逢
甲
大
學
資訊工程學系
博 士 論 文

H.264 / AVC影像資料隱藏之研究
Novel Data Hiding Techniques for
H.264/AVC Video

指導教授：許芳榮
研 究 生：阮庭戰

中 華 民 國 一 百 零 八 年 六 月




Novel Data Hiding Techniques for H.264/AVC Video

Acknowledgement
This dissertation provides my research at Feng-Chia University, Taichung, Taiwan.
It is an opportunity to give thanksgiving to persons who help me in this period.
First, I would like to thank my advisor, Professor Fang-Rong Hsu, for his
invaluable advices, helpful guidance, and supports throughout every stage of PhD study.
Without Professor Hsu, my study would not be finished.
I also would like to give thanks to Professors and staffs of Department of
Information Engineering and Computer Science, Feng Chia University, especially, YaJane Chen for her supports. I appreciate the support and friendship of my best friends
in the Bioinformation Lab, my Vietnamese colleagues and friends.
The unceasing attention, encouragements and supports of my parents, my parentsin-law, my wife and my children are motivations for my studying in Taiwan.

Especially, my wife Phuong-Quynh Mai, my children Hoang-Long Nguyen and MaiPhuong Nguyen are my great sources of love for scholarly endeavors. I want to give
them all my pride.

i

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

摘要
資訊隱藏技術透過將秘密資訊載入到以多媒體為載體來安全地傳送秘密資訊，
已經為視頻序列提出了許多資訊隱藏技術。然而，通過現有這些方案僅能獲得有
限的載入容量。在本研究中，我們提出一項創新的數據隱藏技術，以已經提出進
一步提高載入容量，同時保證載入式視頻 H.264 / AVC 序列的高視覺質量。
技術 1 提出一種用於在 H.264 / AVC 視頻中隱藏 DNA（脫氧核糖核酸）序列的
算法。首先在隱藏之前透過規則對 DNA 序列進行加密，然後將加密的 DNA 序列載
入到內部框架（I 階）的 4×4 區塊（MB）的量化離散餘弦變換（QDCT）係數中。
為避免失真偏移，我們使用 I 階預測的方向。透過滿足-1,0 和 1 的所有成對係
數，不需要使用任何閾值來限制載入數據的容量。實驗結果證明該演算法在載入
容量方面優於其他現有演算法。
隱藏叢集、隱藏叢和防止叢集確認兩兩相異的群集。然後，當防止叢集用於避
免視頻序列中的失真偏移，隱藏叢載入秘密資訊時出現微失真。在技術 2 中，為
了架構載入資訊建立修改方向彙整表。透過這樣，載入視頻序列中的失真盡可能
維持在極小的差距。實驗結果證明，該技術在保持良好視覺質量的同時，於載入
容量方面優於其他兩種技術。
在技術 3 中，載入群組被使用於夾帶秘密資訊，而避免組用於防止失真偏移。
實驗結果證明，該方案可以避免內部框架失真偏移，確保載入時的低失真。因此，
與先前的方案相比，所提出的該技術提更強而有力的載入容量。而且，在不需要
原始資訊的情況下，可以完全讀取載入資訊。

隱藏可逆資訊是一種在主機中載入秘密資訊的技術，例如數據庫，音頻，圖像
和視頻，但它可以恢復原始主機。透過直方圖位移方法，在技術 4 中，提出了一
種用於 H.264 / AVC 的可逆資訊隱藏演算法，其目的是使載入容量盡可能更高，
同時視頻可以更好地恢復到原始狀態。該技術還可以防止失真偏移。所有係數都
大於或等於高峰值，為了隱藏秘密資訊，如果秘密位元為 1 則增加 1，而當秘密
位元為 0 時則不增加。在接收站中，通過檢查具有高峰值（位元 0）和高峰值 +1
（位元 1）的係數，可以輕鬆地進行數據讀取。在讀取資料後，所有係數都比高
峰 +1 大，將被移回以恢復原始視頻。該技術的實驗結果證明，該技術可以恢復
ii

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

到原始視頻，進一步提高了載入容量。該演算法的缺點是它無法更正遭受網路攻
擊的錯誤位元。未來我們將使用 BCH 碼技術實現提高資訊隱藏的穩定性。

關鍵詞：隱寫術，資訊隱藏，DCT，H.264 / AVC，係數分群，內部框架，失真偏
移，載入容量，可逆資訊隱藏，直方圖位移。

iii

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

Abstract
Data hiding technique allows secret data to be delivered securely by embedding the

