CN107392949A - Image zone duplicating and altering detecting method based on local invariant feature - Google Patents
Image zone duplicating and altering detecting method based on local invariant feature Download PDFInfo
- Publication number
- CN107392949A CN107392949A CN201710579410.XA CN201710579410A CN107392949A CN 107392949 A CN107392949 A CN 107392949A CN 201710579410 A CN201710579410 A CN 201710579410A CN 107392949 A CN107392949 A CN 107392949A
- Authority
- CN
- China
- Prior art keywords
- image
- point
- characteristic
- line segment
- detected
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000001514 detection method Methods 0.000 claims abstract description 38
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 17
- 238000002187 spin decoupling employing ultra-broadband-inversion sequences generated via simulated annealing Methods 0.000 claims abstract description 14
- 239000013598 vector Substances 0.000 claims description 42
- 238000009416 shuttering Methods 0.000 claims description 29
- 230000001629 suppression Effects 0.000 claims description 10
- 239000000203 mixture Substances 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 abstract description 4
- 238000012545 processing Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 230000010076 replication Effects 0.000 description 3
- 230000009977 dual effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000009394 selective breeding Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
- G06T7/337—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving reference images or patches
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of image zone duplicating and altering detecting method based on local invariant feature, set of characteristic points is calculated by improved SUSAN Corner Detection Algorithms first, then feature description is carried out to characteristic point by a kind of local tertiary mode, the result that color compares is divided into three classes, the details of preferably reaction texture variations is allowed to, finally by characteristic matching detection image with the presence or absence of duplication tampered region.The inventive method can be more careful description image grain distribution, improve image zone duplicating and altering detection accuracy.
Description
Technical field
The present invention relates to computer visual image processing technology field, it is specifically related to a kind of based on local invariant feature
Image zone duplicating and altering detecting method.
Background technology
It is commonly used with various image processing softwares, people all kinds of digital pictures can be carried out easily processing and
Editor, cause occur a large amount of artificial tampered images on network.These artificial tampered images to scientific research, law evidence obtaining,
Insurance claim and medium public's trust etc. adversely affect, therefore carry out authenticity verification with very for digital picture
Important realistic meaning.
In various distorted image modes, region duplication operation is the most widely used to distort one of means.By multiple
A certain region in imaged is simultaneously pasted it on another disjoint region in same piece image, to reach covering mesh
Mark information or forge the purpose of scene.Distorted for such, 2 identical areas can be whether there is in the image by detecting
Domain, judge the image whether by artificially distorting.
The region duplication altering detecting method of existing digital picture：
1) method based on blocking characteristic matching：Mainly include feature matching method and base based on scattered cosine transform DCT
In the detection algorithm of principal component analysis, this kind of method is typically fairly simple, but passes through geometry for replication region in tampered image
The situation detection of conversion such as rotation, scaling and upset is invalid.
2) method of distinguished point based detection and matching, feature is carried out by the characteristic point in detection image first and to it
Description, then positioned by matching algorithm and replicate sticking area.Typically compared using SIFT and SURF etc. feature point detection part
Complicated algorithm, although ability of these algorithms with certain resistance replication region rotation and scale transformation.But exist tight
The defects of weight：Replicate sticking area and must have the covering of a number of characteristic point, if replication region have it is non-significant
Visual structure, then tampered region may be by complete missing inspection.
The content of the invention
In order to solve the above-mentioned technical problem, the invention provides a kind of image zone duplicating based on local invariant feature to usurp
Change detection method.
