CN110348464A - A kind of image forge detection algorithm based on more support area local luminance sequences - Google Patents
A kind of image forge detection algorithm based on more support area local luminance sequences Download PDFInfo
- Publication number
- CN110348464A CN110348464A CN201910647407.6A CN201910647407A CN110348464A CN 110348464 A CN110348464 A CN 110348464A CN 201910647407 A CN201910647407 A CN 201910647407A CN 110348464 A CN110348464 A CN 110348464A
- Authority
- CN
- China
- Prior art keywords
- support area
- characteristic
- image
- liop
- characteristic point
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
Abstract
The invention discloses a kind of image forge detection algorithms based on more support area local luminance sequences, belong to image forge detection technique field, the detection including S1, characteristic area determines characteristic point;The transformation of S2, characteristic area;The building of LIOP description in S3, each characteristic area;S4, characteristic matching;S5, characteristic point are sorted out;S6, geometric transformation estimation;S7, detection are completed.Son is described by using LIOP and is divided using global brightness ordered pair support area, this division does not need to calculate the principal direction of support area, not only save calculation amount, and it can theoretically guarantee that constructing description has real rotational invariance and monotonic luminance invariance, while obtaining multiple support areas of different scale, different resolution and different directions using NSCT to improve the taste of LIOP description.To improve the robust of image forge Region detection algorithms.
Description
Technical field
It is the present invention relates to image forge detection technique field, in particular to a kind of based on more support area local luminance sequences
Image forge detection algorithm.
Background technique
Digital picture has very important role in current communication process.With digital imaging processing software and modification
The development of equipment, digital picture can easily be tampered without leave it is any significantly distort trace, even if layman
Image editing tools (such as Photoshop) can be readily available to modify existing image.The quantity of image operation control and forgery
Also in rapid growth, this judges that the originality of piece image and accuracy bring great puzzlement to people, especially to judgement department
The true and false in method identification as evidence image is more challenging.Therefore, qualification figure seem it is no be forged be it is vital,
Scene of a crime exploration, judicial expertise and many other fields can be widely used in.
Image forge detection technique refers to can be confirmed the credible of original image in the case where no any priori knowledge
Degree.Common digital picture, which forges method, resampling, splicing, copy-paste etc., and wherein copy-paste forgery is most simple
With the most common digital image tampering method.Copy-paste forgery refers to that the region of arbitrary shape and size is replicated in image
Be then placed within another region of image, the purpose is to by enhance image visual effect come cover image significant data or
Influence the real goal of identification image.Since replication region is in same image, essential attribute such as noise, color, texture
Deng all consistent with general image, this causes discrimination process extremely troublesome.Replication region will do it certain ruler before stickup simultaneously
Degree and rotation process, this, which makes detection forge region, has certain challenge.
The image copy-paste counterfeiting detection method based on characteristic point is increasingly taken seriously in recent years, and such methods are being permitted
Stronger convincingness is proven to have under more image transformation.Image copy-paste counterfeiting detection method based on characteristic point is main
Consider characteristic point region and indicate suspicious region using Feature Descriptor, common Feature Descriptor has Scale invariant special
Sign transformation and acceleration robust features.However these methods in order to make building Feature Descriptor have rotational invariance, it is necessary to count
It calculates the principal direction of support area and is corrected support area according to the direction, the calculating that this not only considerably increases algorithm is multiple
Miscellaneous degree, and inevitably generate error.In addition, these methods make the Feature Descriptor of building using normalization operation
There is certain invariance to linear luminance variation, but non-linear brightness variation issue cannot be handled well (as dullness is bright
Degree variation).
Summary of the invention
The object of the invention is that in order to solve above-mentioned image forge detection algorithm calculate it is complicated, be easy to produce error with
And the problem of non-linear brightness changes cannot be handled well and a kind of image based on more support area local luminance sequences is provided
Detection algorithm is forged, having reduces calculation amount, the advantages of reducing error, solve the problems, such as luminance transformation.