secret data into cover digital media. Many data hiding techniques have been proposed
for video sequences. However, the limited embedding capacity is obtained by these
techniques. In this study, novel data hiding techniques are proposed to further improve
the embedding capacity while guaranteeing the high visual quality of embedded video
H.264/AVC sequences.
Technique 1 proposes an algorithm for hiding DNA (Deoxyribonucleic acid)
sequence in H.264/AVC video. The DNA sequence is first encrypted by a rule before
hiding, and then we embed encrypted DNA sequence into the Quantized Discrete
Cosine Transform (QDCT) coefficients of the 4×4 macroblocks (MBs) of Intra-frames
(I-frames). To prevent the distortion drift, we use the directions of I-frame prediction.
By using all paired-coefficients that meet -1, 0 and 1, we do not need to use any
threshold for limiting the capacity of embedded data. The experimental results show
that the proposed algorithm has improved existing algorithms in terms of embedding
capacity.
Two different clusters, Hiding cluster and Preventing cluster, are determined. Then,
the hiding cluster is used for embedding the secret data with small distortion while the
preventing cluster is used for avoiding the distortion drift in video sequence. In
Technique 2, the Embedding Modification Direction (EMD) table is constructed for
embedding data. By doing so, the distortion in the embedded video sequence is
maintained as small as possible. Experimental results demonstrated that the proposed
technique outperforms the other two existing techniques in terms of embedding
capacity while keeping the good visual quality.
In Technique 3, the embedding group is used to carry the secret data, and the averting
group is used to prevent distortion drift. The experimental results indicated that the
proposed scheme can avoid intra-frame distortion drift, and guarantee low distortion
due to embedding. In addition, the proposed technique provides enhanced embedding
capacity compared to previous schemes. Moreover, the embedded data can be extracted
completely without the requirement of the original data.
Reversible data hiding is a technique for embedding secret data in a host, such as
iv


FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

database, audio, image, and video, but it can recover the original host. By histogram
shifting method, in the Technique 4, a reversible data hiding algorithm for H.264/AVC
is proposed with a purpose that the embedding capacity can achieve as high as possible,
simultaneously, the video can recover to the original better possible. This technique can
also prevent distortion drift.

All coefficients are greater than or equal to peak_point+1,

will first be shifted to zero_point values. For all coefficient equal to the peak_point
value, in order to conceal secret data, we increase by one if the secret bit is one, and do
not change when the secret bit is zero. In the receiver site, the data extraction is easy by
checking coefficients with peak_point values (bit 0) and peak_point+1 (bit 1). After
extraction, all coefficients are greater than peak_point +1, will be shifted back in order
to recover the original video. The experimental results of this technique show that the
proposed technique can approximately recover to the original video. By comparing with
the other techniques, the proposed technique can recover to the original video, and
further improve the embedding capacity. A disadvantage of the algorithm that cannot
correct the error bits for network attacks. So, in the future, we will use BCH syndrom
code technique for robustness of data hiding with the proposed algorithm.

Keywords: Steganography, data hiding, DCT, H.264/AVC, coefficient grouping, Intraframe, distortion drift, embedding capacity, reversible data hiding,
histogram shifting.

v


FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

Abbreviations
DCT

Discrete Cosine Transform

DNA

DeoxyriboNucleic Acid

DST

Discrete Sine Transform

EC

Embedding Capacity

EMD

Embedding Modification Direction

HEVC

High Efficiency Video Coding


HS

Histogram Shifting

IDCT

Integer Discrete Cosine Transform

IPM

Intra Prediction Mode

LSB

Least Significant Bit

MB

MacroBlock

MV

Motion Vector

MPEG

Moving Picture Experts Group

MSE


Mean Square Error

MVD

Motion Vector Difference

NAL

Network Abstraction Layer

NIH

National Institutes of Health

PSNR

Peak Signal to Noise Ratio

QCIF

Quarter Common Intermediate Format

QDCT

Quantize Discrete Cosine Transform

QP

Quality Parameter


RDH

Reversible Data Hiding

SSIM

Structural Similarity index

VOD

Video-On-Demand

vi

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

Table of Content
Acknowledgement ........................................................................................................ i
摘要

.................................................................................................................. ii

Abstract

................................................................................................................. iv


Abbreviations ............................................................................................................. vi
Table of Content ........................................................................................................ vii
List of Figures ............................................................................................................. ix
List of Tables ............................................................................................................... xi
Chapter 1 Introduction ............................................................................................... 1
Research Motivation ........................................................................................ 1
Dissertation Objectives .................................................................................... 5
Dissertation Organization ................................................................................ 5
Chapter 2 Related Works ........................................................................................... 6
Intra-frame Mode Prediction............................................................................ 6
Procedure of Embedding.................................................................................. 8
Histogram Shifting ......................................................................................... 10
Chapter 3 An Algorithm for DNA Sequence Hiding in H.264/AVC Video ........... 11
Introduction .................................................................................................... 11
The Algorithm for DNA Sequence Hiding in H.264/AVC ............................ 12
3.2.1. Macroblock prediction ....................................................................... 12
3.2.2. Case classification .............................................................................. 13
3.2.3. To-be-embedded capacity checking ................................................... 15
3.2.4. Embedding process ............................................................................ 15
3.2.5. Extraction process .............................................................................. 17
3.2.6. Encrypt and decrypt DNA sequences ................................................ 17
Experimental Results and Discussions .......................................................... 18
Chapter Summary .......................................................................................... 22
Chapter 4 A Novel Steganography Scheme for H.264/AVC Video Without
Distortion Drift ..................................................................................... 23
Introduction .................................................................................................... 23
A Novel Steganography Scheme for H.264/AVC Video ................................ 25
vii

