Video Frame Watermarking via CABAC Slice Re-encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The stateful nature of the Context-based Adaptive Binary Arithmetic Coding (CABAC) entropy encoding system in H.264 and H.265 video encoders makes it challenging to embed watermarks into video streams without resetting the CABAC engine, which typically occurs only at the slice level, and creating multiple video versions that converge to the same bitstream before reset.
Innovation Solution
The method involves selecting spatially distinct regions of a video frame, decoding and modifying macroblocks, and re-encoding them to create additional output-slices that can replace existing slices, allowing for watermark embedding and efficient re-slicing to produce multiple video versions that share most of their encoded representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If the CABAC entropy encoding system is used to encode video streams, then encoding efficiency and compression ratio are improved, but the stateful nature of CABAC makes it difficult to embed watermarks without resetting the engine
Solution Approach 1:
The video stream is divided into multiple slices, and the CABAC engine is reset at slice boundaries. This segmentation allows watermark embedding to occur in specific slice regions without affecting the entire video stream, resolving the conflict between maintaining CABAC compression efficiency and enabling watermark embedding by creating independent encoding segments where watermarks can be inserted.
2Reliability
If multiple versions of video sections are broadcast for watermark embedding, then watermark detection capability is improved, but computational power and security are compromised
Solution Approach 1:
Watermark embedding is applied locally to specific regions within video slices rather than to entire video sections. This allows multiple watermark versions to be embedded in different spatial regions, improving detection capability while reducing the computational burden compared to processing entire video sections, as only localized areas require multiple version generation.
3Adaptability or versatility
If the CABAC engine is reset frequently to enable watermark embedding, then watermark embedding flexibility is improved, but encoding efficiency deteriorates
Solution Approach 1:
The video is segmented into slices with CABAC reset at slice boundaries, providing regular opportunities for watermark embedding while maintaining efficient encoding within each slice. This segmentation strategy balances encoding efficiency (by limiting reset frequency to natural slice boundaries) with watermark embedding flexibility (by ensuring resets occur frequently enough to allow embedding in various locations).
Solution Approach 2:
The positions where watermarks will be embedded are predetermined based on slice boundaries and regional divisions. This preliminary planning allows the encoding process to efficiently proceed without frequent unexpected resets, as the watermark embedding locations are pre-coordinated with the slice structure, thereby maintaining encoding efficiency while ensuring embedding flexibility.
Data Source
AI summary
In one embodiment of a method, system, and apparatus described herein, an encoded video is received. At least one set of areas of a spatially distinct region of a video frame to be modified is selected. An entropy layer from the spatially distinct region of a video frame. A state of the removed entropy layer and the location of a first area of the set of areas of the spatially distinct region of the video frame is stored. One area of the set of areas is decoded, the decoding occurring recursively until all areas of the set of areas of the spatially distinct region of the video frame for predicting and decoding the one area are decoded. A modification of the decoded one area is performed. The modified decoded one area is re-encoded according to a re-encoding condition. The stored state of the entropy layer is retrieved, and the entropy layer is re-applied at the first area to the spatially distinct region of the video frame which comprises the set of areas. Related methods, systems and apparatus are also described.