The present invention is achieved through the following technical solutions above-mentioned purpose, a kind of figure based on more support area local luminance sequences
As forging detection algorithm, comprising the following steps:
The detection of S1, characteristic area determine characteristic point: eliminating the influence of noise first with Gaussian filter, recycle maximum
Stable extremal region (MSER) algorithm extracts the maximum stable extremal region of image as support area, and the center of support area
Point is characterized a little;
The transformation of S2, characteristic area: the spatial information of detection zone is obtained, is obtained using non-sample Contourlet transformation
More support areas of different scale, different resolution and different directions;
The building of LIOP description in S3, each characteristic area: non-drop is carried out to brightness values all in each support area
Sequence, is equally spaced divided into B sub-regions according to brightness value size for each support area, by superposition each subregion
The LIOP value of all pixels point obtains description of each subregion, obtains each by description for all subregions of connecting
Hold final LIOP description in region;
S4, characteristic matching: LIOP description in two neighboring characteristic point is carried out using two-way Euclidean distance ratio method
Matching obtains one group of characteristic matching pair if matching;
S5, characteristic point are sorted out: being sorted out using space clustering method to characteristic point;
S6, geometric transformation estimation: to step S5 each classification, using random sampling consistency (RANSAC) algorithm into
Row geometric transformation estimation, further rejects inaccurate matching pair;
S7, detection are completed: being calculated optimal convex hull, fillet to one group of characteristic point in class and are carried out morphology operations, obtain
To forgery region.
Preferably, the non-sample Contourlet transformation of the step S2 is by the tower-like filter (NSP) of non-sampling and non-pumping
Sample directional filters group (NSDFB) composition, and the condition of NSP and NSDFB energy Perfect Reconstruction signal are as follows:
H0(z)G0(z)+H1(z)G1(z)=1
Wherein H0(z)、H1(z) resolution filter, G are indicated0(z)、G1(z) reconstruction filter is indicated.
Preferably, the step S3 is for a characteristic point X0, support area is R, any one in the support area
Pixel X ∈ R can establish the coordinate system of an invariable rotary, i.e., withFor x-axis, perpendicular toFor y-axis, then exist
Under invariable rotary coordinate system, on unit circle with, for starting point, obtained at equal intervals in positive y-axis pixel X N number of neighborhood point (i.e. away from
The N number of point nearest from X).
Compared with prior art, the beneficial effects of the present invention are: describing son by using LIOP and utilizing global brightness sequence
Support area is divided, this division does not need to calculate the principal direction of support area, not only saves calculation amount, but also resonable
Son is described with real rotational invariance and monotonic luminance invariance by can above guarantee to construct, while being obtained using NSCT
Multiple support areas of different scale, different resolution and different directions describe sub taste to improve LIOP, reduce meter
It calculates complexity and is effectively prevented from error, finally improve the robustness of image forge Region detection algorithms.
Detailed description of the invention
Fig. 1 is image forge detection algorithm flow chart of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
As shown in Figure 1, a kind of image forge detection algorithm based on more support area local luminance sequences, including following step
It is rapid:
The detection of S1, characteristic area determine characteristic point: eliminating the influence of noise first with Gaussian filter, recycle maximum
Stable extremal region (MSER) algorithm extracts the maximum stable extremal region of image as support area, and the center of support area
Point is characterized a little;
The transformation of S2, characteristic area: the spatial information of detection zone is obtained, is obtained using non-sample Contourlet transformation
More support areas of different scale, different resolution and different directions;
The building of LIOP description in S3, each characteristic area: non-drop is carried out to brightness values all in each support area
Sequence, is equally spaced divided into B sub-regions according to brightness value size for each support area, by superposition each subregion
The LIOP value of all pixels point obtains description of each subregion, obtains each by description for all subregions of connecting
Hold final LIOP description in region;
S4, characteristic matching: LIOP description in two neighboring characteristic point is carried out using two-way Euclidean distance ratio method
Matching obtains one group of characteristic matching pair if matching;
S5, characteristic point are sorted out: being sorted out using space clustering method to characteristic point;
S6, geometric transformation estimation: to step S5 each classification, using random sampling consistency (RANSAC) algorithm into
Row geometric transformation estimation, further rejects inaccurate matching pair;
S7, detection are completed: being calculated optimal convex hull, fillet to one group of characteristic point in class and are carried out morphology operations, obtain
To forgery region.
Preferably, the non-sample Contourlet transformation of the step S2 is by the tower-like filter (NSP) of non-sampling and non-pumping
Sample directional filters group (NSDFB) composition, and the condition of NSP and NSDFB energy Perfect Reconstruction signal are as follows:
H0(z)G0(z)+H1(z)G1(z)=1
Wherein H0(z)、H1(z) resolution filter, G are indicated0(z)、G1(z) reconstruction filter is indicated.
Step S3 is for a characteristic point X0, support area is R, and any one pixel X ∈ R is equal in the support area
Can establish the coordinate system of an invariable rotary, i.e., withFor x-axis, perpendicular toFor y-axis, then in invariable rotary coordinate
Under system, N number of neighborhood point (i.e. nearest N number of of distance X of pixel X for starting point, is obtained in positive y-axis at equal intervals on unit circle
Point).
