Adaptive Watermark Embedding in DCT-SVD Domain for Raw Video

In this paper, we propose an adaptive watermark embedding algorithm operating in the discrete cosine transform (DCT) and singular value decomposition (SVD) domain for raw video. The core methodology involves segmenting video streams into distinct scenes using scene change detection algorithms, followed by transforming each video frame within scenes into the DCT-SVD domain through mathematical transformations. The watermark is quantized and embedded into the maximum singular values using quantization index modulation techniques, enabling blind detection without requiring original video content. Implementation typically involves calculating DCT coefficients using fast Fourier transform-based approaches, performing SVD decomposition through numerical computing libraries, and applying adaptive thresholding for quantization steps. Experimental validation confirms the algorithm simultaneously satisfies transparency requirements and demonstrates robustness against multiple attacks, including low-pass filtering, median filtering, salt-and-pepper noise, Gaussian noise, H.264 compression, statistical analysis attacks, and frame cropping operations.