CN103065156B - Based on digital camera images and the computer generated image differentiating method of white balance - Google Patents
Based on digital camera images and the computer generated image differentiating method of white balance Download PDFInfo
- Publication number
- CN103065156B CN103065156B CN201210563893.1A CN201210563893A CN103065156B CN 103065156 B CN103065156 B CN 103065156B CN 201210563893 A CN201210563893 A CN 201210563893A CN 103065156 B CN103065156 B CN 103065156B
- Authority
- CN
- China
- Prior art keywords
- image
- white balance
- digital camera
- computer generated
- camera images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The present invention relates to a kind of multi-media forensic technology, particularly a kind of digital camera images based on white balance and computer generated image differentiating method, to solve the problem that digital image content authenticity is differentiated.Utilize digital camera to take vestige that the image obtained can leave over white balance process, the image that Practical computer teaching obtains then need not through the principle of this processing procedure, according to presence or absence detection to white balance process vestige in testing image, thus realize digital camera images and computer generated image is distinguished.The present invention can be applicable to the formulation of digital picture right discriminating system, legal cartoon the application such as to distinguish with pirate, obtains initiative to the use of the digital image content authenticity detection method of better performances.
Description
Technical field
The invention belongs to multimedia evidence obtaining field, particularly relate to a kind of digital camera images based on white balance and computer generated image differentiating method.
Background technology
The universal arrival indicating digital Age of PC and internet.The facility of digital Age is self-evident, but consequent social concern is also following, such as ' electronics crime '.Along with constantly soaring electronics crime case, digital evidence obtaining (digitalforensics) technology is arisen at the historic moment.It to comprise the relevant information of digital signal generation, storage, transmission etc. as the evidence needed for crime survey and legal affairs, such as Email, office documents, sound, video, image, internet records, internal storage data, GPS information etc.In the multiple directions of digital evidence obtaining subject, multi-media forensic technology is the popular direction of just having risen in recent years.Authenticity and the reliability of numerical information (as digital picture, video, audio frequency etc.) are mainly verified in multimedia evidence obtaining, and whether main application is the reliability for verifying digital evidence in court, can accept and believe.Nowadays, digital picture because of its rapidly universal, wide-scale distribution and very easily the characteristic such as amendment cause the extensive concern of people.Digital picture is seen everywhere, and newpapers and periodicals, magazine, network etc. be its route of transmission all.Simultaneously, even photo-editing software from strength to strength makes ordinary people still can be edited digital picture by PC and revise, even photo is forged.Along with the upgrading of image editing software, tampered image, computer generated image are more and more true to nature, are even difficult to the naked eye distinguish its authenticity.But except some specific Entertainment Scene, time most of, people need to know the seen digital picture true and false definitely.Particularly once involve economic interests, individual reputation, spin, crime survey etc., the information that the authenticity of digital image content, the reliability in source must confirm especially.Unfortunately, forge the life that image is deep into us already, the numerous areas such as science, politics, business, law, media all receive impact in various degree.And traditional verification method is difficult to the digital picture " prestige crisis " dealing with today.In order to better tackle the safety problem of digital picture, the passive forensic technologies of digital picture is applied and gives birth to.The passive forensic technologies of digital picture belongs to the research frontier occurred in recent years, and it all has in fields such as medical science, military affairs, court evidence, news report, insurance risk investigation, ecommerce dabbles, and applies very extensive.Domestic and international research is still in the junior stage, though be technically faced with lot of challenges, does not affect the active demand of market to it at all, and application prospect is had an optimistic view of.
Digital camera images and computer generated image are distinguished and are in the passive forensic technologies of digital picture one and popularly study branch.In today that graphics software spreads unchecked, the vivid effect that computer generated image can be accomplished " mixing the spurious with the genuine ".How differentiate between images content is digital camera is taken the natural picture obtained, and is the record to real world and reflection; Or the just embodiment of computer mapping technology, content is also untrue, or even the distortion to reality, and can mislead the public.These are all in the problem must answered out in non-amusement (as criminal investigation, court put to the proof) scene.Also be the emphasis of this patent research.