The present invention provides a kind of image zone duplicating and altering detecting method based on local invariant feature, including following step
Suddenly：
Step S100：Feature point detection is carried out to all pixels point in image to be detected using SUSAN algorithms, obtains spy
Point set is levied, is designated as { (x1,y1),…,(xN,yN)}；
Step S200：Set of characteristic points is traveled through with partial 3 d pattern, obtained corresponding with each characteristic point each for describing
The image feature vector of characteristic point；
Step S300：Travel through set of characteristic points and carry out image feature vector and characteristic point bi-directional matching, obtain each characteristic point
Similar features point pair, the Euclidean distance between the characteristic point of each similar features point centering is minimum, and by similar features point to collection
Conjunction is designated as { [(xai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t represents the number finally matched；
Step S400：By each pair similar features point to being connected with line segment in image to be detected, and calculate every line
The slope f of sectioniAnd fi-fi+1, judge fi-fi+1Whether in -0.05~0.005, this line segment is designated as unanimously if meeting
Line segment, it is unsatisfactory for, does not mark；
Step S500：The quantity of consistent line segment is counted, as slope quantity, judges that the slope quantity in image to be detected is
No≤50, image to be detected is duplicating image if meeting, image to be detected distorts area in the absence of duplication if being unsatisfactory for
Domain,
The remaining line segment in addition to consistent line segment in duplicating image is deleted, calculates the line segment distance of wantonly two consistent line segment, and is united
The consistent line segment quantity that line segment distance is less than 100 is counted, whether judge the consistent line segment quantity of gained >=40, it is to be detected if meeting
Image, which exists, replicates tampered region, is otherwise not present in image to be detected and replicates tampered region.
Further, step S100 comprises the following steps：
Step S110：Image to be detected is traveled through with circular shuttering, calculates the USAN at each pixel in image to be detected
Value u (i, j)；
Step S120：To USAN values u (i, j) thresholding of all pixels in testing image, obtain characteristic point response Rx (i,
j)；
Rx (i, j)=max (0, th2-u (i, j)) (4)
Wherein, th2 is gained threshold value；
Step S130：All pixels point in traversal image to be detected judges whether USAN value u (i, j) meet u (i, j) ＜
The point is then labeled as preliminary characteristic point by th2, satisfaction, otherwise carries out step S120 to next pixel, and gained is all preliminary
Feature point set is combined into preliminary set of characteristic points；
Step S140：Preliminary set of characteristic points is traveled through with non-maxima suppression, acquired results are assembled into set of characteristic points,
It is designated as { (x1,y1),…,(xN,yN)}；
Circular shuttering is all satisfactions (x-x1) centered on (x1, y1) pixel2+(y-y1)2≤ 10 pixel (x,
Y) region of composition.
Further, step S200 comprises the following steps：To wait to appoint the characteristic point (x takenn,yn) centered on, by image P (x,
Y, z) on 7 × 7 image-region be designated as characteristic area Fn(x, y, z), traversal characteristic area FnEach pixel in (x, y, z),
First Model Comparison, second Model Comparison and the 3rd Model Comparison are sequentially carried out to each pixel, obtain feature to
Measure { RT1 ..., RT3 }, image feature vector will be obtained after characteristic vector set obtained by each pixel.
Another aspect of the present invention additionally provides a kind of image-region based on local invariant feature of method described above
Tampering detection apparatus is replicated, including：
Feature point detection module：For carrying out characteristic point to all pixels point in image to be detected using SUSAN algorithms
Detection, obtains set of characteristic points, is designated as { (x1,y1),…,(xN,yN)}；
Feature vector module：For traveling through set of characteristic points with partial 3 d pattern, use corresponding with each characteristic point is obtained
In the image feature vector for describing each characteristic point；
Similar features point is to module：Image feature vector and characteristic point bi-directional matching are carried out for traveling through set of characteristic points,
Obtaining the similar features point pair of each characteristic point, the Euclidean distance between the characteristic point of each similar features point centering is minimum, and by phase
{ [(x is designated as to set like characteristic pointai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t represents the number finally matched；
Consistent line segment module：For in image to be detected by each pair similar features point to being connected with line segment, and count
Calculate the slope f of every line segmentiAnd fi-fi+1, judge fi-fi+1Whether in -0.05~0.005, this line segment if meeting
Consistent line segment is designated as, is unsatisfactory for, is not marked；
Judge module：For counting the quantity of consistent line segment, as slope quantity, the slope number in image to be detected is judged
Amount whether≤50, if meet if image to be detected be duplicating image, image to be detected is usurped in the absence of duplication if being unsatisfactory for
Change region,
The remaining line segment in addition to consistent line segment in duplicating image is deleted, calculates the line segment distance of wantonly two consistent line segment, and is united
The consistent line segment quantity that line segment distance is less than 100 is counted, whether judge the consistent line segment quantity of gained >=40, it is to be detected if meeting
Image, which exists, replicates tampered region, is otherwise not present in image to be detected and replicates tampered region.