FCU e-Theses & Dissertation (2019)



Novel Data Hiding Techniques for H.264/AVC Video

4.2.1 The proposed DCT-based EMD modulation algorithm ...................... 26
4.2.2 The proposed embedding procedure ................................................... 30
4.2.3 The proposed extracting procedure ..................................................... 32
Experimental Results and Discussion ............................................................ 34
Chapter Summary .......................................................................................... 41
Chapter 5 High Embedding Capacity Data Hiding Algorithm for H.264/AVC
Video Sequences Without Intra-frame Distortion Drift ................... 42
Introduction .................................................................................................... 42
High Embedding Capacity Data Hiding Algorithm for H.264/AVC Video ... 44
5.2.1. Intra Frame Prediction ....................................................................... 44
5.2.2. The proposed scheme ......................................................................... 47
5.2.2.1. Category selection and coefficient grouping .................................. 47
5.2.2.2. Embedding phase ............................................................................ 49
5.2.2.3. Extracting phase .............................................................................. 50
Experimental Results and Discussion ............................................................ 51
5.3.1. Embedding capacity evaluation ......................................................... 51
5.3.2. Bitrate increment ratio ....................................................................... 53
5.3.3. Visual quality evaluation .................................................................... 54
Chapter Summary .......................................................................................... 58
Chapter 6 An Algorithm for Reversible Data Hiding in H.264/AVC Video ........ 59
Introduction .................................................................................................... 59
An Algorithm for Reversible Data Hiding in H.264/AVC Video .................. 61
6.2.1. Histogram generation and shifting ..................................................... 61
6.2.2. Embedding process ............................................................................ 63
6.2.3. Extraction and recovering process ..................................................... 65
Experimental Results and Discussion ............................................................ 67

Chapter Summary .......................................................................................... 73
Chapter 7 Conclusions and Future Studies ............................................................ 74
Conclusions .................................................................................................... 74
Future Studies ................................................................................................ 75
Bibliography .............................................................................................................. 76
Publication List ......................................................................................................... 86

viii

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

List of Figures
Figure 2.1. A 4×4 luminance block and its reference pixels in four adjacent
blocks .......................................................................................................... 6
Figure 2.2. Nine 4×4 luminance prediction modes ....................................................... 7
Figure 2.3. Embedding with histogram shifting techniques ....................................... 10
Figure 3.1. The processes of proposed algorithm. ...................................................... 12
Figure 3.2. A 4×4 luminance block and its reference pixels in four adjacent
blocks. ....................................................................................................... 13
Figure 3.3. The diagram of the proposed algorithm. (a) Embedding. (b)
Extraction. ................................................................................................. 16
Figue 3.4. Structure of embedding binary string. ....................................................... 18
Figure 3.5. Visual effect of the resultant intra-frames ................................................ 21
Figure 4.1. The flowchart of our steganography algorithm ........................................ 26
Figure 4.2. An example of two constructed clusters, H and P, in Case1..................... 28
Figure 4.3. The quantized DCT coefficients. (a) before and (b) after embedding
process....................................................................................................... 30

Figure 4.4. The flowchart of embedding procedure.................................................... 31
Figure 4.5. The flowchart of extracting procedure. .................................................... 33
Figure 4.6. Visual effect on videos sequences of the proposed scheme and
scheme in [52] ........................................................................................... 39
Figure 4.7. Example of “Akiyo” for statistical attack using. (a) Chi-square
result of “Akiyo”. (b)Chi-square result of stego-video of modified
coefficients of “Akiyo”. ............................................................................ 41
Figure 5.1. The current block and the four adjacent blocks ........................................ 44
Figure 5.2. Main processes of the proposed scheme: (a) Embedding phase; (b)
Extracting phase ........................................................................................ 47
Figure 5.3. Percentage of blocks that meet the first four categories ........................... 48
Figure 5.4. Example of coefficient grouping .............................................................. 49
Figure 5.5. Embedding capacity of the proposed scheme for different QPs .............. 51
Figure 5.6. Visual quality of the proposed scheme for different QPs ......................... 54
Figure 5.7. Visual quality (PSNR) performances of the proposed scheme and
two previous schemes [51, 52] with different video sequences ................ 57
Figure 5.8. Visual quality (SSIM) performances of the proposed scheme and
ix

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

two previous schemes [51, 52] with different video sequences ................ 57
Figure 5.9. Visual effect of the resultant intra frames by the proposed scheme
and two previous schemes [51, 52] ........................................................... 58
Figure 6.1. Coefficient pairs in Case 1, Case 2 and Case 4 ........................................ 61
Figure 6.2. Illustration for the histogram shifting procedure ...................................... 62
Figure 6.3. The diagram of the embedding process. ................................................... 63

