KR101755980B1 - Copy-Move Forgery Detection method and apparatus based on scale space representation - Google Patents

Copy-Move Forgery Detection method and apparatus based on scale space representation Download PDF

Info

Publication number
KR101755980B1
KR101755980B1 KR1020160011496A KR20160011496A KR101755980B1 KR 101755980 B1 KR101755980 B1 KR 101755980B1 KR 1020160011496 A KR1020160011496 A KR 1020160011496A KR 20160011496 A KR20160011496 A KR 20160011496A KR 101755980 B1 KR101755980 B1 KR 101755980B1
Authority
KR
South Korea
Prior art keywords
match
pair
image
calculating
match pair
Prior art date
Application number
KR1020160011496A
Other languages
Korean (ko)
Inventor
박천수
Original Assignee
세종대학교산학협력단
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 세종대학교산학협력단 filed Critical 세종대학교산학협력단
Priority to KR1020160011496A priority Critical patent/KR101755980B1/en
Application granted granted Critical
Publication of KR101755980B1 publication Critical patent/KR101755980B1/en

Links

Images

Classifications

    • G06K9/00577—
    • G06K9/4671—
    • G06K9/6201—
    • G06K9/64—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G06T7/248—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving reference images or patches

Abstract

A method and apparatus for similar region detection based on scale space representation are disclosed. A similar region detection method based on a scale space representation includes extracting feature points for each key point in an image by applying SIFT; Generating a matched pair set for each keypoint using each of the extracted minutiae; Selecting a subset for the match pair at random in the match pair and calculating a geometric transformation metric based on the affine transform using the selected subset; And generating a distorted image based on the geometric distortion metric, and detecting a similar region by applying a zero mean normalized cross-correlation (ZNCC) between the original image and the distorted image.

Description

Field of the Invention < RTI ID = 0.0 > [0002] < / RTI &

Field of the Invention [0002] The present invention relates to a similar region detection method and apparatus capable of detecting similar regions based on a scale-space representation.

The remarkable development of digital image editing software makes it very easy to edit digital images. Many digital images are created every day, but most of these digital images can be generated without including separate digital watermarks or signatures. Digital images can be authored in a wide variety of ways. The copy-move forger (CMF) is one of the most common methods of digital image tampering.

In such a CMF, the tempered region undergoes post-processing such as rotation, scaling, blurring, and noising, and thus does not exactly coincide with the other regions. This makes it increasingly difficult to find a tempered area in a digital image.

The present invention is to provide a similar region detection method and apparatus capable of easily detecting a similar region copied to a geometric transformation.

According to an aspect of the present invention, there is provided a similar region detection method capable of easily detecting a similar region copied in a geometric transformation.

According to an embodiment of the present invention, there is provided a method of detecting a similar region in an image processing apparatus, comprising: extracting feature points for each key point in an image by applying SIFT; Generating a matched pair set for each keypoint using each of the extracted minutiae; Selecting a subset for the match pair at random in the match pair and calculating a geometric transformation metric based on the affine transform using the selected subset; And a step of generating a distorted image based on the geometric distortion metric and detecting a similar region by applying a zero mean normalized cross-correlation (ZNCC) between the original image and the distorted image, A detection method can be provided.

The generating of the match pair may include: generating a match candidate list by calculating a distance ratio between extracted minutiae for each key point; And deriving a match pair of keypoints included in the mutual match candidate list through a cross check between the match candidate lists of the keypoints, calculating a distance ratio between the match pairs, And a step of constructing the set.

The step of constructing the match pair may be configured by calculating the distance ratios between the match pairs and then adding the match pair to the match pair in descending order to limit it to M (natural number).

Wherein selecting the subset comprises:

Selecting a first match pair at random in the match pair; And

And selecting a second and a third match pair having the same scale ratio as the first match pair and having a same length ratio of the line segment in the match pair set.

The geometric transformation scale can be calculated by calculating the affine transform using the first match pair, the second match pair, and the third match pair.

Wherein the detecting the similar region comprises: calculating a zero mean normalized cross-correlation between the original image and the distorted image; And generating the binary image using the calculated zero-mean normalized cross-correlation to detect the similar region.

The step of detecting the similar region may remove an isolated region below the reference value in the binary image, and fill holes below the reference value using a mathematical morphological operation.

According to another aspect of the present invention, there is provided an image processing apparatus capable of detecting a similar region that can easily detect a similar region copied in a geometric transformation.

