CN103020641A - Digital photo source distinguishing method based on DFT (Discrete Fourier Transform) cross correlation analysis - Google Patents
Digital photo source distinguishing method based on DFT (Discrete Fourier Transform) cross correlation analysis Download PDFInfo
- Publication number
- CN103020641A CN103020641A CN201210499692XA CN201210499692A CN103020641A CN 103020641 A CN103020641 A CN 103020641A CN 201210499692X A CN201210499692X A CN 201210499692XA CN 201210499692 A CN201210499692 A CN 201210499692A CN 103020641 A CN103020641 A CN 103020641A
- Authority
- CN
- China
- Prior art keywords
- photo
- prnu
- dft
- sigma
- correlation analysis
- 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.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a digital photo source distinguishing method based on DFT (Discrete Fourier Transform) cross correlation analysis. The method comprises the following steps of in sequence: (1) segmenting a reference photo and a testing photo according to an R colour channel, a G colour channel and a B colour channel; (2) subtracting relative to original photo channel Ii and PRNU (Photo Response Non Uniformity)-free Iirj and Iitk to obtain residual images nir and nit; (3) calculating reference PRNU wci and fusing reference PRNU wc of each channel, wherein i is a member of the braces of R, G, B; (4) calculating and fusing the cross-correlation coefficients cwn of the reference PRNU wc and the residual noise nt of the testing image; and (5) positioning the part with the maximum value of the cross-correlation coefficient, calculating the Eeuclidean distance from the position to the original point, and determining the decision threshold. According to the digital photo source distinguishing method based on the DFT cross correlation analysis, the characteristic information is adopted and utilized for distinguishing the sources of the photos of seven models shot by the digital cameras of different brands, and the average distinguishing rate reaches 96.3%. The method disclosed by the invention is high in application value.
Description
Technical field
The present invention relates to digital photograph source recognition technology, in particular a kind of digital photograph based on the DFT cross correlation analysis comes source discrimination.
Background technology
Along with the fast development of social progress and science and technology, digital photograph becomes the human a kind of important mode of obtaining with exchange message in digital communication technology.Mass storage device is under the impetus of technical progress, and price becomes and more and more comes more cheaply, and digital photograph memory requirement storage space is the bigger the better, and these provide condition for convenient storage and widespread uses of digital photograph.Yet its superiority also provides means and instrument for illegal activity along with the expansion of digital photograph range of application and when enriching our life.Consider from judicial angle, the Emergence and Development of digital photograph Identification of The Origin technology then provides a solution for this problem.
Digital camera is accepted by vast consumer now, and the source distribution that comes of digital photograph very extensively and is easily obtained, and these benefits are undoubtedly.It also is a very large defective that but digital photograph is easy to obtain.Because we do not know that digital photograph accurately originates.We can know by inference, can become contention in the daily life and a kind of general source of evidence in the court judgment at the society digital photograph.Same, the authenticity that becomes new report that the TV news digital photograph is general proves.Therefore carrying out the technical research of digital photograph Identification of The Origin is a very significant job.
Digital photograph Identification of The Origin technology was just risen in recent years in the world, still was at present the starting stage, and only there are several families in relevant research troop, and was in the majority with university and the research institution of the U.S..The people such as Kharrazi of U.S. Polytechnic university are by analyzing digital camera internal image processing procedure, extracting the authentication of classifying of the multidimensional characteristic can reflect the camera characteristics from digital picture, is a kind of typical algorithm of carrying out source evidence forensics by the extraction feature; The people such as Tsai have also proposed similar method, utilize characteristics of image to differentiate; The people such as K.S.Choi have added camera lens radial distortion feature on this basis, have improved the accuracy rate of camera origin classification; The periodic feature that the people such as S.Bayram adopt EM algorithm detection camera cfa interpolation to introduce carries out source evidence forensics; The professor HanyFarid leader's of U.S. Dartmouth university Research Team has early proposed the digital picture source evidence forensics method based on the wavelet multi-scale analysis, and logarithmic code photo and computer generated image are classified, and its achievement in research is remarkable; The Research Team of U.S. Columbia university take Shih-Fuchang as core, obtained certain achievement in the authentication of the source of image, they propose to utilize unlike signal disposal route in the camera response function reflection digital camera imaging process, the digital picture that different cameral the is taken authentication of originating; The people such as Wu Hao of U.S. Marytand university realize classification by estimating camera cfa interpolation coefficient; The J.Fridrich team of U.S. Binghamton university, in these years also always follow-up study digital picture is come source technology, and huge contribution has been made in the multi-media information security field.In recent years, digital camera images Identification of The Origin technology also has been subject to attention successively in internationally famous periodical and meeting.Top professional international conference etc. has had relevant meeting special topic successively, and these are all declaring publicly importance and the frontier nature of digital camera Identification of The Origin technology.Domestic relevant digital photograph Identification of The Origin technology monographic study is also at the early-stage, and relevant achievement in research and technology are also considerably less.
