and convert to a PDF file with watermarks.