Figure 6.4. Embedding algorithm ............................................................................... 65
Figure 6.5. The diagram of extraction and recovering process. .................................. 65
Figure 6.6. Extraction and recovering algorithm ........................................................ 66
Figure 6.7. Comparing the PSNR before and after recovering video with QP=28 .... 68
Figure 6.8. Comparing the SSIM before and after recovering video with QP=28 ..... 68
Figure 6.9. PSNR of videos after embedding and extraction. .................................... 69
Figure 6.10. SSIM of videos after embedding and extraction. ................................... 69
Figure 6.11. The PSNR deviation of the proposed algorithm with two
algorithms. ................................................................................................ 72
Figure 6.12. The SSIM deviation of the proposed algorithm with two
algorithms. ................................................................................................ 72

x

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

List of Tables
Table 2.1. Three conditions of the selected modes and its corresponding
reference pixels ........................................................................................... 7
Table 2.2. Relationship between Conditions and Cases for adjacent 4×4 block
modes. ......................................................................................................... 8
Table 3.1. Modes and their Conditions in [51] ........................................................... 14
Table 3.2. Relationship between Conditions and Cases for adjacent 4×4 block
modes. ....................................................................................................... 14
Table 3.3. The capacity of to-be-embedded data (bits) ............................................... 15
Table 3.4. Comparison of the embedding capacity of proposed algorithm with
algorithms in [51] and [52]. ...................................................................... 19

Table 3.5. Comparison of the PSNR, SSIM of proposed algorithm with
algorithms in [51] and [52]. ...................................................................... 20
Table 3.6. Comparison of the PSNR, SSIM, and EC of proposed algorithm and
Ma et al.2010’s algorithm for QP=28. ...................................................... 21
Table 3.7. Comparison proposed algorithm with EC the same as that of Ma et al.
for QP=28.................................................................................................. 22
Table 4.1. Comparison of the embedding capacity of the proposed scheme and
three existing schemes [51, 52, 58] in terms of the number of bits per
4×4 macroblock. ....................................................................................... 35
Table 4.2. Comparison of embedding capacity of the proposed scheme and three
existing schemes [51, 52, 58] for QP = 28. ............................................... 35
Table 4.3. Comparison of visual quality between the proposed scheme and three
existing schemes [51, 52, 58]. ................................................................... 37
Table 4.4. Comparison of visual quality of the proposed scheme with three
existing schemes [51, 52, 58] for QP=28 with 14 sequences ................... 37
Table 4.5. Comparison of bitrate increment rate between the proposed scheme
and two schemes in [51, 52]. .................................................................... 38
Table 4.6. Execution time comparison between the proposed scheme two
existing schemes [51, 52] (milliseconds per frame). ................................ 39
Table 4.7. Performance comparison between the proposed scheme with some
xi

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

existing schemes [51, 52, 58, 62, 101] with QP =28 ................................ 40
Table 5.1. Five general categories of reference pixels and selected prediction
modes ........................................................................................................ 45

Table 5.2. Comparison of embedding capacity of the proposed scheme and two
previous schemes [51, 52] in terms of embedded bits per 44 block ....... 52
Table 5.3. Embedding capacity comparison of the proposed scheme and two
previous schemes [51, 52] for QP = 28 ..................................................... 53
Table 5.4. Comparison of bit rate increment ratio of the proposed scheme and
two previous schemes [51, 52].................................................................. 53
Table 5.5. Comparison of visual quality (PSNR: dB) of the proposed scheme
and two previous schemes [51, 52] ........................................................... 55
Table 5.6. Comparison of visual quality (PSNR: dB) of the proposed scheme
and two previous schemes [51, 52] for QP = 28 ....................................... 55
Table 5.7. Comparison of visual quality (PSNR and SSIM) of the proposed
scheme and two previous schemes [51, 52] for QP = 28 .......................... 56
Table 6.1. The capacity of to-be-embedded data ........................................................ 64
Table 6.2. Quality of videos after embed for randomly secret data bits ..................... 67
Table 6.3. Quality of videos after embedding and recovering for DNA sequences
NC_007020 (11440nts) for QP=28 ........................................................... 70
Table 6.4. Quality of videos after embedding and recovering for DNA sequences
NC_007203 (6909nts) for QP=28 ............................................................. 70
Table 6.5. Comparing the proposed algorithm with Liu et al.’s and Lin et al.’s
algorithms for QP=28 ............................................................................... 71