Further, feature point detection module includes：
USAN value modules：For traveling through image to be detected with circular shuttering, calculate in image to be detected at each pixel
USAN value u (i, j)；
Threshold module：For USAN values u (i, j) thresholding to all pixels in testing image, characteristic point response is obtained
Rx(i,j)；
Rx (i, j)=max (0, th2-u (i, j)) (4)
Wherein, th2 is gained threshold value；
Preliminary feature point module：Judge whether USAN value u (i, j) are full for traveling through the point of all pixels in image to be detected
The point is then labeled as preliminary characteristic point, otherwise enters threshold module to next pixel, by institute by sufficient u (i, j) ＜ th2, satisfaction
Obtain all preliminary feature point sets and be combined into preliminary set of characteristic points；
Non-maxima suppression module：For traveling through preliminary set of characteristic points with non-maxima suppression, by acquired results set
Into set of characteristic points, { (x is designated as1,y1),…,(xN,yN)}；
Circular shuttering is all satisfactions (x-x1) centered on (x1, y1) pixel2+(y-y1)2≤ 10 pixel (x,
Y) region of composition.
The technique effect of the present invention：
The present invention provides the image zone duplicating and altering detecting method based on local invariant feature, passes through SUSAN angles first
Point detection algorithm calculates set of characteristic points, then carries out feature description to characteristic point by local tertiary mode, color is compared
Result be divided into three classes, be allowed to preferably reaction texture variations details, whether there is finally by characteristic matching detection image
Replicate tampered region.The inventive method can be more careful description image grain distribution, improve image zone duplicating and altering
The accuracy of detection.
The present invention provides the image zone duplicating and altering detecting method based on local invariant feature, can be more accurately to face
Color comparison information is encoded.Feature description it is more accurate, behind characteristic matching result it is better, whole result certainly it is more accurate
Really.
Specific refer to proposes according to the image zone duplicating and altering detecting method based on local invariant feature of the present invention
Various embodiments it is described below, will cause apparent in terms of the above and other of the present invention.
Brief description of the drawings
Fig. 1 is the image zone duplicating and altering detecting method flow signal provided by the invention based on local invariant feature
Figure；
Fig. 2 is the circular shuttering used in operator in SUSAN methods provided by the invention；
Fig. 3 is the image zone duplicating and altering detection means schematic diagram provided by the invention based on local invariant feature.
Embodiment
The accompanying drawing for forming the part of the application is used for providing a further understanding of the present invention, schematic reality of the invention
Apply example and its illustrate to be used to explain the present invention, do not form inappropriate limitation of the present invention.
Referring to Fig. 1, the present invention proposes a kind of image zone duplicating and altering detecting method based on local invariant feature, wraps
Include following steps：
Step S100：Feature point detection is carried out to all pixels point in image to be detected using SUSAN algorithms, obtains spy
Point set is levied, is designated as { (x1,y1),…,(xN,yN)}；
The color RGB image that note needs to carry out region duplication tampering detection is P (x, y, z), method category proposed by the present invention
In this kind of method that distinguished point based is detected and matched, therefore firstly the need of progress feature point detection, low level image procossing
The method (i.e. SmallUnivalReSegmentAssimilatingNucleus, abbreviation SUSAN algorithm) in the similar area of small nut value,
It is a kind of classical angular-point detection method, it can extract the angle point and edge feature of target in compared with very noisy, and position
Accurately.
Preferably, the present invention has carried out some improvement to SUSAN algorithms, with stronger Stability and veracity,
Comprise the following steps：
Step S110：Image to be detected is traveled through with circular shuttering, calculated in described image to be detected at each pixel
USAN value u (i, j)；
Circular shuttering used is defined as follows：Center pixel is designated as (x1, y1), all to meet condition (x-x1)2+(y-y1)2≤
The region of 10 pixel (x, y) composition is the scope of circular shuttering.In the present embodiment, circular shuttering is as shown in Fig. 2 circle
Template size is 7 × 7, altogether 37 pixels, and the pixel of wherein label 19 is center pixel.Other can also be used existing often
Use template.