Summary of the invention
Technical matters to be solved by this invention is to provide a kind of digital photograph based on the DFT cross correlation analysis to come source discrimination for the deficiencies in the prior art.
Technical scheme of the present invention is as follows:
A kind of digital photograph of DFT cross correlation analysis comes source discrimination, may further comprise the steps:
(1) at first reference and test photograph collection are carried out R, G and the B Color Channel is cut apart: photo I can be divided into I
i, i ∈ { R, G, B}; Select a BLS-GSM small echo low-pass filter to I
i, { R, G, B} carry out filtering and obtain reference photo I without PRNU i ∈
i Rj, j=1,2...n and test photo I
i Tk, k=1,2...m;
(2) the former passage Ii of corresponding photo and without PRNUI
i Rj, I
i Tk, subtract each other and obtain residual image n
i r, n
i t
(3) to each path computation with reference to PRNU w
Ci, { B} and fusion are with reference to PRNU w for R, G for i ∈
cCalculating formula is as follows:
w
c=0.3*w
cr+0.6*w
cg+0.1*w
cb;
(4) calculate fusion with reference to PRNU w
cWith test pattern residual noise n
tCrosscorrelation coefficient c
WnCalculating formula is as follows:
In the formula, N and M are the horizontal and vertical pixel counts of photo, and (*) conjugation is got in expression, and capitalization is denoted as the DFT conversion that corresponding small letter represents; The DFT varying type is:
(5) locate at last crosscorrelation coefficient maximum value position, and calculate it to the Euclidean distance of initial point; Determine decision threshold; Formula is as follows:
m(x
0,y
0)=max||c
rn(j,k)||j=1K?M,k=1KN
In the formula, (x
0, y
0) be crosscorrelation coefficient maximum value position, Th is decision threshold, ρ (x0, y0)>Th represents not to be what this camera was taken.
Described method, select the method for decision threshold Th to be: by central limit theorem as can be known, not the approximate Gaussian distributed of the same digital camera photo crossing dependency numerical digit Euclidean distance value of putting, test the pass that its crossing dependency numerical digit of photo puts cumulative distribution density function, Gaussian distribution density function and the gauss of distribution function of Euclidean distance value for n and be:
Get Th=z, make: P
FA=1-c (Th) minimum.
Utilizing digital camera output photo to extract camera merges with reference to PRNU, merge the crosscorrelation coefficient peak with test photo residual noise with reference to PRNU with the method location of frequency domain cross correlation analysis, show by experiment the photo residual noise of digital camera of the same race and merge with reference to PRNU to reach crosscorrelation coefficient peak value at nearly initial point place, and the photo residual noise of variety classes digital camera and fusion reach crosscorrelation coefficient peak value with reference to PRNU at initial point place far away.The photo that the characteristic information that the present invention uses is taken the digital camera of model different brands in 7 identification of originating, average recognition rate has reached 96.3%.Therefore, this method has higher using value.
Description of drawings
Fig. 1 is the process flow diagram of the inventive method;
Fig. 2 is that C1 reference model noise model distributes with the Euclidean distance of testing photo.
Embodiment
Below in conjunction with specific embodiment, the present invention is described in detail.
A kind of digital photograph of DFT cross correlation analysis comes source discrimination, and method flow diagram such as Fig. 1 may further comprise the steps:
(1) at first reference and test photograph collection are carried out R, G and the B Color Channel is cut apart.Photo I can be divided into I
i, i ∈ { R, G, B}.Select a BLS-GSM (Bayes Least Squares-Gaussian Scale Mixtures) small echo low-pass filter to I
i, { R, G, B} carry out filtering and obtain reference photo I without PRNU i ∈
i Rj, j=1,2...n and test photo I
i Tk, k=1,2...m.
(2) the former passage Ii of corresponding photo and without PRNUI
i Rj, I
i Tk, subtract each other and obtain residual image n
i r, n
i t
(3) to each path computation with reference to PRNU w
Ci, { B} and fusion are with reference to PRNU w for R, G for i ∈
cCalculating formula is as follows:
w
c=0.3*w
cr+0.6*w
cg+0.1*w
cb
(4) calculate fusion with reference to PRNU w
cWith test pattern residual noise n
tCrosscorrelation coefficient c
WnCalculating formula is as follows:
In the formula, N and M are the horizontal and vertical pixel counts of photo, and (*) conjugation is got in expression, and capitalization is denoted as the DFT conversion that corresponding small letter represents.The DFT varying type is:
(5) locate at last crosscorrelation coefficient maximum value position and calculate its Euclidean distance to initial point.Determine decision threshold.Formula is as follows:
m(x
0,y
0)=max||c
rn(j,k)||j=1K?M,k=1KN
In the formula, (x
0, y
0) be crosscorrelation coefficient maximum value position, Th is decision threshold, ρ (x0, y0)>Th represents not to be what this camera was taken.
