Motion Compensation based Adaptive Video Watermarking in Compressed Domain

Many video watermarking algorithms have been developed over the years to provide authenticity of video contents. The major challenges were to maintain the perpetual quality of the video and also sustaining the imperceptibility of watermark to Human Visible Systems (HVS). With rapid growth of online streaming and other video applications compression of video contents became very much essential for saving bandwidth as well as storage space. But most of these compressions are lossy and hamper embedded watermark. The most common watermarking threat is MPEG-4/H.264 codec that highly reduces the size of video content based upon motion prediction and quantization. The objective of this thesis is to present a robust watermarking technique in compressed domain that can withstand different attacks. To sustain watermark after H.264 conversion residual frames are chosen to embed data in the proposed method. Besides this, to maintain quality of video frames watermark embedding is done based on an adaptive value, derived from respective sub-blocks, that varies from block to block. Most of the watermarking techniques developed so far embed watermarks in each frame of the video sequence without using any property of video. For this reason, motion prediction is used to generate the residual frames that actually count on H.264 format. Pixel values vary from frame to frame and that is why an adaptive value is deduced from each macro-block. This adaptive value is used as threshold for embedding watermark in those blocks. This makes the watermark much more sustainable for each block. Finally the proposed technique is tested on different benchmark videos and robustness is tested against different attacks like Gaussian Noise addition, salt and pepper noise and