For any pixel (i0, j0) in image P (x, y, z), the center pixel of circular shuttering is placed on (i0, j0),
Then calculate each pixel in image P (x, y, z) inside circular shuttering position pixel color value and center pixel (i0,
J0 the difference of color value), if the pixel color value of a pixel and the face of center pixel (i0, j0) in image to be detected
Difference between colour is less than or equal to the similarity degree threshold value th1 of setting, then the pixel belongs to USAN regions；Otherwise the pixel
Point is not belonging to USAN regions.
By the above method so as to which whether each pixel judged in image to be detected belongs to USAN regions, in existing method
He Tongzhi areas formula is absorbed to be expressed as follows：
C (i, j) represents whether the pixel belongs to USAN regions.
Wherein, th1 (i0, j0) represents the threshold value of similarity degree, typically one constant value of artificial selection as threshold value,
This largely constrains it and is automatically processing the application of aspect.
The method that a kind of automatic selected threshold of energy is proposed to this present invention, similarity degree threshold value th1 is calculated by formula (2)
(i0,j0)：
Wherein, Ω1Represent the contiguous range that size is 3 × 3 centered on (i0, j0).
Then, all pixels point c (i, j) being in motion vector figure Vx (i, j) in circular shuttering position is carried out
Statistics：
Wherein, Ω2Pixel point set in expression motion vector figure Vx (i, j) in circular shuttering position, u (i0,
J0) be pixel (i0, j0) USAN values.
Using circular shuttering traversing graph as Vx (i, j) all pixels, the USAN value u (i, j) of all pixels are obtained.
Step S120：To USAN values u (i, j) thresholding of all pixels in the testing image, characteristic point response is obtained
Rx(i,j)；
Rx (i, j)=max (0, th2-u (i, j)) (4)
Wherein, th2 is gained threshold value；Value is 28 in the present embodiment.
Step S130：Travel through the point of all pixels in described image to be detected judge USAN value u (i, j) whether meet u (i,
The point is then labeled as preliminary characteristic point by j) ＜ th2, satisfaction, otherwise carries out step S120 to next pixel, gained is owned
Preliminary feature point set is combined into preliminary set of characteristic points；
All pixels point in traversal image to be detected judges whether USAN value u (i, j) meet u (i, j) ＜ th2, meets
The point is then labeled as preliminary characteristic point, step S120 otherwise is carried out to lower pixel；By all preliminary set of characteristic points of gained
For preliminary set of characteristic points.
During due to u (i, j) ＜ th2, Rx (i, j) is only possible to be more than 0, and will meet the point is the preliminary of preliminary judgement
Characteristic point.
Step S140：The preliminary set of characteristic points is traveled through with non-maxima suppression, acquired results are assembled into the spy
Point set is levied, is designated as { (x1,y1),…,(xN,yN)}；
For each preliminary judgement in preliminary set of characteristic points characteristic point such as (i1, j1), observe centered on it,
In the field of size 5 × 5, if also have other pixels Rx (i, j) value than Rx (i1, j1) greatly, if not provided, namely Rx
(i1, j1) is maximum, then retains this characteristic point (i1, j1)；Otherwise this characteristic point (i1, j1) is deleted, also i.e. by Rx (i, j)
Reset to 0.All preliminary characteristic points in preliminary set of characteristic points are handled, final all Rx (i, j) are more than 0 picture
Element set is set of characteristic points.
Step S200：The set of characteristic points is traveled through with partial 3 d pattern, obtains use corresponding with each characteristic point
In the image feature vector for describing each characteristic point；
By existing partial 3 d pattern.
Preferably, existing local binary mould is encoded by the information of color comparison chart picture, but dual mode is only
Simple magnitude relationship is characterized, the difference degree of texture variations can not be reacted, and is easily affected by noise.Thus in order to gram
These problems are taken, the present invention is extended to local binary, and the result that color compares is divided into 3 classes, so as to preferably anti-
Answer the details of texture variations.