(6) by selecting decision threshold Th so that the error rate that identification is striven for is minimum.By central limit theorem as can be known, not the approximate Gaussian distributed of Euclidean distance value that same digital camera photo crossing dependency numerical digit is put.The pass that n test photo its crossing dependency numerical digit put cumulative distribution density function, Gaussian distribution density function and the gauss of distribution function of Euclidean distance value is:
Get Th=z, make: P
FA=1-c (Th) minimum.
Experiment and analysis:
In order to verify algorithm complexity in this paper, reduce calculated amount, the photo in the experiment is only got 512 * 512 pixels in center.7 each 170 photos (totally 1190 width of cloth) that camera is taken in this experiment, have been adopted, camera and photo sample parameter such as table 1, in order to meet the actual conditions in the life, photo content and the environment on every side of every kind of camera shooting all should enrich as far as possible, comprising: landscape, building, indoor and outdoors, different illumination conditions etc.Every camera selects 50 width of cloth (totally 350 width of cloth) photo to extract usefulness as camera reference model noise model, and then remaining 120 photos (totally 840 width of cloth) are as test sample book.
Table 1 camera and photo sample parameter
As can be seen from Figure 2 the test photo taken of C1 camera and its reference model noise model are located frequency domain crosscorrelation coefficient at nearly initial point O (0,0) and are reached peak value, and the photo that other cameras are taken reaches peak value at initial point O (0,0) far away.As seen can effectively identify the photo origin camera with crosscorrelation coefficient peak feature herein.By being tested, each digital camera obtains the photo origin recognition correct rate that each digital camera as shown in table 2 is taken.
Table 2 algorithm identified accuracy rate
Should be understood that, for those of ordinary skills, can be improved according to the above description or conversion, and all these improvement and conversion all should belong to the protection domain of claims of the present invention.
Claims (2)
1. the digital photograph of a DFT cross correlation analysis comes source discrimination, it is characterized in that, may further comprise the steps:
(1) at first reference and test photograph collection are carried out R, G and the B Color Channel is cut apart: photo I can be divided into I
i, i ∈ { R, G, B}; Select a BLS-GSM small echo low-pass filter to I
i, { R, G, B} carry out filtering and obtain reference photo I without PRNU i ∈
i Rj, j=1,2...n and test photo I
i Tk, k=1,2...m;
(2) the former passage I of corresponding photo
iWith without PRNUI
i Rj, I
i Tk, subtract each other and obtain residual image n
i r, n
i t
(3) to each path computation with reference to PRNU w
Ci, { B} and fusion are with reference to PRNU w for R, G for i ∈
cCalculating formula is as follows:
w
c=0.3*w
cr+0.6*w
cg+0.1*w
cb.
(4) calculate fusion with reference to PRNU w
cWith test pattern residual noise n
tCrosscorrelation coefficient c
WnCalculating formula is as follows:
In the formula, N and M are the horizontal and vertical pixel counts of photo, and (*) conjugation is got in expression, and capitalization is denoted as the DFT conversion that corresponding small letter represents; The DFT varying type is:
(5) locate at last crosscorrelation coefficient maximum value position, and calculate it to the Euclidean distance of initial point; Determine decision threshold; Formula is as follows:
m(x
0,y
0)=max||c
m(j,k)||j=1K?M,k=1KN
In the formula, (x
0, y
0) be crosscorrelation coefficient maximum value position, Th is decision threshold, ρ (x
0, y
0)>Th represents not to be what this camera was taken.
