Content watermarking

1. A method of watermarking a video, the method comprising:

generating three or more pilot signals;

generating a watermark;

embedding the three or more of the pilot signals in a first set of frames of the video according to a pilot just-noticeable-difference (JND) model; and

embedding the watermark in a second set of frames of the video according to a watermarking just-noticeable-difference (JND) model, the second set of frames being non-overlapping with the first set of frames, wherein:

embedding the watermark comprises embedding the watermark in a Y luminance component of each of the second set of frames; and

embedding the watermark in a Y luminance component of each of the second set of frames comprises, for each frame in the second set of frames;

partitioning the corresponding frame into 8×8 blocks B(i,j), where 0≦i≦m, 0≦j≦n;

generating an 8×8 pseudorandom block B r (x,y) having elements x, y=0, 1, . . . , 7 with a pseudo normal distribution N(0, 1);

applying a discrete cosine transform (DCT) to each block B(i,j) of the corresponding frame to generate a corresponding DCT block C(x,v); and

embedding the watermark in coefficients of each DCT block C(x,y) to generate a corresponding watermarked DCT block C w (x,y) according to an equation C w (x,y)=C(x,y)+s w ·W(i,j)·B r (x,y)·JND (x,y), where s w is a non-negative number representing watermarking strength, W(i,j) is a corresponding block of the watermark, and JND w (x,y) is a formula representing the watermarking JND model, and

JND w (x,y)=max{C To (x,y), C To (x,y)|(C(x,y)/C To (x,y)| ε }, where C To (x,y)=Tf(x,y) x (DC/C 0 ) a , Tf(x,y) is half of a quantization step-size of C(x,y), DC is a discrete cosine coefficient, C 0 is a global mean of a luminance for 8-bit images, a=0.649 and ε=0.33.

2. The method of claim 1 , wherein the watermark comprises a two-dimensional barcode.

3. The method of claim 1 , further comprising, prior to embedding the watermark or any of the pilot signals, determining a message zone within the frames of the video for embedding a watermark message zone of the watermark and the pilot signals, the message zone excluding a perimeter area of the frames of the video.

4. The method of claim 1 , wherein the pilot JND model is based on any one or more of: a luminance adaptation factor, a texture adaptation factor, and a temporal adaptation factor.

5. The method of claim 1 , wherein:

the three or more pilot signals comprise five pilot signals including a center pilot signal and first, second, third, and fourth pilot signals defining corners of a rectangle centered on the center pilot signal;

the first set of frames of the video includes four non-overlapping subsets of frames including first, second, third, and fourth subsets of frames; and

embedding the three or more of the pilot signals in the first set of frames of the video comprises:

embedding the center pilot signal and the first pilot signal in the first subset of frames exclusive of the other pilot signals;

embedding the center pilot signal and the second pilot signal in the second subset of frames exclusive of the other pilot signals;

embedding the center pilot signal and the third pilot signal in the third subset of frames exclusive of the other pilot signals; and

embedding the center pilot signal and the fourth pilot signal in the fourth subset of frames exclusive of the other pilot signals.

6. The method of claim 1 , further comprising splitting the video into an audio stream and a video stream, the video stream including the first and second sets of frames, wherein the pilot signals and the watermark are embedded in the video stream exclusive of the audio stream.

7. A non-transitory computer-readable medium having computer instructions stored thereon that are executable by a processing device to perform the method of claim 1 .