According to an embodiment of the present invention, a feature point extracting unit extracts feature points for each key point in an image using SIFT; A sampling unit for generating a matched pair set for each keypoint using the extracted minutiae; A transform calculation unit for randomly selecting a subset for the match pair in the match pair and calculating a geometric transform scale based on the affine transform using the selected subset; And a detector for detecting a similar region by applying a zero mean normalized cross-correlation (ZNCC) between the original image and the distorted image after generating the distorted image based on the geometric distortion metric, A device may be provided.

The sampling unit generates a match candidate list by calculating a distance ratio between extracted minutiae for each key point and derives a match pair from keypoints included in the mutual match candidate list through cross check between the match candidate lists of the key points After calculating the distance ratios between the match pairs, a match pair set can be configured using the distance ratios between the match pairs.

Wherein the transformation calculator selects a first match pair randomly in the match pair and generates a second match pair having a same scale ratio of the first match pair and a same length ratio of the line segments, A match pair may be selected from the match pair set to construct the subset to calculate the geometric transformation metric.

The detector calculates a zero-mean normalized cross-correlation between the original image and the distorted image, generates a binary image using the calculated zero-mean normalized cross-correlation, and then removes an isolated region below the reference value in the binary image And holes below the reference value can be filled using a mathematical morphological operation.

It is possible to quickly and easily detect a similar region copied to a geometric transformation by providing a similar region detection method and apparatus capable of detecting a similar region based on a scale-space representation according to an embodiment of the present invention. can do.

1 is a flowchart illustrating a method of detecting a similar region of an image in an image processing apparatus according to an embodiment of the present invention.
FIG. 2 illustrates a sequence of keypoint matching according to an embodiment of the present invention; FIG.
3 is a diagram illustrating a length of a line segment according to an embodiment of the present invention;
4 is a block diagram schematically illustrating an internal configuration of an image processing apparatus according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating a similar region detection result according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The present invention is capable of various modifications and various embodiments, and specific embodiments are illustrated in the drawings and described in detail in the detailed description. It is to be understood, however, that the invention is not to be limited to the specific embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. In addition, numerals (e.g., first, second, etc.) used in the description of the present invention are merely an identifier for distinguishing one component from another.

Also, in this specification, when an element is referred to as being "connected" or "connected" with another element, the element may be directly connected or directly connected to the other element, It should be understood that, unless an opposite description is present, it may be connected or connected via another element in the middle.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a flowchart illustrating a method of detecting a similar region of an image in an image processing apparatus according to an exemplary embodiment of the present invention. FIG. 2 is a view illustrating a process of keypoint matching according to an exemplary embodiment of the present invention. And FIG. 3 is a view for explaining a length of a line segment according to an embodiment of the present invention.

In operation 110, the image processing apparatus 100 detects keypoints for the target image and extracts feature points for keypoints by applying Scale Invariant Feature transform (SIFT). Here, since the SIFT algorithm is obvious to those skilled in the art, a detailed description thereof will be omitted.

For example, the image processing apparatus 100 can detect a local extremity extrema in a scale space with a key point, and can detect the SFIT feature descriptor with respect to the detected key point.

In operation 115, the image processing apparatus 100 generates a match candidate list for each key point using the extracted minutiae.

Due to the SIFT characteristic, each keypoint can be matched with a myriad of feature points. Accordingly, the image processing apparatus 100 can construct a similarity vector D for each key point, and use it to generate a match candidate list.

The image processing apparatus 100 may calculate a distance ratio. At this time, the distance ratio can be calculated using Equation (1).

Figure 112016010006671-pat00001

here,

Figure 112016010006671-pat00002
,
Figure 112016010006671-pat00003
Represent keypoints or feature points, respectively.

In calculating the distance ratio, the image processing apparatus 100 starts i = F, decrements i by 1, and repeats it until it reaches 1. [ Where F may be the maximum number of overlapping regions and i may be a natural number.

In addition, the image processing apparatus 100

Figure 112016010006671-pat00004
, The process of calculating the distance ratio can be terminated.

For example, suppose that the distance ratio calculation process is ended when i = m (natural number). At this time, the match candidate list for the current key point is

Figure 112016010006671-pat00005
As shown in FIG.

For example, if m = 1, then the match candidate list for that keypoint will contain a single matched key-point. However, if m> 1, then the match candidate list for the keypoint includes a plurality of matched keypoints.