If the feature point set (i.e. the center of MSER algorithm detection zone) of image to be detected is T:
T={ t1,t2,…,tn} (1)
Wherein each characteristic point ti∈ T, i=1,2 ..., the corresponding region MSER n, available corresponding more supports
Region LIOP describes sub- wi, and then obtain Feature Descriptor set ψ:
ψ={ w1,w2,…,wn} (2)
To each Feature Descriptor wi∈ ψ, i=1,2 ..., n, the most simple common two-way Euclidean distance ratio method of use
It is matched.I.e. if characteristic point tAAnd tBIt is a pair of of matching pair, then has
Euclidean distance, w are asked in wherein d () expressionAAnd wBIt is t respectivelyAAnd tBCorresponding LIOP description, and wBAnd wCIt is
Distance wANearest and secondary close Feature Descriptor, while guaranteeing wAAnd wDIt is distance wBRecently and secondary close Feature Descriptor, η are
Threshold value takes 0.6.
It can be obtained by one group of characteristic matching pair: { m in this way1,m2,…,mqAnd { p1,p2,…,pq, wherein characteristic point mr
With prIt is a pair of of characteristic matching, mr∈T,pr∈ T, r=1,2 ..., q.Next utilize space length by { m1,m2,…,mqOr
{p1,p2,…,pqSorted out, calculate first each characteristic point apart from probability density E (mi):
Being chosen using formula (5) makes E (mi) the smallest characteristic point m0As initial point, characteristic point then is determined by formula (6)
ml, l=1,2 ..., q whether with m0For one kind.
||ml-m0||≤δ (6)
Wherein δ is decision threshold, related with distorted image regional location, selects 200.On similarly remaining characteristic point repeats
Process is stated until all characteristic points are all sorted out, { p1,p2,…,pqClassification and { m1,m2,…,mqCorrespond.
To each classification, geometric transformation estimation is carried out using RANSAC algorithm, it is inaccurate the purpose is to further reject
Matching pair, make detection forgery region it is more accurate.The algorithm rejects inaccurate matching pair using formula (7):
WhereinWithIt is the homogeneous coordinate system of matching characteristic point, | | | |2Indicate the operation of 2- norm, threshold epsilon takes
3, U be the homography matrix between class, using affine transformation estimation, i.e.,
WhereinIt is the transformation such as rotation and scale, s in control coordinate transformxAnd syIt is x and y-axis in coordinate transform
Translational movement.Necessary post-processing is finally carried out, i.e., optimal convex hull, fillet are calculated to one group of characteristic point in class and carries out shape
State student movement is calculated, and obtains forging region.
The detection method of the detection method and SURF of choosing this algorithm and SIFT compares the standard this it appears that this algorithm
True rate and recall ratio are above other algorithms
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (3)
1. a kind of image forge detection algorithm based on more support area local luminance sequences, which comprises the following steps:
The detection of S1, characteristic area determine characteristic point: eliminating the influence of noise first with Gaussian filter, recycle maximum stable
Extremal region (MSER) algorithm extracts the maximum stable extremal region of image as support area, and the central point of support area is
Characteristic point;
The transformation of S2, characteristic area: obtaining the spatial information of detection zone, obtains difference using non-sample Contourlet transformation
More support areas of scale, different resolution and different directions;
The building of LIOP description in S3, each characteristic area: carrying out non-drop to brightness values all in each support area and sort,
Each support area is equally spaced divided into B sub-regions according to brightness value size, by owning in superposition each subregion
The LIOP value of pixel obtains description of each subregion, obtains each support area by description for all subregions of connecting
The final LIOP in domain describes son;
S4, characteristic matching: matching LIOP description in two neighboring characteristic point using two-way Euclidean distance ratio method,
One group of characteristic matching pair is obtained if matching;
S5, characteristic point are sorted out: being sorted out using space clustering method to characteristic point;
S6, geometric transformation estimation: each classification to step S5 is carried out several using random sampling consistency (RANSAC) algorithm
What transformation estimation, further rejects inaccurate matching pair;
S7, detection are completed: being calculated optimal convex hull, fillet to one group of characteristic point in class and are carried out morphology operations, obtain puppet
Make region.
2. a kind of image forge detection algorithm based on more support area local luminance sequences according to claim 1, special
Sign is: the non-sample Contourlet transformation of the step S2 is by the tower-like filter (NSP) of non-sampling and non-sampling directionality
Filter group (NSDFB) composition, and the condition of NSP and NSDFB energy Perfect Reconstruction signal are as follows:
H0(z)G0(z)+H1(z)G1(z)=1 (1)
Wherein H0(z)、H1(z) resolution filter, G are indicated0(z)、G1(z) reconstruction filter is indicated.