The step S200 comprises the following steps：To appoint the characteristic point (x takenn,yn) centered on, by 7 on image P (x, y, z)
× 7 image-region is designated as characteristic area Fn(x, y, z), traversal characteristic area FnEach pixel in (x, y, z), to each
Pixel sequentially carries out first Model Comparison, second Model Comparison and the 3rd Model Comparison, obtains characteristic vector
{ RT1 ..., RT3 }, image feature vector will be obtained after characteristic vector set obtained by each pixel.
Appoint below and take a characteristic point (xn,yn) exemplified by illustrate.On image P (x, y, z), with characteristic point (xn,yn)
Centered on, the image-region that size is 7 × 7 is designated as characteristic area Fn(x, y, z), therefore characteristic point (xn,yn) and characteristic area Fn
(x, y, z) is to correspond.Characteristic point (xn,yn) information be all reflected in characteristic area FnOn (x, y, z), therefore by spy
Levy region Fn(x, y, z) carries out feature statement, to realize to characteristic point (xn,yn) purpose that is described.
Characteristic area Fn(x, y, z) shares 49 pixels, and the present invention carries out same processing side to each pixel
Method, to be illustrated exemplified by one of pixel (x0, y0),
Centered on (x0, y0), image P (x, y, z) select its surrounding close to 8 pixels be designated as { (x01,
y01),…,(x08,y08) contrast pixel as the point, then carry out first Model Comparison by formula (5)：
Wherein, m ∈ { 1,2 ..., 8 } represent pixel serial number, and th1 represents first discrepancy threshold.Taken in the present embodiment
Value 10, r1 (m) represent pixel (xm,ym) coding under first pattern.Final 8 pixels can obtain the two of one 8
System number { r1 (1), r1 (2) ..., r1 (8) }, it is pixel (x0, y0) to be translated into the value obtained by decimal number
Characteristic value under one pattern, is designated as RT1.
Followed by the comparison of second pattern：
R2 (m) represents pixel (xm,ym) coding under first pattern.Same 8 pixels can obtain one 8
Binary number { r2 (1), r2 (2) ..., r2 (8) }, it is pixel (x0, y0) to be translated into the value obtained by decimal number
Characteristic value in a second mode, is designated as RT2.
Carry out the comparison of the 3rd pattern：
Wherein, r3 (m) represents pixel (xm,ym) coding under the 3rd pattern.Same 8 pixels can obtain one
The binary number { r3 (1), r3 (2) ..., r3 (8) } of individual 8, it is pixel to be translated into the value obtained by decimal number
The characteristic value of (x0, y0) under the 3rd pattern, is designated as RT3.
It can be seen that the distribution of texture has been carried out thinner division by these three patterns, by discrepancy threshold th1 to color
Comparison result carries out the classification of three kinds of situations, is no longer simple under local binary be more than or less than.Therefore the present invention carries
The character description method gone out can preferably characterize the details of grain distribution.
The characteristic vector { RT1 ..., RT3 } that a length is three is obtained in final pixel point (x0, y0) one.
To characteristic area Fn(x, y, z) all pixels, are all handled according to the method described above, obtain respective feature
Vector, all characteristic vectors is arranged together, it is final to obtain the characteristic vector that a length is 49 × 3=147
{ RT1 ..., RT147 }, namely complete to characteristic point (xn,yn) description.
According to the method described above, all characteristic points are described respectively, available length is 49 × 3=147's
Characteristic vector.
Step S300：Travel through set of characteristic points and carry out described image characteristic vector and the characteristic point bi-directional matching, obtain
The similar features point pair of each characteristic point, the Euclidean distance between the characteristic point of each similar features point centering is most
It is small, and the similar features point is designated as { [(x to setai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t represents final
The number matched somebody with somebody；
Following feature based vector matches to characteristic point, for set of characteristic points { (x1,y1),…,(xN,yN)}
Each characteristic point, the minimum characteristic point of distance is found inside set of characteristic points by bi-directional matching.