In operation 120, the image processing apparatus 100 constructs a match pair between keypoints using a match candidate list configured for each keypoint.

The process of configuring a match pair between each keypoint will now be described in detail.

The image processing apparatus 100 cross-checks the match candidate lists of the keypoints to eliminate keypoints having low reliability.

For example, if the current keypoint is

Figure 112016010006671-pat00006
.
Figure 112016010006671-pat00007
In the match candidate list C, the current key point < RTI ID = 0.0 >
Figure 112016010006671-pat00008
And the reliability can be tested.

if

Figure 112016010006671-pat00009
Figure 112016010006671-pat00010
Is regarded as an actual matched pair. However,
Figure 112016010006671-pat00011
, Then the pair is not considered an actual matched pair.

Then, the image processing apparatus 100 calculates the distance ratios of the match pairs regarded as matched pairs, and inserts the distance ratios into the match pair set R in descending order.

The ratio of the distance between each pair of matrices can be calculated as shown in Equation (2).

Figure 112016010006671-pat00012

In order to construct a match pair for each keypoint, the process of calculating the distance ratio between each match pair is in a tradeoff relationship between computational complexity and detection performance. Therefore, the ratio of distances between the respective match pairs may be calculated up to M (natural numbers).

FIG. 2 illustrates a sequence of keypoint matching according to an exemplary embodiment of the present invention.

In step 125, the image processing apparatus 100 randomly selects a subset in the match pair R and calculates a geometric transformation metric for the selected subset.

In order to detect keypoints stably in the scale space, the SIFT uses a scale space representation implemented as an image pyramid. The initial image repeats the smoother including the Gaussian blur and is subsampled to obtain a high level of the pyramid.

The difference-of-Gaussian (DoG) image can be calculated by subtracting the adjacent image scale. For example, the DoG image can be calculated using Equation (3).

Figure 112016010006671-pat00013

here,

Figure 112016010006671-pat00014
Represents a scale level,
Figure 112016010006671-pat00015
The
Figure 112016010006671-pat00016
Of Gaussian blur (
Figure 112016010006671-pat00017
) ≪ / RTI >
Figure 112016010006671-pat00018
). ≪ / RTI >

DoG image (

Figure 112016010006671-pat00019
), Each sample point is compared to eight neighbors in the current scale and compared to nine neighbors in the scale. After this process, a larger point than all neighbors can be selected as a key point.

Each octave of the scale space can be divided by n (integer), and the resulting interval is

Figure 112016010006671-pat00020
. To detect the local extremum
Figure 112016010006671-pat00021
DoG images need to be calculated for each octave.

For example, suppose that the octave and blur levels for extracted keypoint k are o and l, respectively. The scale level for the key point k extracted by the given o and l can be calculated.

For example, the scale level may be calculated using Equation (4).

Figure 112016010006671-pat00022

Match Pairs Set Matches in R

Figure 112016010006671-pat00023
However,
Figure 112016010006671-pat00024
Wow
Figure 112016010006671-pat00025
The
Figure 112016010006671-pat00026
Wow
Figure 112016010006671-pat00027
Pixel coordinates. At this time, the variation ratio of the scale level to the matched pair p can be calculated as shown in Equation (5).

Figure 112016010006671-pat00028

here,

Figure 112016010006671-pat00029
Wow
Figure 112016010006671-pat00030
Respectively
Figure 112016010006671-pat00031
Wow
Figure 112016010006671-pat00032
. ≪ / RTI >

In general, in the CMF scenario, the local area can be geometrically deformed and copied to other areas. Accordingly, the image processing apparatus 100 according to an embodiment of the present invention uses an affine transform to model a geometric transformation between a source and a copied region.

For example, assume that two match pairs that can be generated by a CMF attack are p1 and p2.

The two match pairs share a common geometric transformation and the scale parameters between the two match pairs are the same. Therefore, the relationship between the two match pairs can be summarized as shown in Equation (6).

Figure 112016010006671-pat00033

here,

Figure 112016010006671-pat00034
and
Figure 112016010006671-pat00035
Represents the strain rate of the scale level of p1 and p2.

FIG. 3 illustrates a diagram illustrating a line segment of a circle region and a copied region.

Since the affine transform forms the length ratio of the line segment, the length of the segment deformed using the original segment can be roughly predicted.

This is expressed by Equation (7).

Figure 112016010006671-pat00036

The image processing apparatus 100 selects three non-collinear pairs to calculate the affine transform.