Traditional method distinguished for digital camera image and computer generated image is the differentiating method based on Wavelet Detection.These class methods carry out multi-level decomposition often through to small echo, and on each subband, ask for multiple statistic to form distinguishing characteristic.In these class methods existing, common way is the progression constantly increasing wavelet decomposition in order to improve differentiation precision, and statistical nature is asked on more subbands, be that cost exchanges the raising distinguishing precision for being on the increase assumed condition, increasing distinguishing characteristic dimension, this will inevitably cause statistical nature dimension bigger than normal; Meanwhile, wavelet transformation well can not excavate the marginal information of image, and different just with edge details of one of important difference between the artificial image of the real world images obtained captured by digital camera and Practical computer teaching, this also result in the deficiency distinguished precision and still have much room for improvement.
Summary of the invention
The object of the invention is to overcome above-mentioned weak point, thus solve during digital camera images and computer generated image distinguish, the problem that statistical nature dimension is excessive, differentiation precision is not high.
The invention provides a kind of digital camera images based on white balance and computer generated image differentiating method, comprise following steps: step 1, set up image data base, image data base comprises some digital camera images and some computer generated images;
Step 2, carries out statistical nature extraction to the every piece image in image data base, obtains set of eigenvectors;
Step 3, sending into the set of eigenvectors of gained in step 2 in sorter and trains, obtaining the differentiation model for distinguishing digital camera images and computer generated image;
Step 4, carries out statistical nature extraction to image to be measured, and gained proper vector sends into sorter, distinguishes model judge by step 3 gained.
And in step 2 and step 4, the implementation of arbitrary image being carried out to statistical nature extraction comprises following sub-step,
Step 2.1, selects n kind white balance algorithm, and light source image being carried out to n kind white balance algorithm is estimated;
Step 2.2, estimate for step 2.1 gained n kind light source, calculate R respectively, the gain coefficient on G, B tri-passages, enumerates this 3 × n gain coefficient
wherein c ∈ r, g, b}, i={1,2 ..., n};
Step 2.3, according to 3 × n the gain factor of trying to achieve in step 2.2, passes through formula
calculate 3 other × n statistical nature;
Step 2.4, the gain coefficient that step 2.2 is tried to achieve
as the 1st to 3rd × n feature, the statistical nature that step 2.3 is tried to achieve
as 3rd × n+1 to 6th × n feature, be combined into final proper vector.
And, get n=5.
The present invention pursues the more excellent digital picture discrimination method of performance, for judging in digital picture authentication whether image is that the true picture obtained taken by digital camera, or the artificial image generated by graphics software, and then truly whether provide reliable basis for estimation for picture material, be beneficial to the development of China's technique of criminal investigation research, advance the formulation of relevant law.
Embodiment
The present invention is by analyzing the principle of digital camera imaging and computerized mapping software image, indicate that digital camera is taken pictures in imaging process, have to pass through this step of white balance process, computer generated image then need not be like this, then with this difference for starting point, design a kind of digital camera images based on white balance and computer generated image differentiating method.
The ultimate principle of the method design is as follows:
The basic skills of most white balance process follows VonKries model, that is: 1. investigate image light source, the colour temperature of evaluate image; 2. according to the result of color temperature estimation, the color modulation gain factor of each color layers on setting image; 3. according to the color modulation gain factor of setting, modulation image color, the image of gained is the image of white balance process.
For with digital camera images, its image have passed through a white balance process, the colour temperature of the light source identical or standard sources colour temperature that is comparable to set by white balance algorithm on image, and computer generated image is quite different.Therefore, if logarithmic code camera image carries out white balance process, be equivalent to the process of second time white balance, its color modulation is changed not quite inevitable or even is not needed modulation.And white balance process is carried out to computer generated image, be now equivalent to first time white balance process, its color modulation is changed relatively large.In other words, in image generation process white balance process with or without can by judging the assessment of color temp.When image be whether the true picture of camera shooting, have passed through in imaging process the problems such as which kind of white balance Processing Algorithm all unknown, by carrying out multiple WHITE TONE estimation to known image, and extract statistical nature; Sorter is then adopted to train the differentiation that the method distinguishing model realizes digital camera images and computer generated image.
