当前位置:网站首页>Reversible digital watermarking method based on histogram modification
Reversible digital watermarking method based on histogram modification
2022-07-28 06:16:00 【zzuls】
0. I think the histogram of the image , It reflects the distribution of pixel values of the image , It tells you that the pixel value of this image is 0-255 How many pixels are there . The abscissa represents the pixel value , The ordinate represents the number of pixels corresponding to the pixel value in the image .ps: The pixel value of gray-scale image can also be said to be gray-scale value .

( Image source reference watermark )
1. Early foundation : You need to find the zero point of the histogram ( The corresponding pixel value is z) And peak point ( The corresponding pixel value is p). This zero point can be simply understood as the histogram ordinate 0 The point of , If there is no zero , Choose the minimum value instead of zero , Also record 0 The value corresponding to the point . The peak point is the point with the most pixels in the histogram , The reason for finding the peak value is that the secret message needs to be embedded into the pixel value p Of pixels , It can maximize the watermark capacity .

Embedding watermark steps ( Suppose there is ):
(1) First, calculate the histogram of the image , And find the zero point , That is, pixels without any gray value in the image are recorded as z; Then find the peak value of the gray value with the most pixels in the histogram , Write it down as p. For the convenience of narration , Hypothetical hypothesis p<z.
ps: The peak point may also be on the right of zero , But the actual steps of embedding watermark are similar
(2) From top to bottom 、 Scan each pixel in the image from left to right , The gray value of each pixel is
Express , When
>z or
<p when , The value of the pixel remains unchanged ; otherwise , The gray value of the pixel plus 1:
(3) Pixels in the image whose gray value is equal to the peak point , Is the point where secret information can be embedded , Convert secret information into binary streams , use
Express . Embed information sequentially have to
.
ps: How do you understand this , If p=128, There are... In the image 3 The gray value of points is 128 Respectively p1,p2,p3. If I want to embed watermark information as 101, Then I can make p1 = p1+1=129、p2 = p2+0 = 128、p3 = p3+1 = 129;
(4) The obtained image composed of gray value is the image after embedding secret information . meanwhile p、z Save in the form of a key .
Steps of extracting watermark information :
(1) Read key , obtain p、z Value ;
(2) Progressive image , When
when , It indicates that this point is a hidden information point , Extract information 0 And keep the gray value of this point unchanged ; When when , This point is also a hidden information point , Extract information 1 And reduce the pixel value 1.
ps: Why?
and Time is the point of hiding information ? Because in the original image
All points plus one
, therefore It must be a secret message 1 Only some
(3) When
or
when , The value of the pixel remains unchanged ; When
when , The gray value of the pixel is reduced 1.
(4) Get a new image composed of gray values , That is, the carrier image after extracting secret information .
边栏推荐
- Deep learning (self supervision: simpl) -- a simple framework for contractual learning of visual representations
- How much does it cost to make a small program mall? What are the general expenses?
- Basic usage of word2vec and Bert
- Prime_ Series range from detection to weight lifting
- Reinforcement learning -- SARS in value learning
- 深度学习数据窃取攻击在数据沙箱模式下的威胁分析与防御方法研究阅读心得
- What are the advantages of small program development system? Why choose it?
- ESXi 7.0 Update 1c中加入的systemMediaSize启动选项
- 利用辅助未标记数据增强无约束人脸识别《Boosting Unconstrained Face Recognition with Auxiliary Unlabeled Data》
- File upload vulnerability summary
猜你喜欢

深度学习——Patches Are All You Need

Adaface: quality adaptive margin for face recognition image quality adaptive edge loss for face recognition

Reinforcement learning - dqn in value learning

Word2vec+ regression model to achieve classification tasks

Neural network learning

Apache log4j arbitrary code execution replication

Reinforcement learning - Multi-Agent Reinforcement Learning

二、OpenVINO简述与构建流程

强化学习——不完全观测问题、MCTS

Small program development solves the anxiety of retail industry
随机推荐
TensorFlow2.1基本概念与常见函数
Cluster operation management system, to answer questions about the process
How much does small program development cost? Analysis of two development methods!
Deep learning (self supervision: Moco V2) -- improved bases with momentum contractual learning
C语言EOF的理解
Differences between processes and threads
What are the detailed steps of wechat applet development?
Deep learning (incremental learning) - (iccv) striking a balance between stability and plasticity for class incremental learning
UNL-类图
Bert based data preprocessing in NLP
On low resolution face recognition in the wild:comparisons and new technologies
Notes on the thinking of software analysis and library acquisition of a Taobao customer
Distributed cluster architecture scenario optimization solution: distributed scheduling problem
Automatic scheduled backup of remote MySQL scripts
How to use Bert
深度学习(一):走进机器学习与深度学习理论部分
The difference and connection between cookies, sessions and tokens
Nsctf web Title writeup
强化学习——多智能体强化学习
Deep learning - metaformer is actually what you need for vision