First, the image processing apparatus 100 randomly selects the first pair (

Figure 112016010006671-pat00037
).

Then, the image processing apparatus 100 generates a second match pair (step < RTI ID = 0.0 >

Figure 112016010006671-pat00038
) From the match pair R. < RTI ID = 0.0 >

Figure 112016010006671-pat00039

Where E represents a threshold for a preset reprojection error.

Likewise, the image processing apparatus 100 can select the third match pair.

Thus, when three match pairs are selected, the image processing apparatus 100 calculates the geometric transformation scale using the affine transform.

The image processing apparatus 100 models a geometric transformation for a copied region including various geometric transformations such as rotation, scaling, shearing, and reflection.

Let us define a 2 x 2 linear matrix A as shown in equation (9).

Figure 112016010006671-pat00040

here,

Figure 112016010006671-pat00041
Is a parameter describing rotation and scaling deformation.

The relationship between the matched pairs can be described as Equation (10).

Figure 112016010006671-pat00042

here,

Figure 112016010006671-pat00043
ego,
Figure 112016010006671-pat00044
Represents a distortion factor.

The image processing apparatus 100 acquires the unique solution of Equation (10) using the three selected match pairs. In particular, the solution of equation (10) can be obtained using the maximum likelihood prediction.

After calculating the affine transforms of A and t, the image processing apparatus 100 can calculate the number of inner laminates in the match pair R satisfying the constraint of Equation (11).

Figure 112016010006671-pat00045

In operation 130, the image processing apparatus 100 calculates the geometric transformation scale for the selected subset (three match pairs) and detects the overlapping region.

For example, the image processing apparatus 100 generates a distorted image W for each transformation matrix.

Then, the image processing apparatus 100 defines overlapping regions using zero mean normalized cross-correlation (ZNCC) between the original image and the distorted image.

This can be expressed by Equation (12).

Figure 112016010006671-pat00046

Where B represents a set of seven neighboring pixels,

Figure 112016010006671-pat00047
and
Figure 112016010006671-pat00048
Represents the intensity of the average pixel of the original image I and the distorted image W. [

Next, the image processing apparatus 100

Figure 112016010006671-pat00049
To generate a binary image, to discard a small isolated area, and to fill a small hole using a mathematical morphological operation.

Accordingly, in the embodiment of the present invention, the image processing apparatus 100 can detect the overlapped area.

FIG. 4 is a block diagram schematically illustrating an internal configuration of an image processing apparatus according to an embodiment of the present invention.

4, an image processing apparatus 100 according to an exemplary embodiment of the present invention includes a feature extraction unit 410, a sampling unit 415, a transformation calculation unit 420, a detection unit 425, a memory 430, And a processor 435.

The feature extraction unit 410 is a unit for extracting feature points for each key point in the image by applying SIFT.

The sampling unit 415 is means for generating a matched pair set for each key point using each extracted feature point.

For example, the sampling unit 415 generates a match candidate list by calculating the distance ratios between extracted minutiae for each key point, and searches for keypoints included in the mutual match candidate list through cross check between the match candidate lists of the keypoints After deriving a match pair, a distance ratio between the match pairs is calculated, and then a match pair set can be configured using the distance ratio between the match pairs.

The transformation calculator 420 is a means for randomly selecting a subset for a match pair in the match pair and calculating a geometric transformation scale based on the affine transform using the selected subset.

The detection unit 425 generates a distorted image based on the geometric distortion scale, and then detects a similar region by applying zero mean normalized cross-correlation (ZNCC) between the original image and the distorted image .

The memory 430 is a means for storing various applications necessary for performing a method for detecting a similar region based on the scale space representation according to an embodiment of the present invention, various data derived from the process, and the like.

The processor 435 may include internal components (e.g., a feature extraction unit 410, a sampling unit 415, a deformation calculation unit 420, a memory 410, and a memory 410) of the image processing apparatus 100 according to an embodiment of the present invention. (430), etc.).

FIG. 5 is a diagram illustrating a similar region detection result according to an embodiment of the present invention.

5 (a) shows a test image, (b) shows a result of detection of a conventional SIFT-T, and FIG. 5 (c) (D) shows the detection result of the conventional SIFT-2, and (e) shows the detection result according to the embodiment of the present invention. As shown in FIG. 5, it can be seen that the similar region detection method according to an embodiment of the present invention detects the similar region according to the geometric transformation more accurately than the conventional method.

