压缩域脆弱水印算法：多视点视频编码保护

需积分: 8 122 浏览量更新于2024-07-14 收藏 6.85MB PDF 举报

"多视点视频编码的压缩域无损脆弱水印算法" 这篇研究论文主要探讨了在多视点视频编码（MVC）中应用一种无损脆弱水印算法，特别关注了基于分层B帧（Hierarchical B-picture, HBP）预测结构的编码方案。多视点视频编码是一种技术，用于捕捉并编码多个不同角度或视点的视频流，以提供立体或全景的视觉体验。HBP是MVC中常用的一种编码策略，能够有效地减少视频数据的冗余，提高压缩效率。论文提出了一种针对HBP的脆弱水印算法。脆弱水印通常用于数字媒体的完整性验证，一旦被篡改，可以立即检测到。在H.264/AVC编码标准中，B_DIRECT_16×16和B_SKIP是两种宏块类型，分别对应于DIRECT和SKIP预测模式。作者创新性地提出，在转换B_SKIP宏块为B_DIRECT_16×16宏块的过程中嵌入水印。这种方法允许在不影响视频解码的情况下，将水印嵌入到编码语法元素中，增强了水印的隐蔽性。为了确保多视点视频内容的完整性，水印的生成基于跨视图的多视点视频比特流的内容特征。这样设计的水印能够反映出原始视频的关键特性，从而在验证视频是否被篡改时提供依据。实验结果显示，该水印算法不会对视频的质量（如结构相似性指数SSIM）产生任何负面影响。与其他方法比较，该算法在保持视频质量的同时，提供了有效的脆弱水印嵌入和验证机制。这在版权保护、内容认证以及多视点视频传输的安全性方面具有重要意义。通过在压缩域进行操作，这种水印方法减少了计算复杂性，并且由于其无损特性，不会导致解码后视频质量的下降，这对于高保真度的多视点视频应用至关重要。这篇论文为多视点视频编码的保护和验证提供了一个新的解决方案，对于多媒体安全和数字版权管理领域具有一定的理论价值和实际应用潜力。未来的研究可能会进一步探索如何增强水印的鲁棒性，同时保持其脆弱性，以及如何适应不断发展的视频编码标准。

To achieve reversible data hiding, Zhao et al. [41] proposed a novel three-dimensional

histogram shifting (3D HS) scheme that utilizes stereo H.264 video as covers media. The

proposed 3D HS algorithm is used to embed data in quantized DCT (QDCT) coefficients of

MVC video. Three coefficients chosen from each embeddable block are used for hiding two

bits of information, where just one coefficient may be changed by adding 1 (at most) in most

cases. Compared with the conventional 3D HS algorithm, the proposed scheme could achieve

superior payload-distortion performance. As information is embedded on the inter-MBs that

are based on inter and inter-view predictions, the algorithm could be applied to P and B frames

in the MVC sequence. For the selected inter-MBs, the intra-frame distortion drift caused by

data hiding can be avoided. However, the inter-MBs of P/B frames other than B4 frames will

still lead to inter-frame distortion drift.

To address the drawbacks of the above algorithms, we developed a novel fragile

watermarking algorithm for H.264/AVC multiview coding in the compressed domain. In the

proposed algorithm, the hierarchical B picture (HBP) structure of MVC videos is explored and

the algorithm is lossless. To ensure the integrity of the video content, the watermark is

extracted from the features of the I/P-frame and embedded in the B-frame, which is the

primary frame type in the HBP coding structure. The algorithm is characterized by low

complexity, high payload, and low video bitstream overhead. An experimental implementation

confirmed its compatibility with the H.264/AVC video compression standard, and applicability

to both motion and still video contents with acceptable embedding capacities.

The remainder of this paper is organized as follows. Section 2 reviews the inter-prediction

modes of H.264/AVC and MVC with the HBP coding structure, and describes the proposed

watermarking scheme. Section 3 describes the experiments that were conducted to demon-

strate the performance of the proposed scheme, and Section 4 concludes the paper.

2 Motivation

The high compression efficiency of the H.264/AVC video coding standard constitutes a major

obstacle to the application of video watermarking in the compressed domain. The present

study was motivated by some restrictions that were highlighted by previous works on

watermarking for the MVC video stream. Firstly, based on the specific H.264/AVC codec

architecture, many skipped macroblocks that are widely distributed in the inter-prediction

mode may not carry any watermark data when using the current watermarking technique.