Appoint and take a characteristic point (xa,ya) exemplified by illustrate, characteristic vector corresponding to note is RTa, and it is special to calculate itself and other
The Euclidean distance of point is levied, then the minimum characteristic point of chosen distance is designated as (xb,yb) as its preliminary matches characteristic point.Then
With same method to characteristic point (xb,yb) handled, when the minimum characteristic point of distance is exactly (xa,ya) when, it is final to determine
Characteristic point (xa,ya) and characteristic point (xb,yb) it is a pair of similar features points, otherwise characteristic point (xa,ya) there is no similar features point.
Bi-directional matching is carried out to each characteristic point by the above method, a series of similar features points can be obtained, be designated as
{[(xai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t represents the number finally matched.
Step S400：By similar features point described in each pair to being connected with line segment in described image to be detected, and count
Calculate the slope f of every line segmentiAnd fi-fi+1, judge fi-fi+1Whether in -0.05~0.005, this line segment if meeting
Consistent line segment is designated as, is unsatisfactory for, is not marked；
Step S500：The quantity of the consistent line segment is counted, as slope quantity, judges the institute in described image to be detected
Whether state slope quantity≤50, described image to be detected is duplicating image if meeting, described to be detected if being unsatisfactory for
Image, which is not present, replicates tampered region,
The remaining line segment in addition to the consistent line segment in the duplicating image is deleted, calculates the line of the wantonly two consistent line segments
Segment distance, and the consistent line segment quantity of the line segment distance less than 100 is counted, whether judge the consistent line segment quantity of gained >=
40, described image to be detected, which exists, if meeting replicates tampered region, is otherwise usurped in described image to be detected in the absence of duplication
Change region.
After similar features point set is obtained, every a pair of similar features point is connected with line segment on image P (x, y, z)
Come, calculate the direction of every line segment, namely slope fi：
Wherein, { fi| i=1 ..., t } slope is represented, the difference of these slopes is counted, difference is in scope (- 0.05,0.05)
Interior line segment, the direction for regarding as line segment are basically identical.It is basically identical if there is at least 50 slopes, namely at least 50 lines
The direction of section is basically identical, then preliminary to judge duplication tampered region in image to be detected P (x, y, z) be present；Otherwise directly judge
Image P (x, y, z) is without duplication tampered region.
If after tentatively judging that image to be detected P (x, y, z) has duplication tampered region, retain on image P (x, y, z)
The basically identical line segment information of those slopes, deletes other line segments.Then the distance of these line segments is calculated, formula is as follows：
dij=| yai-yaj|+|xai-xaj| (9)
Wherein i, j represent the label information of line segment, dijRepresent the distance of two lines section.If there is between 40 line segments
Mutual distance is less than 100, then final to judge that image P (x, y, z) be present has duplication tampered region；Otherwise judge image P (x, y,
Z) without duplication tampered region.
Referring to Fig. 3, another aspect of the present invention additionally provides a kind of image based on local invariant feature of above method
Region duplication tampering detection apparatus, including：
Feature point detection module：For carrying out characteristic point to all pixels point in image to be detected using SUSAN algorithms
Detection, obtains set of characteristic points, is designated as { (x1,y1),…,(xN,yN)}；
Feature vector module：For traveling through the set of characteristic points with partial 3 d pattern, obtain and each characteristic point
The corresponding image feature vector for being used to describe each characteristic point；
Similar features point is to module：Described image characteristic vector and the characteristic point pair are carried out for traveling through set of characteristic points
To matching, the similar features point pair of each characteristic point is obtained, between the characteristic point of each similar features point centering
Euclidean distance is minimum, and the similar features point is designated as into { [(x to setai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t
Represent the number finally matched；
Consistent line segment module：For in described image to be detected by similar features point described in each pair to being connected with line segment
Come, and calculate the slope f of every line segmentiAnd fi-fi+1, judge fi-fi+1Whether in -0.05~0.005, if meeting
This line segment is designated as consistent line segment, is unsatisfactory for, and does not mark；
Judge module：For counting the quantity of the consistent line segment, as slope quantity, judge in described image to be detected
The slope quantity whether≤50, if meet if described image to be detected be duplicating image, it is described if being unsatisfactory for treat
Detection image, which is not present, replicates tampered region,
The remaining line segment in addition to the consistent line segment in the duplicating image is deleted, calculates the line of the wantonly two consistent line segments
Segment distance, and the consistent line segment quantity of the line segment distance less than 100 is counted, whether judge the consistent line segment quantity of gained >=
40, described image to be detected, which exists, if meeting replicates tampered region, is otherwise usurped in described image to be detected in the absence of duplication
Change region.