The method of detecting similar regions based on the above-described scale space representation according to the present invention can be implemented as computer-readable codes on a computer-readable recording medium. The computer-readable recording medium includes all kinds of recording media storing data that can be decoded by a computer system. For example, it may be a ROM (Read Only Memory), a RAM (Random Access Memory), a magnetic tape, a magnetic disk, a flash memory, an optical data storage device, or the like. In addition, the computer-readable recording medium may be distributed and executed in a computer system connected to a computer network, and may be stored and executed as a code readable in a distributed manner.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the following claims And changes may be made without departing from the spirit and scope of the invention.

410: Feature extraction unit
415:
420:
425:
430: Memory
435: Processor

Claims (12)

A method for detecting a similar region in an image processing apparatus,
Extracting feature points for each keypoint from the image by applying SIFT;
Generating a matched pair set for each keypoint using each of the extracted minutiae;
Selecting a subset for the match pair at random in the match pair and calculating a geometric transformation metric based on the affine transform using the selected subset; And
Generating a distorted image based on the geometric distortion scale, and detecting a similar region by applying a zero mean normalized cross-correlation (ZNCC) between the original image and the distorted image,
Wherein selecting the subset comprises:
Selecting a first match pair at random in the match pair; And
And selecting a second match pair and a third match pair having the same scale ratio as the first match pair and having a same length ratio of the line segments in the match pair, Detection method.

The method according to claim 1,
Wherein generating the match pair comprises:
Generating a match candidate list by calculating a distance ratio between extracted minutiae for each key point; And
After calculating the distance ratios between the match pairs after deriving the match pairs from the key points included in the mutual match candidate list through the cross check between the match candidate lists of the key points, Wherein the similarity detection step comprises:
3. The method of claim 2,
Wherein configuring the match pair comprises:
Wherein a ratio of distances between the pair of matched pairs is calculated and then added to the matched pair set in a descending order so as to be limited to M (natural number).
delete The method according to claim 1,
And calculating the geometric transformation scale by calculating an affine transform using the first match pair, the second match pair, and the third match pair.
The method according to claim 1,
Wherein the step of detecting the similar region comprises:
Calculating a zero mean normalized cross-correlation between the original image and the distorted image; And
And generating a binary image using the calculated zero-mean normalized cross-correlation to detect the similar region.
The method according to claim 6,
Wherein the step of detecting the similar region comprises:
Wherein an isolated region below the reference value is removed from the binary image, and a hole below the reference value is filled using a mathematical morphological operation.
A computer-readable recording medium having recorded thereon a program code for performing the method according to any one of claims 1 to 3 and 5 to 7.
A feature extraction unit for extracting feature points for each key point in the image by applying SIFT;
A sampling unit for generating a matched pair set for each keypoint using the extracted minutiae;
A transformation calculator for randomly selecting a subset for the match pair in the match pair and calculating a geometric transformation metric based on the affine transform using the selected subset; And
And a detector for detecting a similar region by applying a zero mean normalized cross-correlation (ZNCC) between the original image and the distorted image after generating a distorted image based on the geometric distortion metric,
Wherein the deformation calculating unit
Selecting a first match pair randomly in the match pair and comparing a second match pair and a third match pair having the same scale ratio of the first match pair and a same length ratio of line segments, Wherein the geometric distortion metric is calculated by constructing the subset by selecting from a pair set.
10. The method of claim 9,
Wherein the sampling unit comprises:
Generating a match candidate list by calculating a distance ratio between extracted minutiae for each key point, deriving a match pair of key points included in the mutual match candidate list through cross checking between the match candidate lists of the key points, And calculates a distance ratio between the match pairs, and then constructs the match pair using the ratio of distances between the match pairs.
delete 10. The method of claim 9,
Wherein:
Calculating a zero mean normalized cross-correlation between the original image and the distorted image, generating a binary image using the calculated zero-mean normalized cross-correlation, removing an isolated region below the reference value in the binary image, Characterized in that the holes of the image are filled using mathematical morphological operations.