xii

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

Chapter 1 Introduction
Research Motivation

In the recent years, the information booming requires necessary of information
security. Beside cryptography field, hiding data into a host is also a way to ensure data
integrity when two computers communicate on the internet. In the data hiding area, the
host may be a text file, a database, an image, an audio, a DNA sequence, or a 3D mesh.
The image is a common environment for data hiding. There are many techniques to
hide secret data on digital images. These techniques may be irreversible [1-8] or
reversible [9-14] the original image. The irreversible technique is normally provided for
high embedding capacity with low distortion, such as in applications with tamper
detection [15] and image authentication [16, 17]. Irreversible techniques cannot recover
the original image after extraction the secret data. However, Reversible data hiding
(RDH) techniques have been provided in order to resolve the weakness of irreversible
techniques. To avoid attacking from malicious tools, the researchers used algorithms for
robustness data hiding [19-21]. The study in [19] increased the robustness of secret data
by using perceptual marking, but in 2017, Kim et al.’s [21] used BCH code to correct
error coding for data distribution when solving robustness problem.
Comparing with the above techniques, the substitution Least Significant Bit (LSB)
techniques are easy by reading the words in the bitstream of the image after encoding
and then substitute LSB bit [22-24]. However, this technique is usually used for hiding
information in audio files [25-27].
With the advent of 3D technology, the 3D mesh also became a good environment for
watermarking and hiding secret data. Many techniques have been proposed for
watermarking information in 3D mesh [28-31]. Yu et al. [32] presented a robust
watermarking technique for 3D mesh models by distribution information. In order to
recover 3D mesh, Jiang et al. [33] shown a scheme for reversible data hiding for
encrypted 3D mesh models. Like the 3D mesh, DNA sequences are also widely
developed in recent years. Therefore, they are quickly used for hiding secret data. Data
hiding can be covered by mapping to fake DNA sequences [34-36] or using an available
DNA sequence to carry secret data. Many techniques used virtual nucleotide bases insert
into an existing DNA sequence [37-39], but there are techniques that use codons, the set
of three nucleotide bases for translation to Amino Acid codes, to encrypt data before

1

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

embedding as in [40-43].
Digital video technology, born in the late 1970s, laid the foundation for the birth of
the MPEG (Moving Picture Expert Group) standards. In 1996, the first release of
MPEG-2 part 1 was published [44], and beginning used for embedding secret data [4547]. The study in [45] presented a watermark scheme for MPEG-2 encoded video with
averting drift problem. The data hiding technique in [46] can support user-defined levels
of security an accessibility. The degradation and distortions are also avoided in this study.
By using the Discrete Cosine Transform (DCT) block, Chae et al. [47] embedded data
in individual video frames.
In 2004, Richardson et al. [48] published a study in H.264/AVC (Advanced Video
Coding) and MPEG-4 video coding. H.264/AVC, was known the MPEG-4 part 10, is
the common standard for high video efficiency compression. By its advantages, the
H.264/AVC is a perfect environment for concealing secret data. There are many studies
research on data hiding in H.264/AVC [49-57]. The study in [53] presented a method
for concealing data in H.264/AVC streams. Because this method focuses on blind data
hiding scheme, it does not need the original host video when extracting the message. In
the data hiding for H.264/AVC, the problem is Intra-frame distortion drift. This problem
had resolved by Ma et al. [51] by exploiting the coefficient-pairs of 4 × 4 DCT
macroblock. This study is groundwork for researchers who work on data hiding in H.264
field.

The study in [51] only embedded 46% of the 4 × 4 DCT luminance blocks.

Therefore, in 2013, Lin et al. [52] had proposed a study to improve the embedding

capacity by exploiting the remaining 54% of luminance blocks while keeping the video
qualities, especially Structural Similarity (SSIM) index. In 2014, Tew et al [49] surveyed
information hiding techniques for H.264/AVC compressed domain. The study showed
the general framework for hiding secret data in video and data representation schemes.
By using codeword substitution technique, the study in [50] encrypted codewords with
stream ciphers, and then embedded data into the encrypted data. To improve data hiding
scheme in the encrypted H.264, Xu et al. [56] exploited fully CAVLC codeword for data
hiding. The study can keep the bitrate and also improve the embedding capacity.
For more robustness for data hiding in H.264, many studies [58-60] had been
proposed. In 2013, Liu et al. [59] used BCH syndrome code for encoding the secret data.
The encoded data then embedded into coefficients of 4 × 4 luminance DCT macroblocks
of intra-frames. This study not only get more robustness, but also get high quality and
avert intra-frame distortion drift. Study in [60] showed that the BCH(63,7,15) can at
2

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

least correct 99.17% error bits with quality parameter (QP) is 25 or 27.

By using

Shamir’s (t, n) secret sharing, the studies in [61-62] distributed embedded data into n
sub-secrets for more robustness. While study in [61] is a method for irreversible, the
study in [62] proposed a method for reversible data hiding (RDH) in H.264/AVC.
Many studies for the RDH in H.264/AVC were also presented in recent years. The
data can embed in the encoded video [63-75] or encrypted video [77-78]. The studies in
[63-65] presented reversible data hiding in H.264/AVC approaches for intra-frame error