Preferably, feature point detection module includes：
USAN value modules：For traveling through image to be detected with circular shuttering, each pixel in described image to be detected is calculated
USAN value u (i, j) at point；
Threshold module：For USAN values u (i, j) thresholding to all pixels in the testing image, characteristic point is obtained
Respond Rx (i, j)；
Rx (i, j)=max (0, th2-u (i, j)) (4)
Wherein, th2 is gained threshold value；
Preliminary feature point module：All pixels point for traveling through in described image to be detected judges that USAN value u (i, j) are
No to meet u (i, j) ＜ th2, the point is then labeled as preliminary characteristic point, otherwise enters threshold module to next pixel by satisfaction,
All preliminary feature point sets of gained are combined into preliminary set of characteristic points；
Non-maxima suppression module：For traveling through the preliminary set of characteristic points with non-maxima suppression, by acquired results
The set of characteristic points is assembled, is designated as { (x1,y1),…,(xN,yN)}；
The circular shuttering is all satisfactions (x-x1) centered on (x1, y1) pixel2+(y-y1)2≤ 10 pixel
The region of point (x, y) composition.
Simulation example will be carried out by the condition in above-described embodiment below, specific example is done further specifically to the present invention
It is bright.
The color RGB image that note needs to carry out region duplication tampering detection is P (x, y, z), method category proposed by the present invention
In this kind of method that distinguished point based is detected and matched, therefore firstly the need of progress feature point detection, low level image procossing
The method (i.e. SmallUnivalReSegmentAssimilatingNucleus, abbreviation SUSAN algorithm) in the similar area of small nut value,
It is a kind of classical angular-point detection method, it can extract the angle point and edge feature of target in compared with very noisy, and position
Accurately.The present invention has carried out some improvement to SUSAN algorithms, and with stronger Stability and veracity, specific steps are such as
Under：
(1) using circular shuttering traversing graph as P (x, y, z), the value in the absorption He Tongzhi areas (USAN) at every place of calculating；
Circular shuttering is defined as follows：Center pixel is designated as (x1, y1), all to meet condition (x-x1)2+(y-y1)2≤ 10
The region of pixel (x, y) composition is the scope of circular shuttering.Circular shuttering is as shown in Fig. 2 circular shuttering size is 7 × 7, one
Totally 37 pixels, the pixel of wherein label 19 is center pixel.For image P (x, y, z) any pixel (i0, j0), by circle
The center pixel of shape template is placed on (i0, j0), then calculates each picture in image P (x, y, z) inside circular shuttering position
The difference of the pixel color value of element and the color value of center pixel (i0, j0), if the pixel color value of a pixel and center
Difference between the color value of pixel (i0, j0) is less than or equal to the threshold value th1 of the similarity degree of setting, then the pixel belongs to
USAN regions；Otherwise the pixel is not belonging to USAN regions.By the above method so as to judging whether pixel belongs to USAN areas
Domain, specific formula are as follows：
C (i, j) represents whether the pixel belongs to USAN regions.Wherein th1 (i0, j0) represents the threshold value of similarity degree,
Typically one constant value of artificial selection, this largely constrains it and is automatically processing the application of aspect.To this
The method that invention proposes a kind of automatic selected threshold of energy, calculation formula are as follows：
Wherein Ω represents the contiguous range that size is 3 × 3 centered on (i0, j0).
Then all pixels point being in motion vector figure Vx (i, j) in circular shuttering position is counted：
Here the pixel point set in Ω expression motion vector figure Vx (i, j) in circular shuttering position, u (i0,
J0) be pixel (i0, j0) USAN values.