KR1020160011496A 2016-01-29 2016-01-29 Copy-Move Forgery Detection method and apparatus based on scale space representation KR101755980B1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020160011496A KR101755980B1 (en) 2016-01-29 2016-01-29 Copy-Move Forgery Detection method and apparatus based on scale space representation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020160011496A KR101755980B1 (en) 2016-01-29 2016-01-29 Copy-Move Forgery Detection method and apparatus based on scale space representation

Publications (1)

Publication Number Publication Date
KR101755980B1 true KR101755980B1 (en) 2017-07-10

Family

ID=59356452

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020160011496A KR101755980B1 (en) 2016-01-29 2016-01-29 Copy-Move Forgery Detection method and apparatus based on scale space representation

Country Status (1)

Country Link
KR (1) KR101755980B1 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108109141A (en) * 2017-12-18 2018-06-01 辽宁师范大学 Based on the matched homologous partial copy detection method of super-pixel multiple features
CN108564814A (en) * 2018-06-06 2018-09-21 清华大学苏州汽车研究院（吴江） A kind of parking position detection method and device based on image
CN109816706A (en) * 2019-02-01 2019-05-28 辽宁工程技术大学 A kind of smoothness constraint and triangulation network equal proportion subdivision picture are to dense matching method
CN110599478A (en) * 2019-09-16 2019-12-20 中山大学 Image area copying and pasting tampering detection method

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101554811B1 (en) * 2014-01-13 2015-09-21 한국과학기술원 Method and system for detecting copying and moving modulation about local area of image
JP5796611B2 (en) * 2013-08-28 2015-10-21 株式会社リコー Image processing apparatus, image processing method, program, and imaging system

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5796611B2 (en) * 2013-08-28 2015-10-21 株式会社リコー Image processing apparatus, image processing method, program, and imaging system
KR101554811B1 (en) * 2014-01-13 2015-09-21 한국과학기술원 Method and system for detecting copying and moving modulation about local area of image

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108109141A (en) * 2017-12-18 2018-06-01 辽宁师范大学 Based on the matched homologous partial copy detection method of super-pixel multiple features
CN108109141B (en) * 2017-12-18 2021-11-19 辽宁师范大学 Homologous local replication detection method based on superpixel multi-feature matching
CN108564814A (en) * 2018-06-06 2018-09-21 清华大学苏州汽车研究院（吴江） A kind of parking position detection method and device based on image
CN109816706A (en) * 2019-02-01 2019-05-28 辽宁工程技术大学 A kind of smoothness constraint and triangulation network equal proportion subdivision picture are to dense matching method
CN110599478A (en) * 2019-09-16 2019-12-20 中山大学 Image area copying and pasting tampering detection method
CN110599478B (en) * 2019-09-16 2023-02-03 中山大学 Image area copying and pasting tampering detection method

Similar Documents

Publication Publication Date Title
Melekhov et al. Siamese network features for image matching
US10019646B2 (en) Method for object recognition and/or verification on portable devices
Amerini et al. Geometric tampering estimation by means of a SIFT-based forensic analysis
US10147018B2 (en) Image processing apparatus, image processing method, and storage medium
Davarzani et al. Copy-move forgery detection using multiresolution local binary patterns
Kwon et al. Learning jpeg compression artifacts for image manipulation detection and localization
KR102048638B1 (en) Method and system for recognizing content
CN106530194B (en) Method and device for detecting suspected infringement product picture
KR101755980B1 (en) Copy-Move Forgery Detection method and apparatus based on scale space representation
CN105069457B (en) Image recognition method and device
Ulutas et al. Frame duplication detection based on bow model
EP2235680A1 (en) Invariant visual scene and object recognition
Kim et al. Adaptive weighted fusion with new spatial and temporal fingerprints for improved video copy detection
CN106663195A (en) Improved method for fingerprint matching and camera identification, device and system
Park et al. Fast and robust copy-move forgery detection based on scale-space representation
Soni et al. Keypoints based enhanced multiple copy‐move forgeries detection system using density‐based spatial clustering of application with noise clustering algorithm
Brogan et al. Spotting the difference: Context retrieval and analysis for improved forgery detection and localization
US8913851B1 (en) Fingerprinting image using points of interest for robust image identification
Li et al. SIFT keypoint removal via directed graph construction for color images
JP6793925B2 (en) Verification equipment, methods, and programs
Maigrot et al. Tampering detection and localization in images from social networks: A CBIR approach
US20140270541A1 (en) Apparatus and method for processing image based on feature point
Mahdi et al. Detection of Copy-Move Forgery in Digital Image Based on SIFT Features and Automatic Matching Thresholds
Nataraj et al. Seam carving detection and localization using two-stage deep neural networks
Al-Qershi et al. Comparison of matching methods for copy-move image forgery detection

Legal Events

Date Code Title Description
E701 Decision to grant or registration of patent right
GRNT Written decision to grant