3. a kind of image forge detection algorithm based on more support area local luminance sequences according to claim 1, special
Sign is: the step S3 is for a characteristic point X0, support area is R, any one pixel X ∈ in the support area
R can establish the coordinate system of an invariable rotary, i.e., withFor x-axis, perpendicular toFor y-axis, then in invariable rotary
Under coordinate system, (i.e. distance X is nearest for the N number of neighborhood point for for starting point, obtaining pixel X in positive y-axis at equal intervals on unit circle
N number of point).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910647407.6A CN110348464A (en) | 2019-07-17 | 2019-07-17 | A kind of image forge detection algorithm based on more support area local luminance sequences |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910647407.6A CN110348464A (en) | 2019-07-17 | 2019-07-17 | A kind of image forge detection algorithm based on more support area local luminance sequences |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110348464A true CN110348464A (en) | 2019-10-18 |
Family
ID=68176693
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910647407.6A Pending CN110348464A (en) | 2019-07-17 | 2019-07-17 | A kind of image forge detection algorithm based on more support area local luminance sequences |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110348464A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111768368A (en) * | 2020-05-26 | 2020-10-13 | 西安理工大学 | Image area copying and tampering detection method based on maximum stable extremal area |
CN115035533A (en) * | 2022-08-10 | 2022-09-09 | 新立讯科技股份有限公司 | Data authentication processing method and device, computer equipment and storage medium |
-
2019
- 2019-07-17 CN CN201910647407.6A patent/CN110348464A/en active Pending
Non-Patent Citations (1)
Title |
---|
颜普等: "基于多支持区域局部亮度序的图像伪造检测", 《计算机应用》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111768368A (en) * | 2020-05-26 | 2020-10-13 | 西安理工大学 | Image area copying and tampering detection method based on maximum stable extremal area |
CN111768368B (en) * | 2020-05-26 | 2022-05-03 | 西安理工大学 | Image area copying and tampering detection method based on maximum stable extremal area |
CN115035533A (en) * | 2022-08-10 | 2022-09-09 | 新立讯科技股份有限公司 | Data authentication processing method and device, computer equipment and storage medium |
CN115035533B (en) * | 2022-08-10 | 2022-10-21 | 新立讯科技股份有限公司 | Data authentication processing method and device, computer equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Gao et al. | Automatic change detection in synthetic aperture radar images based on PCANet | |
WO2019169816A1 (en) | Deep neural network for fine recognition of vehicle attributes, and training method thereof | |
CN104933721B (en) | Stitching image altering detecting method based on color filter array characteristic | |
CN103426186B (en) | A kind of SURF fast matching method of improvement | |
CN109214420A (en) | The high texture image classification method and system of view-based access control model conspicuousness detection | |
CN109740572B (en) | Human face living body detection method based on local color texture features | |
CN104809731B (en) | A kind of rotation Scale invariant scene matching method based on gradient binaryzation | |
Vaishnavi et al. | Application of local invariant symmetry features to detect and localize image copy move forgeries | |
CN107506765B (en) | License plate inclination correction method based on neural network | |
CN104217221A (en) | Method for detecting calligraphy and paintings based on textural features | |
CN103400384A (en) | Large viewing angle image matching method capable of combining region matching and point matching | |
CN104408728A (en) | Method for detecting forged images based on noise estimation | |
CN106327534A (en) | Tire inner wall texture identification method based on locating block | |
CN108550165A (en) | A kind of image matching method based on local invariant feature | |
Fan et al. | Registration of multiresolution remote sensing images based on L2-siamese model | |
CN114863464B (en) | Second-order identification method for PID drawing picture information | |
CN103353941A (en) | Natural marker registration method based on viewpoint classification | |
CN110348464A (en) | A kind of image forge detection algorithm based on more support area local luminance sequences | |
CN111275070B (en) | Signature verification method and device based on local feature matching | |
CN106709915B (en) | Image resampling operation detection method | |
CN108491857A (en) | A kind of multiple-camera target matching method of ken overlapping | |
CN109711420B (en) | Multi-affine target detection and identification method based on human visual attention mechanism | |
Jwaid et al. | Study and analysis of copy-move & splicing image forgery detection techniques | |
Changjie et al. | Algorithm of remote sensing image matching based on corner-point | |
Soni et al. | Improved block-based technique using surf and fast keypoints matching for copy-move attack detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191018 |
|
RJ01 | Rejection of invention patent application after publication |