10.1007@s11042-019-08338-x.pdf资源-CSDN文库

需积分: 10 134 浏览量 2020-03-11 19:26:24 上传评论收藏 2.11MB PDF 举报

资源推荐

资源详情

资源评论

An improved robust image watermarking by using

different embedding strengths

Dhani Ariatmanto

& Ferda Ernawan

Received: 15 February 2019 /Revised: 22 August 2019 /Accepted: 1 October 2019

Springer Science+Business Media, LLC, part of Springer Nature 2020

Abstract

Image watermarking technique is an alternative solution to protecting digital image

strengths for embedding a watermark. An image is divided into non-overlapping blocks of

8 × 8 pixels. The variance pixel value was computed for each image block. Image blocks

with the highest variance value were selected for the embedding regions. Therefore, it was

transformed by discrete cosine transforms (DCT). Five DCT coefficients in the middle

frequency were selected and the average of selected DCT blocks was calculated to generate

different embedding strengths by using a set of rules. The watermark bits were embedded

by using a set of embedding rules with the proposed different embedding strengths. For an

additional security, the binary watermark was scrambled by using an Arnold Transform

before it was embedded. The experimental results showed that the proposed scheme

achieved a higher imperceptibility than the other existing schemes. The proposed scheme

achieved a watermarked image quality with a PSNR value of 46 dB. The proposed scheme

also produced a high watermark extracting resistance under various attacks.

Keywords Different embedding strengths

Adaptive scaling factor

Embedding scheme

Extracting scheme

Image watermarking

Discrete cosine transforms

1 Introduction

Rapid advances in digital technology has encouraged the protection of multimedia data

ownership. Digital multimed ia data is vulnerable to manipulation, illegal copying and

Multimedia Tools and Applications

https://doi.org/10.1007/s11042-019-08338-x

* Ferda Ernawan

ferda1902@gmail.com

Faculty of Computer Science, Universitas AMIKOM Yogyakarta, Ring Road Utara Condong Catur,

Sleman, Yogyakarta 55283, Indonesia

Faculty of Computing, Universiti Malaysia Pahang, Lebuhraya Tun Razak, 26300 Gambang,

Kuantan, Malaysia

distribution [34, 54]. Digital watermarking provides a solution to secure and protect intellec-

tual property [21, 43]. Digital watermarking is a data-hiding method for copyright protection

and authentication [20, 32, 38]. Furthermore, researchers have designed highly robu st

watermarking schemes to protect against malicious and non-malicious attacks [8, 31, 37, 45,

56]. Image watermarking methods can be performed in spatial [23, 28, 36] and frequency

domains [16, 27, 44]. The spatial domain is a popular method in fragile watermarking for

authenticating images, detecting tamper attacks and recovering image pixels [39, 40]. Digital

watermarking based on spatial domain directly embeds a watermark by altering the pixels.

This method provides low computational cost, but it is easily destroyed by signal processing

attacks [15]. In contrast, watermarking based on frequency domain provides better durability

against various attacks and produces good imperceptibility [3]. Many techniques are used to

implement a frequency domain in digital watermarking, such as discrete cosine transforms

(DCT) [46], discrete wavelet transforms (DWT) [4], singular value decomposition (SVD) [17],

integer wavelet transforms (IWT) [19] and redundant discrete wavelet transforms (RDWT)

[18]. Consequences of using more than one transforms in a watermarking system are high

computational cost and complexity [11, 55].

Researchers have begun to research on quantum image watermarking [1, 35, 48, 49, 51]. The

quantum image is presented by capturing colours from the monochromatic electromagnetic

wave that requests for a special machine and location [9, 50]. The existing digital watermarking

has widely applied DWT, RDWT, IWT, SVD for copyright protection. These methods request

large computational costs and complexity. They need a special hardware for implementation. A

digital watermarking technique that requires the least computational complexity and can be

implemented on a minimum hardware requirement has become the current challenge in digital

watermarking. A fast embedding and extracting watermark without a significant change in the

host image quality has become the main issue in digital watermarking.

The existing image watermarking schemes require a large computational time for embed-

ding and extracting a watermark. The watermarking techniques cannot be implemented in a

mobile application that provides minimum hardware specifications. With the minimum com-

putational cost in a mobile phone, it is quite difficult to deal with wavelet do main or

optimisation methods in digital watermarking. Digital images can be transferred from one

user to anoth er through mobile applications. A digital image needs protection from

unauthorised users. Moreover, digital watermarking technique in a mobile application requires

less computational costs and transmission. The existing image watermarking techniques do not

consider improvement on the robustness of watermarked images in mobile applications. Image

watermarking technique must consider a low-cost computational complexity due to limited

hardware in mobile application. Image watermarking techniques based on a wavelet domain

require a special hardware for implementation. The existing watermarking techniques based on

optimisation methods also request for high computational costs and complexity to embed a

watermark image.