2. method according to claim 1, it is characterized in that, select the method for decision threshold Th to be: by central limit theorem as can be known, not the approximate Gaussian distributed of the same digital camera photo crossing dependency numerical digit Euclidean distance value of putting, test the pass that its crossing dependency numerical digit of photo puts cumulative distribution density function, Gaussian distribution density function and the gauss of distribution function of Euclidean distance value for n and be:
Get Th=z, make: P
FA=1-c (Th) minimum.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210499692XA CN103020641A (en) | 2012-11-18 | 2012-11-18 | Digital photo source distinguishing method based on DFT (Discrete Fourier Transform) cross correlation analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210499692XA CN103020641A (en) | 2012-11-18 | 2012-11-18 | Digital photo source distinguishing method based on DFT (Discrete Fourier Transform) cross correlation analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103020641A true CN103020641A (en) | 2013-04-03 |
Family
ID=47969231
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210499692XA Pending CN103020641A (en) | 2012-11-18 | 2012-11-18 | Digital photo source distinguishing method based on DFT (Discrete Fourier Transform) cross correlation analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103020641A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107735783A (en) * | 2015-06-23 | 2018-02-23 | 都灵理工学院 | Method and apparatus for searching for image |
CN108154080A (en) * | 2017-11-27 | 2018-06-12 | 北京交通大学 | A kind of method that video equipment is quickly traced to the source |
CN112989308A (en) * | 2021-05-12 | 2021-06-18 | 腾讯科技(深圳)有限公司 | Account authentication method, device, equipment and medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6148270A (en) * | 1996-10-30 | 2000-11-14 | Yamatake-Honeywell Co., Ltd. | Fast target distance measuring device and high-speed moving image measuring device |
US20070237399A1 (en) * | 2006-04-07 | 2007-10-11 | Fuji Xerox Co., Ltd. | Failure analysis system, failure analysis method, and program product for failure analysis |
CN101441720A (en) * | 2008-11-18 | 2009-05-27 | 大连理工大学 | Digital image evidence obtaining method for detecting photo origin by covariance matrix |
CN101616238A (en) * | 2009-07-17 | 2009-12-30 | 中山大学 | A kind of digitized authentication image method of digital camera |
-
2012
- 2012-11-18 CN CN201210499692XA patent/CN103020641A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6148270A (en) * | 1996-10-30 | 2000-11-14 | Yamatake-Honeywell Co., Ltd. | Fast target distance measuring device and high-speed moving image measuring device |
US20070237399A1 (en) * | 2006-04-07 | 2007-10-11 | Fuji Xerox Co., Ltd. | Failure analysis system, failure analysis method, and program product for failure analysis |
CN101441720A (en) * | 2008-11-18 | 2009-05-27 | 大连理工大学 | Digital image evidence obtaining method for detecting photo origin by covariance matrix |
CN101616238A (en) * | 2009-07-17 | 2009-12-30 | 中山大学 | A kind of digitized authentication image method of digital camera |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107735783A (en) * | 2015-06-23 | 2018-02-23 | 都灵理工学院 | Method and apparatus for searching for image |
CN107735783B (en) * | 2015-06-23 | 2021-10-15 | 都灵理工学院 | Method and apparatus for searching image |
CN108154080A (en) * | 2017-11-27 | 2018-06-12 | 北京交通大学 | A kind of method that video equipment is quickly traced to the source |
CN108154080B (en) * | 2017-11-27 | 2020-09-01 | 北京交通大学 | Method for quickly tracing to source of video equipment |
CN112989308A (en) * | 2021-05-12 | 2021-06-18 | 腾讯科技(深圳)有限公司 | Account authentication method, device, equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Cozzolino et al. | Noiseprint: A CNN-based camera model fingerprint | |
Chen et al. | Shadow information recovery in urban areas from very high resolution satellite imagery | |
CN106097379B (en) | It is a kind of to use the distorted image detection of adaptive threshold and localization method | |
CN103561271B (en) | The moving target of static camera shooting is removed video spatial domain altering detecting method | |
CN108319986B (en) | Image source identification method and system based on PRNU | |
Villalba et al. | Smartphone image clustering | |
CN108154080B (en) | Method for quickly tracing to source of video equipment | |
CN103530638A (en) | Method for matching pedestrians under multiple cameras | |
CN105184308B (en) | Remote sensing image building detection classification method based on global optimization decision | |
CN102567738B (en) | Rapid detection method for pornographic videos based on Gaussian distribution | |
Zhang et al. | Improved Fully Convolutional Network for Digital Image Region Forgery Detection. | |
CN106683074B (en) | A kind of distorted image detection method based on haze characteristic | |
Qiao et al. | Identifying individual camera device from raw images | |
CN103020641A (en) | Digital photo source distinguishing method based on DFT (Discrete Fourier Transform) cross correlation analysis | |
Mehrish et al. | Robust PRNU estimation from probabilistic raw measurements | |
Fan et al. | Modeling the EXIF-image correlation for image manipulation detection | |
CN113537088A (en) | Digital image sensor hardware fingerprint matching method and system based on small sample learning | |
CN103561274A (en) | Video time domain tamper detection method for removing moving object shot by static camera lens | |
CN111861976B (en) | Digital image source shooting equipment identification method based on hardware fingerprint correlation | |
Chen et al. | Hybrid saliency detection for images | |
CN104408736A (en) | Characteristic-similarity-based synthetic face image quality evaluation method | |
Zhang et al. | Exposing digital image forgeries by using canonical correlation analysis | |
Sun et al. | The detecting system of image forgeries with noise features and EXIF information | |
CN111325185B (en) | Face fraud prevention method and system | |
CN111080586B (en) | Method for obtaining evidence of tampered image source based on convolutional neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130403 |