Embodiment describes technical solution of the present invention in detail.
A kind of digital camera images based on white balance provided by the invention and computer generated image differentiating method, for judging in digital picture authentication whether image is that the true picture obtained taken by digital camera, or the artificial image that generated by graphics software, and then truly whether provide reliable basis for estimation for picture material.Technical solution of the present invention can adopt computer software technology to realize automatically running.Describe technical solution of the present invention in detail by the following examples.
Embodiment realization flow comprises following steps:
Step 1, set up image data base:
The proprietary image data base for training digital camera images and computer generated image to distinguish model can be set up during concrete enforcement, include digital camera images, computer generated image is some (can collect voluntarily, also the database that other people have set up can be adopted, the DVMM database as Columbia University).
Step 2, statistical nature extracts:
All according to statistical nature extracting method, proper vector is extracted to the every piece image in image data base, the set of eigenvectors that the proper vector obtaining all images in image data base is formed.
Step 3, training area sub-model:
The set of eigenvectors of gained in step 2 being sent in sorter and trains, obtaining the differentiation model for distinguishing digital camera images and computer generated image;
Step 4, testing image is differentiated:
As needs judge whether image is computer generated image time, extracted the feature of testing image by statistical nature extracting method consistent in step 2, gained proper vector send into sorter, judged by the differentiation model trained.
For ease of implementing reference, the embodiment of the present invention provides the concrete solution procedure of the statistical nature extracting method mentioned in above-mentioned differentiating method as follows further:
Step 2.1, image light source is estimated:
Embodiment gets n=5, and image is carried out to the first step operation of 5 kinds of white balance algorithm, namely light source estimates (generally have 4 kinds of white balance algorithm, GrayWorld, MaxRGB, ShadesofGrey, and GreyEdge, but GreyEdge is divided into 2 kinds of situations, 5 kinds of light source estimations technique altogether according to parameter difference).
Step 2.2, the extraction of feature set one:
5 kinds of light sources for step 2.1 are estimated, calculate R respectively, the gain coefficient on G, B tri-passages, enumerate this 15 gain coefficients
wherein c ∈ r, g, b}, i={1,2 ..., 5}, using these 15 statistical natures the as the 1st to the 15th feature;
Step 2.3, the extraction of feature set two:
According to 15 gain factors of trying to achieve in step 2.2, pass through formula
calculate 15 other statistical natures, using these 15 statistical natures the as the 16th to the 30th feature;
Wherein, what abs (.) represented is the operation that takes absolute value, and English is absolute, is the function representation taken absolute value in all kinds of computerese (comprising the conventional programming languages such as C language, VB language, Pascal language, Matlab language).
Step 2.4, the formation of final feature:
The feature that step 2.2 is tried to achieve the as the 1st to the 15th feature; The feature of step 2.3 being tried to achieve the as the 16th to the 30th feature, merges and is combined into final proper vector.
Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.
Claims (2)
1., based on digital camera images and the computer generated image differentiating method of white balance, it is characterized in that: comprise following steps,
Step 1, sets up image data base, and image data base comprises some digital camera images and some computer generated images;
Step 2, carries out statistical nature extraction to the every piece image in image data base, obtains set of eigenvectors;
Step 3, sending into the set of eigenvectors of gained in step 2 in sorter and trains, obtaining the differentiation model for distinguishing digital camera images and computer generated image;
Step 4, carries out statistical nature extraction to image to be measured, and gained proper vector sends into sorter, distinguishes model judge by step 3 gained;
In described step 2 and step 4, the implementation of arbitrary image being carried out to statistical nature extraction comprises following sub-step,
Step 2.1, selects n kind white balance algorithm, and light source image being carried out to n kind white balance algorithm is estimated;
Step 2.2, estimate for step 2.1 gained n kind light source, calculate R respectively, the gain coefficient on G, B tri-passages, enumerates this 3 × n gain coefficient
wherein c ∈ r, g, b}, i={1,2 ..., n};
Step 2.3, according to 3 × n the gain factor of trying to achieve in step 2.2, passes through formula
calculate 3 other × n statistical nature;
Step 2.4, the gain coefficient that step 2.2 is tried to achieve
as the 1st to 3rd × n feature, the statistical nature that step 2.3 is tried to achieve
as 3rd × n+1 to 6th × n feature, be combined into final proper vector.
