Machine-learned Video Watermarking via Frequency Domain Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video watermarking technologies often result in perceptible watermarks that negatively affect user experience and are easily detectable, making them ineffective against piracy.
Innovation Solution
Employing machine-learned models for imperceptible message watermarking by processing video data and message vectors through transformation and embedding portions to generate spatial-temporal encoding data, utilizing discriminator and decoding models to adjust parameters for imperceptibility, and incorporating a video distortion model to enhance robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional watermarking methods are used to embed messages in video data, then watermarking functionality is achieved, but the watermarks become perceptible and negatively affect user experience
Solution Approach 1:
The patent transforms the watermark embedding process by changing the domain parameters from spatial to frequency domain using Fourier transforms. This parameter transformation allows the watermark to be embedded in frequency coefficients rather than spatial pixels, making it imperceptible while maintaining detectability through spectral analysis
Solution Approach 2:
The patent replaces traditional mechanical watermark embedding methods with machine learning-based spectral embedding. Instead of direct spatial manipulation, the system uses neural networks to learn optimal frequency domain embedding strategies, substituting mechanical approaches with intelligent adaptive systems
2Reliability
If traditional watermarking methods are used to embed messages in video data, then watermarking functionality is achieved, but the watermarks become easily detectable and vulnerable to piracy
Solution Approach 1:
The patent replaces traditional spatial domain watermark embedding with machine learning-based frequency domain embedding. This substitution creates watermarks that are imperceptible in spatial views but detectable through spectral analysis, fundamentally changing the detection paradigm from visual to frequency-based
Solution Approach 2:
The patent moves the watermark embedding from the spatial dimension to the frequency dimension. By transforming the problem into the frequency domain using Fourier transforms, the system embeds watermarks in a different dimensional space where they remain hidden from spatial perception but accessible through frequency analysis
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
Systems and methods of the present disclosure are directed to a computing system. The computing system can obtain a message vector and video data comprising a plurality of video frames. The computing system can process the input video with a transformation portion of a machine-learned watermark encoding model to obtain a three-dimensional feature encoding of the input video. The computing system can process the three-dimensional feature encoding of the input video and the message vector with an embedding portion of the machine-learned watermark encoding model to obtain spatial-temporal watermark encoding data descriptive of the message vector. The computing system can generate encoded video data comprising a plurality of encoded video frames, wherein at least one of the plurality of encoded video frames includes the spatial-temporal watermark encoding data.