concealment. Xu et al. [63] implemented the histogram shifting (HS) technique for
embedding the secret data into motion vector (MV) of a macroblock (MB), in order to
improve qualities and payload. Histogram shifting of motion vector values is also
presented in [76] for RDH algorithm. This study had a higher capacity, invisibility, and
especially it was simple algorithm. In 2012, Liu et al. [66] proposed a RDH technique
for H.264/AVC with low distortion drift. This system is not only fast and simple to
implement, but also has a high embedding capacity. By using zero QDCT coefficientpairs, in 2018, Chen et al. [75] showed a RDH in video based on H.264/AVC for
improving the embedding capacity. Studies in [77-78] used histogram shifting technique
for RDH in H.264/AVC encrypted domain. In 2016, Yao et al. [77] presented a
theoretical analysis than then embed secret data into residual coefficients after
estimation embedding distortions. The study can against packet loss of video, and make
the visual quality better. The study in [78] used intra-prediction mode (IPM) motion
vector difference (MVD) and histogram shifting technique to conceal secret data into
the watermark video. The study can recover original video, but cannot against the attacks.
To eliminate the outside attacks, some studies proposed the robustness data hiding
scheme [79-80]. Liu et al. [79] used BCH syndrome code to improve robustness with
better visual quality. This study can avoid the distortion drift of video when embedding
secret data. The study in [80] also used BCH code for encoding the grouping of data.
Watermarking is a data hiding technique. This technique was widely used to embed
information into images. Although it also can apply for concealing data in H.264/AVC
[81-89]. Qiu et al. [82] presented two schemes for watermarking in video. The fragile
watermarking applied for motion vector domain, and the robust scheme used in DCT
domain. In 2006, Nguyen et al. [85] presented a fast system for watermarking on
H.264/AVC motion vector. The computation complexity of the study was kept low.
Noorkami et al. [81] showed an algorithm for watermarking information in compressed
domain of H.264. This algorithm have low computational complexity and suitable for
3

FCU e-Theses & Dissertation (2019)



Novel Data Hiding Techniques for H.264/AVC Video

real time applications. For video on Demand (VOD) service, in 2012, He et al. [88]
proposed a real-time watermarking scheme for H.264 compressed video. This study also
has low computational complexity, and kept slight bitrate and PSNR effects. To avoid
attacks from outside, many studies focused on robustness [83-86]. Gong et al. [86]
proposed a scheme for robust watermarking in H.264/AVC video. The scheme can
achieve high visual quality, robustness, and improve computational complexity.
In order to retrieve the original video, some studies proposed the reversible
watermarking for H.264/AVC [90-91]. Study in [90] used histogram shifting technique
to embed watermark bits by modifying the last nonzero coefficients of 4×4 macroblocks
domain. The study has low computational complexity and can recover video. Farrugia
et al. [91] used decoder to retrieve the original video for reversible visible watermarking
method in H.264/AVC encoded domain. The study could get high visual quality in term
PSNR with the previous methods.
In this study, we improve existing scheme by presenting four techniques for
reversible and irreversible data hiding in H.264/AVC to get high embedding capacity
and avert distortion drift. In the Technique 1, the DNA sequence is encrypted by our
form and then embed into the zero coefficients. To reduce distortion drift, the
macroblock (MB) modes are predicted in this study. The experimental results illustrate
that the study can get high embedding capacity.
The second technique, Technique 2, constructs the embedding modification direction
table for embedding data. By doing so, the distortion in the embedded video sequence
is maintained as small as possible. Experimental results indicated that the technique
outperforms to other two existing techniques in terms of embedding capacity while
keeping good visual quality.
The third technique, Technique 3, uses the embedding group to carry the secret data,
and the averting group is used to prevent distortion drift in the adjacent blocks. The
experimental results demonstrated that the proposed scheme can avoid intra-frame

distortion drift and guarantee low distortion due to embedding. In addition, the proposed
technique provides enhanced embedding capacity compared to previous techniques.
The fourth technique, Technique 4, uses histogram shifting technique for reversible
data hiding in H.264/AVC with a purpose that the embedding capacity can achieve as
higher as possible, simultaneously, the video can recover to the original better possible.
This technique can also prevent distortion drift. The experimental results indicate that
the proposed algorithm can approximately recover to the original video. By comparing
4

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

with the other studies, the proposed study further improve the embedding capacity, and
can recover to the original video.

Dissertation Objectives
The objectives of this dissertation are briefly described as below.
1. To provide an algorithm that encrypted DNA sequences and then embeds into
H.264/AVC with high embedding capacity.
2. To propose two schemes to improve further embedding capacity, and prevent
distortion drift by exploring 4x4 luminance blocks
3. To develop a reversible data hiding algorithm for H.264/AVC by using
histogram shifting technique, improving embedding capacity.

Dissertation Organization
This dissertation is organized as followings. Chapter 2 presents some basic
information that relate to our studies. An algorithm for DNA sequence hiding in
H.264/AVC video is proposed in Chapter 3. Chapter 4 develop a novel steganography

scheme for H.264/AVC video without distortion drift. Averting intra-frame distortion
drift is also mentioned in Chapter 5 with high embedding capacity data hiding algorithm
for H.264/AVC video sequences. An algorithm for reversible data hiding in H.264/AVC
is proposed in Chapter 6. Finally, the conclusions and future works are shown in Chapter
7.

