Efficient Video Watermarking Algorithm Based on Convolutional Neural Networks with Entropy-Based Information Mapper.

Area of Science:

• Digital signal processing
• Computer vision
• Information security

Background:

• Digital video content requires robust protection against unauthorized use and piracy.
• Existing watermarking methods often face trade-offs between transparency, robustness, and capacity.
• The need for secure and imperceptible data embedding in video signals is critical.

Purpose of the Study:

• To develop a transparent and robust video watermarking method with high embedding capacity.
• To utilize deep neural networks for embedding watermarks in the luminance channel.
• To implement an information mapper for transforming binary signatures into embedded watermarks.

Main Methods:

• Deep neural networks were employed to embed watermarks in the luminance channel (YUV color space).
• An information mapper transformed multi-bit binary signatures, reflecting entropy, into embedded watermarks.
• Experiments were conducted on 256x256 pixel video frames, testing capacities from 4 to 16,384 bits.

Main Results:

• The proposed method demonstrated transparency, validated by SSIM and PSNR metrics.
• Robustness was confirmed through the bit error rate (BER) analysis.
• High watermarking capacity was achieved, suitable for diverse applications.

Conclusions:

• The developed deep neural network-based video watermarking technique offers a promising solution for secure data embedding.
• The information mapper effectively enhances watermark embedding capacity and robustness.
• The method provides a balance between transparency, robustness, and high capacity for video signals.