Image watermarking techniques use a scaling factor method to embed a watermark in the

host image [24]. A scaling factor is widely applied in all selected blocks of DCT coefficients

for weighting the embedded watermark. However, each DCT block has different characteristic,

and hence the usage of a scaling factor for all selected DCT blocks may not give optimal

invisibility and robustness of the watermarked image. Therefore, each selected DCT block

should use different scaling values. A major priority of the digital watermarking technique is a

balance between robustness and imperceptibility [29]. For copyright protection, embedded

data must be strong and able to withstand against signal alteration [5].

Multimedia Tools and Applications

This paper proposes a different embedding strength for embedding a watermark in each

selected image block. A new embedding technique was developed based on the effect of

selected DCT coefficients in the middle frequency against average DCT coefficients of its

block. The proposed scheme generates a different embedding strength for each selected block.

The different proposed embedding s trengths can improve the watermarked image

imperceptibility, especially for small and medium host images. The proposed scheme can also

achieve high robustness against different types of attack. The proposed watermarking tech-

nique provides less computational time and complexity. It is also compatible with mobile

applications that require minimum hardware specifications and less computational cost. The

main contributions of the proposed scheme are:

1. The proposed scheme can provide a low-cost computational time for embedding and

extracting the watermark image. It uses DCT domain that is applicable in mobile phones.

The advantages of DCT in image watermarking are:

& DCT algorithm is easy to be realised and implemented in the hardware. It provides fast

speed and high precision in digital signal processing [14]. DCT can be implemented in

an environment with minimum hardware specification, such as a mobile phone.

& DCT has lower complexity as compared to other transforms and the ability to exhibit

an excellent energy compaction of image signals [10]. DCT provides low-cost energy

consumption. It is suitable for embedding a watermark into the digital image in

mobile applications.

2. The proposed scheme examines the middle DCT coefficients with psychovisual threshold

to obtain the different embeddings strength for embedding and extracting the watermark

image. The proposed embedding technique can be substantially adapted by considering

the image content itself.

3. The proposed scheme uses Arnold transform and a secret key to secure and protect the

embedded watermark image. The proposed scheme can avoid unauthorised users or

attackers from recognising the watermark image.

4. The proposed scheme improves the robustness of embedded watermark under various

attacks. It produces minimum distortion of the watermarked image, especially for small

and medium images.

5. The proposed scheme also provides faster watermark embedding and extracting than other

existing image watermarking schemes.

This paper is organised as follows. Section 2 discusses related works. Section 3 presents the

proposed watermark embedding and extracting, while Section 5 presents the experimental

results and Section 6 concludes the paper.

2 Related works

Watermarking schemes use a scaling factor to scale the strength of embedding watermark. The

balancing of robustness and imperceptibility is controlled by a scaling factor. A scaling factor

has an important role to determine the quality of a watermarked image and robustness of the

extracted watermark [2, 22]. Schemes by Ernawan [12] and Su [52] presented a threshold as a

Multimedia Tools and Applications

trade-off between normalising cross-correlation (NC) and structural similarity index (SSIM)

for the embedding watermark scaling factor. The scaling factor gives a significant effect on the

quality and robustness of the extracted watermark. The schemes showed that the scaling factor

can achieve a good resistance against different types of attack. Both schemes also produce high

imperceptibility of the watermarked image. The schemes may still be improved by using

different scaling values for each selected block. Each selected DCT block has different effects

towards image reconstruction. Therefore, selected DCT blocks must be differently scaled to

achieve optimal invisibility and robustness. It motivates the development of a new embedding

technique by using different embedding strengths for each image block.

A scheme by Das [6] presented a blind watermarking scheme based on the DCT domain by

using an inter-block correlation. This scheme examined the correlation between two adjacent

blocks, and used a scaling factor for the embedding strength of the watermarked image. The

experimental results reported that it could achieve high robustness against JPEG compression

attacks. However, this scheme still required a high computational cost for embedding a

watermark.

A scheme by Roy [41] presented an image watermarking based on the DCT domain by

using a repetition code in the middle frequencies. The scheme used the coding theory of the

error correction code (ECC) for embedding a watermark. The scheme compared two coeffi-

cients in the middle frequency in the embedding process. The simulation results reported that

this scheme achieved high robustness and good imperceptibility but it also requested high

computational time.

AschemebyLai[25] presented a robust image watermarking scheme based on DCT-SVD

and human visual characteristics. This scheme selects the image blocks for embedding a

watermark by using entropy and edge entropy to represent the human visual characteristics.

Image blocks that have the lowest entropy value are not noticeable by human eyes. The

scheme modified certain coefficients of the orthogonal U matrix by using a threshold for

embedding a watermark. The experimental results reported that this scheme achieved high

robustness and good visual quality of the watermarked image. Although the scheme can

improve robustness, the use of a threshold value for all selected DCT blocks may not achieve

an optimal balance between invisibility and robustness for different images.