5

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

Chapter 2 Related Works
This chapter provides an overview of relationship of works, which relate to our
techniques, such as the prediction of intra-frame mode, procedure of embedding and
histogram shifting.

Intra-frame Mode Prediction
In H.264/AVC, to reduce the redundancy of intra frames, the well-known intra
prediction algorithm is used. In this algorithm, each macroblock (MB) of H.264/AVC
videos can be encoded by 16×16 or 4×4 mode. Since the human eyes are sensitive to
any modification of luminance values in 16×16 intra MBs, thus, only H.264/AVC videos
encoded by 4×4 intra mode is recently considered for embedding data. Figure 2.1 shows
an example 4×4 luminance block. As can be seen in Figure 2.1, Bri,j is a current 4×4
luminance block whose pixels are named from a to p, and its four adjacent luminance
blocks, i.e., BRi-1,j-1, BRi-1,j, BRi-1,j+1, BRi,j-1, have been decoded. These four blocks are
decompressed and the reference pixels, labeled A to M, are then used to predict the
current block Bri,j. Figure 2.2 shows a description of the intra mode decision scheme that
is used to select an optimal prediction mode from the nine modes to construct the

predicted luminance block Bpi,j[52]. Then, to encode the current 4×4 luminance block
Bri,j , the residual block Ri,j is calculated by Ri,j =Bri,j -Bpi,j which is continuously
processed by integer DCT algorithm and quantization operation to generate the
quantized DCT coefficients Rqdcti,j . Then, the encoded version of the block Bri,j is
represented by Bpi,j and Rqdcti,j. Therefore, the recovered luminance block Bri,j can be
computed as Bri,j =Ri,j + Bpi,j , where Ri,j is obtained by applying the de-quantization
operation and the inverse DCT algorithm on Rqdcti,j.

Figure 2.1. A 4×4 luminance block and its reference pixels in four adjacent blocks
6

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

Figure 2.2. Nine 4×4 luminance prediction modes
In 2010, Ma et al. [51] proposed a steganography algorithm for video H.264/AVC by
perturbing the coefficient pair of Rqdcti,j. In their algorithm, the seven pixels, i.e., d, h, l,
m, n, o and p of the recovered luminance block Bri,j are considered whether they are
used in the intra prediction algorithm for the next blocks or not. Table 2.1 shows three
conditions and its corresponding reference pixels. Then, according to the condition that
the current block belongs, three specific coefficient pairs are identified to embed three
secret bits by perturbing process.
Table 2.1. Three conditions of the selected modes and its corresponding reference
pixels
Mode name

Mode value


Reference pixels

Condition1

Right-Mode

0, 3, or7

Condition2

Under-Left-Mode &

1 or 8 and 0, 1, 2, 4, m, n, o, p

Under-Mode

5, 6, or 8

Under-Right-Mode

0, 1, 2, 3, 7, or 8

Condition3

d, h, l, p

p

To further improve embedding capacity of Ma et al.’s scheme, in 2013, Lin et al. [52]
proposed new data embedding algorithm for H.264/AVC intra-frame. Instead of using

three above-mentioned conditions to classify the current block for embedding the secret
7

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

data, Lin et al. has divided the reference pixels and the selected prediction modes of the
four adjacent blocks into five cases, i.e., Case1, Case2, Case3, Case4 and Case5, as
shown in Table 2.2. Instead of only selecting three coefficient pairs for embedding three
secret bits, to increase embedding capacity, Lin et al. designed new data embedding
algorithm to embed one more secret bit into the current block. As a consequence, the
embedding capacity has been further improved in the scheme [52]. However, in their
scheme, each pair of coefficients is modified to embed only one secret bit. It means that
the more secret bits are embedded, the more pairs of coefficients are modified. As a
result, the higher embedding capacity is obtained, the more distortion of the video will
be. Therefore, in this paper, we propose a new steganography algorithm to further
improve embedding capacity while maintaining the good visual quality of the embedded
video.
Table 2.2. Relationship between Conditions and Cases for adjacent 4×4 block modes.
Cases

Condition 1

Condition 2

Condition 3

Reference pixels


Case1

True

False

X

d, h, l, p

Case2

False

True

X

m, n, o, p

Case3

False

False

False

NRP


Case4

False

False

True

p

Case5

True

True

X

All

X - Do not care; NRP - No Reference Pixel; All - all of seven reference pixels d, h, l,
p, m, n, o

Procedure of Embedding
Integer cosine transform (ICT), a kind of Discrete Cosine Transform (DCT), is
usually used in H.264/AVC standard. Since the human eyes are less sensitive to the
brightness, we only use 4×4 luminance macroblocks to embed data, and apply the ICT
transform for 4×4 blocks, shown in (2.1).
𝑊 = 𝐶𝑓 𝑅𝐶𝑓𝑇


(2.1)

Where W is the matrix of undetermined DCT coefficients corresponding to the
1
2
residual block R4×4; CTf is transformed matrix of Cf, and 𝐶𝑓 = [
1
1

