当前位置:网站首页>基于差值扩展的可逆水印方法
基于差值扩展的可逆水印方法
2022-07-28 05:23:00 【zzuls】
1.将图像分成像素点对(x, y) x,y∈Z,0≤x,y≤255
个人认为,这个拆分过程需要专门记录,就是记录了图像中哪两个像素组成了这个像素对,这是为了能后续还原出水印。
2.定义其整数均值l 和差值h

公式一 公式二
3.将秘密比特信息m以差值扩展的方法嵌入到差值h中:
![]()
4.将得到的h′代入公式二中,得到新的图像像素对,形成嵌入秘密信息后的图像。
5.为了保证可逆隐藏后的数据没有溢出,设载体图像为8bit的灰度图像,可得某像素点对进行可逆数字水印的条件。

以任意一个像素对举例
加密过程:
(1)假设一个图像对为(206,201)即x=206,y=201
(2)l = floor((x+y)/2) = 203 h=206-201=5
(3)将h转化为二进制101
(4)假如我们对于这个像素对要嵌入的秘密消息m=1,则把1加入到101的LSB上得到h'
(5)得到h' = 1011 即 h' = h*2+m = 2*5+1 = 11
(6)把h'带入公式二得到:x' = 203+ floor((11+1)/2) = 203+6 = 209 y'=203 - floor((11)/2) = 198
(7)故加密得到的信息为(x',y')=(209,198)
解密过程:
(1)已知(x',y')=(209,198)
(2)计算 l' = floor((x'+y')/2) = 203 h' = x' - y' = 11
(3)m = LSB(h') = LSB(1011) = 1 //就是求h'的最后一个比特位
(4)h = floor(h'/2) = 5
(5)x = l'+ floor((h+1)/2) = 203 + 3 = 206 y = l'-floor(h/2) = 203-2 = 201
(6)故还原出原始像素对(x,y) = (206,201) 秘密消息 m = 1
参考文献:
Jun Tian Reversible Data Embedding Using a Difference Expansion
边栏推荐
- Knowledge point 21 generic
- There is a problem with MySQL paging
- Ssh/scp breakpoint resume Rsync
- Cluster operation management system, to answer questions about the process
- Shutter webivew input evokes camera albums
- UNL class diagram
- ssh/scp断点续传rsync
- Deep learning (self supervision: CPC V2) -- data efficient image recognition with contractual predictive coding
- Improved knowledge distillation for training fast lr_fr for fast low resolution face recognition model training
- Dataset class loads datasets in batches
猜你喜欢

First meet flask

深度学习(自监督:CPC v2)——Data-Efficient Image Recognition with Contrastive Predictive Coding

Reinforcement learning - proximal policy optimization algorithms

Deep learning (self supervised: Moco V3): An Empirical Study of training self supervised vision transformers

强化学习——价值学习中的DQN

小程序开发如何提高效率?

微信小程序开发费用制作费用是多少?

Reinforcement learning - dqn in value learning

5、 Video processing and GStreamer

How much does small program development cost? Analysis of two development methods!
随机推荐
深度学习(自监督:MoCo V3):An Empirical Study of Training Self-Supervised Vision Transformers
CertPathValidatorException:validity check failed
Deep learning (self supervision: simple Siam) -- Exploring simple Siamese representation learning
深度学习(增量学习)——ICCV2022:Contrastive Continual Learning
Kubesphere installation version problem
transformer的理解
What are the advantages of small program development system? Why choose it?
How much does it cost to make a small program mall? What are the general expenses?
Digital collections become a new hot spot in tourism industry
循环神经网络
UNL class diagram
Mysql5.6 (according to.Ibd,.Frm file) restore single table data
What are the points for attention in the development and design of high-end atmospheric applets?
5、 Video processing and GStreamer
【4】 Redis persistence (RDB and AOF)
Uview upload component upload upload auto upload mode image compression
深度学习(增量学习)——ICCV2021:SS-IL: Separated Softmax for Incremental Learning
pytorch深度学习单卡训练和多卡训练
【6】 Redis cache policy
On low resolution face recognition in the wild:comparisons and new technologies