The scheme by Makbol [30] presented a robust block-based watermarking that was

established on discrete wavelet transforms (DWT) -singular value decomposition (SVD)

and the human visual system. The regions o f inserting a watermark are selected from the

lowest entropy value and edge entropy value. SVD was performed on the lo w-sub-band in

the first level of DWT. The elements of the orthogonal U matrix are modified to embed the

watermark image. The experimental results reported that the scheme can achieve high

robustness of the watermarked image. The sc heme used a threshold value as the embed-

ding strength for all selected image blocks. The use of a threshold as the embedding

strength for all selected blocks may not achieve an optimal trade-off between invisibility

and robustness.

AschemebyDey[7] presented a robust image watermarking based on DWT-DCT-SVD

for the authentication system. This scheme presented firefly algorithms to find the optimal

scaling factor for embedding watermark images. The singular values are modified by scaling

factors obtained from the firefly optimisation process. The scheme achieved a high robustness

level of the watermarked image. However, the combination of more than one transform

domain and firefly algorithms had increased the computational cost and time complexity for

embedding and extracting the watermark image.

Multimedia Tools and Applications

Alternatively, a scheme by Yadav [57] presented an adaptive watermarking technique based

on the dynamic scaling factors. The scheme selected the embedding locations based on the

highest entropy value. Their scheme scaled upwards and downwards of the low-frequency on

the wavelet transform to determine the scaling factor. The experimental results reported that it

can improve robustness against different types of attack.

A scheme by Ernawan [13] presented an optimal DCT-psychovisual threshold in image

watermarking. The embedding locations are selected based on the combination of entropy and

edge entropy. The selected DCT coefficients are chosen by examining the gap between the

psychovisual threshold and minimum quantisation value towards error reconstruction. This

scheme uses a threshold obtained from a trade-off between imperceptibility and robustness of

watermarked images under JPEG compression. Although this scheme can improve the

watermarking performance, a scaling factor may not achieve an optimal scale for embedding

a watermark on each selected image b lock. The existing watermarking schemes are

summarised in Table 1.

3 Proposed watermarking scheme

In this section, the embedding and extraction procedures of the proposed DCT watermarking

scheme by using different embedding strengths are explained in detail.

3.1 Embedding procedures

The block diagram of the proposed embedding scheme is shown in Fig. 1.

The embedding procedure can be summarised as follows:

Step 1: A host image is divided into non-overlapping blocks of 8 × 8 pixels from left to

right, top to bottom.

Step 2: Each image block is computed by variance function. The embedding locations are

selected by measuring the lowest variance pixels in each image block. Variance pixels can

be defined by [33]:

∑

i¼1

∑

j¼1

i; j

−F



ð1Þ

where σ denotes the variance value, N and M represent the size of each image block, F denotes

each image pixel and

Frepresents the average pixel value of that image block. The variance

pixel of each block is calculated to identify the pixel variance and classify the different regions.

The lowest variance pixel indicates a part of the background image. The block with the lowest

variance of pixels indicates the less complex components [33]. In this paper, a watermark was

embedded in the regions that have less complex components. In this scheme, the lowest

variance value was used to select image blocks for embedding the watermark. The experiments

used a watermark with a size of 32 × 32 pixels. The number of selected blocks was the same as

the number of watermark bits. The selected blocks with the lowest variance of pixels were

stored into a database.

Step 3: Selected blocks with the lowest variance pixels were computed by DCT. Each

image block F was transformed into the corresponding DCT coefficients as follows:

Multimedia Tools and Applications

剩余26页未读，继续阅读

评论收藏

内容反馈

zxzoxo

粉丝: 0
资源: 1

10.1007@s11042-019-08338-x.pdf

PDA DETECTOR FOR DS-CDMA COOPERATIVE SYSTEM.pdf

P-VCD:基于内容的视频复制检测-开源

BURNINTEST--硬件检测工具

TabuVis:表格数据的交互式可视化-开源

年月日汶川地震导致的周边断层应力变化.pdf

OPPO A7 A82位号图.pdf A32 A92S K7 A82 A7 ACER2 FINDX2 PRO

基于深度学习的问答系统开发.pdf

华为认证助理网络工程师考试模拟题3.pdf

python二级考试习题.pdf

江西省上饶县中学2018届高三数学上学期第二次月考试题理-8页.pdf

pdf转swf的软件

ad hoc多信道多接口研究

JDK 1.8 64位.rar

SoftCnKiller2.54.zip（因为要下载码，弃用不更新了）

编译好的64位pycdc.exe(支持python3.10和python3.11)

广联达授权9.9（2.5）

工程伦理的期末考试题库

雷电模拟器改成真机环境

2024广联达提示盗版文件被锁解锁工具.zip

ISO26262标准(中文版)下载.pdf

Open3D算法测试数据.rar

NFC-Reader-Tool-电脑版.zip

常用经典斯坦福点云数据

化作尘智能门锁项目视频教程笔记.pdf

ghidra 11.0 逆向分析工具

matlab2020b ubuntu.txt

ESP32-ASR.zip

IDEA2020.3详细安装教程.docx

zip压缩包密码破解工具(免费)

最新资源