1
1
1
1 −1 −2
]
−1 −1 1
−2 2 −1

𝑄𝑃

With 𝑞𝑏𝑖𝑡𝑠 = 215+𝑓𝑙𝑜𝑜𝑟( 6 ) , and
8

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

𝑎𝑏⁄
𝑎𝑏⁄

𝑎2
2
2
2
2
𝑏 ⁄
𝑎𝑏⁄
𝑏 ⁄
4
2
4 , 𝑎 = 1⁄ , 𝑏 = √2⁄ ,
2
5
2
𝑎𝑏⁄
𝑎𝑏
⁄2
𝑎
2
𝑏 2⁄
𝑎𝑏⁄
𝑏 2⁄
4
2
4]

𝑎2
𝑎𝑏⁄
2
2

𝑎

𝑃𝐹 =

𝑎𝑏⁄
2

[

We can calculate the basic quantization as the following equation,
𝑊. 𝑃𝐹
̂ = 𝑟𝑜𝑢𝑛𝑑 (
𝑊
)
𝑄𝑠𝑡𝑒𝑝

(2.2)

Qstep is the quantize step size, which is determined by QP, and the factor (PF/Qstep)
can be used in the reference model software as a multiplication by a factor MF and rightshift, we have
𝑀𝐹
𝑄𝑃
15+𝑓𝑙𝑜𝑜𝑟( )
6
2

𝑃𝐹

= 𝑄𝑠𝑡𝑒𝑝


Therefore, the Equation (2.2) can be represented by,
̂ = 𝑟𝑜𝑢𝑛𝑑 (
𝑊

𝑊.𝑀𝐹

𝑄𝑃
15+𝑓𝑙𝑜𝑜𝑟( )
6
2

)

(2.3)

The secret data is embedded into the quantized DCT coefficients as in following
formula,
̂′ = 𝑊
̂ +∆
𝑊

(2.4)

where =(ai,j)4×4 is the 4×4 error matrix added to the 4×4 quantized DCT coefficient
̂ by data hiding.
matrix 𝑊
The re-scaling step shown in (2.5), after that, is applied to calculate the inverse ICT,
and the post-scaling step as depicted in (2.6), we can get the residual block R’.
̂′ = 𝑊
̂ . 𝑄𝑠𝑡𝑒𝑝. 𝑃𝐹. 64

𝑊

(2.5)

̂′ 𝐶𝑖
𝐶𝑖𝑇 𝑊
)
64

(2.6)

𝑅′ = 𝑟𝑜𝑢𝑛𝑑 (
1

1
1
1 1⁄2 −1
where 𝐶𝑖𝑇 =
1 − 1⁄2 −1
[1

−1

1

1⁄
2
−1
1
− 1⁄2]


We able to get the residual block after embedding R’’, after the re-scaling step as
described in (2.7), and the inverse ICT and the post-scaling step as showed in (2.8) of
the decoder,
̂ ′′ = 𝑊
̂′ . 𝑄𝑠𝑡𝑒𝑝. 𝑃𝐹. 64 = 𝑊
̂ . 𝑄𝑠𝑡𝑒𝑝. 𝑃𝐹. 64 + ∆. 𝑄𝑠𝑡𝑒𝑝. 𝑃𝐹. 64
𝑊
9

(2.7)

FCU e-Theses & Dissertation (2019)


Novel Data Hiding Techniques for H.264/AVC Video

𝐶𝑖𝑇 𝑊 ′′𝐶𝑖
̂ + ∆). 𝐶𝑖 )
𝑅 = 𝑟𝑜𝑢𝑛𝑑 (
) = 𝑟𝑜𝑢𝑛𝑑(𝐶𝑖𝑇 . 𝑄𝑠𝑡𝑒𝑝. 𝑃𝐹. (𝑊
64
′′

(2.8)

Histogram Shifting
Ni et al. [96] had generated the grayscale image’s (512 × 512 × 8) histograms. In this
histogram, the zero point and the peak point have found by corresponding to the
grayscale value. The zero point means no pixel in the given image, and the peak point

is the maximum number of pixel in the given image. The finding of peak point was
proposed, in order to increase the embedding capacity as large as possible.
In this study, the histogram is generated by two steps. Firstly, the peak_point and
zero_point are found by histogram values. Then, the whole image is scanned from top
to bottom, or from left to right. The grayscale value of pixels is shifted from peak point
value to zero point values by increasing by 1. After histogram shifting, the to-beembedded secret data will be checked. If a corresponding to-be-embedded bit in the
embedding data is binary “1”, the pixel value is increased by 1, vice versa, the pixel
value is not changed. In order to increase the embedding capacity in each image, Ni et
al. [106] proposed multiple pairs of maximum point and minimum point, which are
peak_point and zero_point in a range of the histogram.

(b) Histogram shifting

(a) Original histogram

(c) Embedding data
Figure 2.3. Embedding with histogram shifting techniques

10

FCU e-Theses & Dissertation (2019)