2., as claimed in claim 1 based on digital camera images and the computer generated image differentiating method of white balance, it is characterized in that: get n=5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210563893.1A CN103065156B (en) | 2012-12-21 | 2012-12-21 | Based on digital camera images and the computer generated image differentiating method of white balance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210563893.1A CN103065156B (en) | 2012-12-21 | 2012-12-21 | Based on digital camera images and the computer generated image differentiating method of white balance |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103065156A CN103065156A (en) | 2013-04-24 |
| CN103065156B true CN103065156B (en) | 2016-01-13 |
Family
ID=48107779
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210563893.1A Active CN103065156B (en) | 2012-12-21 | 2012-12-21 | Based on digital camera images and the computer generated image differentiating method of white balance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103065156B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107635123B (en) | 2017-10-30 | 2019-07-19 | Oppo广东移动通信有限公司 | White balancing treatment method and device, electronic device and computer readable storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101655912A (en) * | 2009-09-17 | 2010-02-24 | 上海交通大学 | Method for detecting computer generated image and natural image based on wavelet transformation |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7778461B2 (en) * | 2006-05-05 | 2010-08-17 | New Jersey Institute Of Technology | System and/or method for image tamper detection |
-
2012
- 2012-12-21 CN CN201210563893.1A patent/CN103065156B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101655912A (en) * | 2009-09-17 | 2010-02-24 | 上海交通大学 | Method for detecting computer generated image and natural image based on wavelet transformation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103065156A (en) | 2013-04-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Guo et al. | Fake colorized image detection | |
| Swaminathan et al. | Digital image forensics via intrinsic fingerprints | |
| Tsai et al. | Fast geometric re-ranking for image-based retrieval | |
| WO2019214557A1 (en) | Method and system for detecting face image generated by deep network | |
| CN106096519A (en) | Live body discrimination method and device | |
| CN103793447A (en) | Method and system for estimating semantic similarity among music and images | |
| CN111444826A (en) | Video detection method and device, storage medium and computer equipment | |
| CN112990484B (en) | Model joint training method, device and equipment based on asymmetric federated learning | |
| CN109558792B (en) | Method and system for detecting internet logo content based on samples and features | |
| Lu et al. | Steganalysis of content-adaptive steganography based on massive datasets pre-classification and feature selection | |
| Monteiro et al. | Design and evaluation of classifier for identifying sign language videos in video sharing sites | |
| Mani et al. | A survey on digital image forensics: Metadata and image forgeries | |
| CN110674692A (en) | Target accurate retrieval method and system based on difficult sample generation | |
| CN103065156B (en) | Based on digital camera images and the computer generated image differentiating method of white balance | |
| Conotter et al. | Detecting photographic and computer generated composites | |
| CN114022923A (en) | Intelligent collecting and editing system | |
| Yu et al. | A multi-scale feature selection method for steganalytic feature GFR | |
| CN107563788A (en) | The stage division and server of shared billboard | |
| CN103065155B (en) | A kind of digital camera type identifier method based on CFA | |
| CN103458279A (en) | Attraction judging method based on advertisement video image data | |
| CN114329050A (en) | Visual media data deduplication processing method, device, equipment and storage medium | |
| Liu et al. | Anti‐noise image source identification | |
| CN112434547A (en) | User identity auditing method and device | |
| Chen et al. | Identification of image global processing operator chain based on feature decoupling | |
| CN109344289A (en) | Method and apparatus for generating information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210707 Address after: 215000 unit 01, 5 / F, building a, 388 Xinping street, Suzhou Industrial Park, Suzhou City, Jiangsu Province Patentee after: BOOSLINK SUZHOU INFORMATION TECHNOLOGY Co.,Ltd. Address before: 430072 Hubei Province, Wuhan city Wuchang District of Wuhan University Luojiashan Patentee before: WUHAN University |
|
| TR01 | Transfer of patent right |