Secondly, the coded video sequence of an MVC video with a hierarchical B picture structure

has a high proportion of B-frames that are exploited for temporal and inter-view predictions,

which is required for developing a novel watermarking scheme with a high embedding

capacity instead of applying a direct extension of a 2D field scheme. In this section, we briefly

introduce the inter-prediction mode of H.264/AVC and the HBP coding structure of MVC,

with a view toward laying a foundation for the introduction of the proposed watermarking

scheme.

2.1 B_SKIP and B_DIRECT modes

As an efficient coding standard, H.264/AVC affords more flexibility for the selection of motion

compensation block sizes and shapes compared to previous standards, with the minimum size

of the luma motion compensation blocks being as small as 4 × 4 [33]. Whereas a small block is

Multimedia Tools and Applications

suitable for regions with significant details and complex motions, a larger block is suitable for

flat and homogeneous regions. Three types of frames are adopted in the H.264 encoder,

namely, I-, P- and B-frames. In an I-frame, the pixel values in a block are coded using intra-

prediction to exploit the spatial redundancies within the frame. In P- and B-frames, the inter-

prediction is used for the estimation of forward and bidirectional motions between frames to

take advantage of temporal redundancies.

In H.264/AVC, each macroblock in an inter-prediction mode can be divided into blocks of

16 × 16, 16 × 8, 8 × 16, or 8 × 8 pixels. This is referred to as macroblock partitioning. When a

macroblock is split into 8 × 8 blocks, each of these so-called sub-macroblocks can be further

partitioned into 8 × 4, 4 × 8, or 4 × 4 pixels. Figure 1 illustrates the partitioning, with each

partition corresponding to one coding mode. The H.264 encoder invokes a motion estimation

algorithm for each partition of the coding macroblock. Lagrangian rate-distortion optimization

is used for mode selection during the motion estimation process.

For further enhancement of coding efficiency, H.264/AVC introduces the DIRECT and

SKIP modes of inter-prediction for low-motion and static macroblocks [5].

DIRECT mode: This is an inter-prediction mode for a macroblock or macroblock

partition that does not require the decoding of a motion vector. Instead, the decoder

calculates the vectors based on previously coded vectors in the forward and backward

reference frame list, namely, list 0 and list 1, and uses them to carry out bi-predictive

motion compensation of the decoded residual samples. Depending on the method used to

calculate the vector of the blocks, the DIRECT inter-prediction mode can be classified as

spatial direct or temporal direct mode, and is applied to two sizes of partition blocks,

namely, 16 × 16 and 8 × 8 blocks.

SKIP mode: The SKIP prediction mode is applied to only 16 × 16 blocks and does not

require the actual encoding of the macroblock. In this mode, a simple piece of information,

which is used in the bitstream to indicate that a copy of the macroblock is available in the

reference frame, is sufficient for decoding the macroblock. The SKIP mode in H.264/AVC is

classified as P_SKIP or B_SKIP mode based on whether it is applied to a P- or B-frame. In the

case of a B_SKIP mode macroblock, the satisfaction of the following criteria is required:

(1) The best selected mode should be a 16 × 16 partition using the DIRECT prediction mode

(B_DIRECT_16 × 16).

4 sub-macroblock

partition of

4 samples

2 sub-macroblock

partition of

8 samples

2 sub-macroblock

partition of

4 samples

1 sub-macroblock

partition of

8 samples

1 macroblock

partition of

16 samples

2 macroblock

partition of

8 samples

2 macroblock

partition of

16 samples

4 macroblock

partition of

8samples

Macroblock

Partitions

Sub-macroblock

Partitions

Fig. 1 Different inter-prediction block sizes

Multimedia Tools and Applications

剩余25页未读，继续阅读

weixin_38618140

粉丝: 9
资源: 908

压缩域脆弱水印算法：多视点视频编码保护

基于多视点视频编码的差错控制算法

面向多视点视频编码的宏块级码率控制算法.doc

基于模式相关性的多视点视频编码宏块模式快速选择算法

联合多视点视频编码中的快速搜索算法分析 (2011年)

多视点视频编码的宏块级码率控制算法优化

多视点视频编码的精确码率控制算法

联合多视点视频编码中的TZSearch快速搜索算法优化分析

基于分层B帧的多视点视频编码快速运动与视差估计算法

多视点视频编码的差错控制算法研究与应用

多视点视频编码：宏块模式快速选择算法

最新资源