Using circular shuttering traversing graph as Vx (i, j) all pixels, the USAN value u (i, j) of all pixels can be obtained.
(2) after the USAN values for calculating all pixels, a preliminary characteristic point is obtained by thresholding and responds Rx (i, j)；
Rx (i, j)=max (0, th2-u (i, j)) (4)
Wherein th2 represents threshold value, here value 28, and only as u (i, j) ＜ th2, Rx (i, j) is only possible to be more than 0,
It is the characteristic point of a preliminary judgement to show the point.Pixel set of all Rx (i, j) more than 0 is preliminary feature point set
Close.
(3) preliminary set of characteristic points is handled using non-maxima suppression, obtains final set of characteristic points.
For each preliminary judgement in preliminary set of characteristic points characteristic point such as (i1, j1), observe using it in
The heart, in the field of size 5 × 5, if there is Rx (i, the j) values of other pixels than Rx (i1, j1) greatly, if not provided, namely
Rx (i1, j1) is maximum, then retains this characteristic point (i1, j1)；Otherwise delete this characteristic point (i1, j1), also i.e. by Rx (i,
J) 0 is reset to.The characteristic point of all preliminary judgements in preliminary set of characteristic points is handled, final all Rx (i, j)
Pixel set more than 0 is final set of characteristic points, is designated as { (x1,y1),…,(xN,yN)}。
Next characteristic vector description is carried out to each characteristic point.The present invention has used for reference the think of of local binary description
Think, encoded by color comparison information, but dual mode only characterizes simple magnitude relationship, can not react texture change
The difference degree of change, and be easily affected by noise.In order to overcome these problems, the present invention expands local binary
Exhibition, the result that color compares is divided into 3 classes, the details of texture variations is preferably reacted, comprises the following steps that：
Appoint and take a characteristic point (xn,yn) exemplified by illustrate.On image P (x, y, z), with (xn,yn) centered on, greatly
It is small be 7 × 7 image-region be designated as characteristic area Fn(x, y, z), therefore characteristic point (xn,yn) and characteristic area Fn(x, y, z) is
Correspond.Characteristic point (xn,yn) information be all reflected in characteristic area FnOn (x, y, z), therefore we pass through to characteristic area
Fn(x, y, z) carries out feature statement, to realize to characteristic point (xn,yn) purpose that is described.
Characteristic area Fn(x, y, z) shares 49 pixels, and the present invention carries out same processing side to each pixel
Method, to be illustrated exemplified by one of pixel (x0, y0),
Centered on (x0, y0), image P (x, y, z) select its surrounding close to 8 pixels be designated as { (x01,
y01),…,(x08,y08) contrast pixel as the point, then carry out first Model Comparison：
Wherein m ∈ { 1,2 ..., 8 } represent pixel serial number, and th1 represents first discrepancy threshold, here value 10, r1
(m) pixel (x is representedm,ym) coding under first pattern.Final 8 pixels can obtain the binary number of one 8
{ r1 (1), r1 (2) ..., r1 (8) }, it is pixel (x0, y0) at first to be translated into the value obtained by decimal number
Characteristic value under pattern, is designated as RT1.
Followed by the comparison of second pattern：
R2 (m) represents pixel (xm,ym) coding under first pattern.Same 8 pixels can obtain one 8
Binary number { r2 (1), r2 (2) ..., r2 (8) }, it is pixel (x0, y0) to be translated into the value obtained by decimal number
Characteristic value in a second mode, be designated as RT2.
Carry out the comparison of the 3rd pattern：
Wherein r3 (m) represents pixel (xm,ym) coding under the 3rd pattern.Same 8 pixels can obtain one
The binary number { r3 (1), r3 (2) ..., r3 (8) } of 8, be translated into the value obtained by decimal number be pixel (x0,
Y0 the characteristic value under the 3rd pattern), is designated as RT3.
It can be seen that the distribution of texture has been carried out thinner division by these three patterns, by discrepancy threshold th1 to color
Comparison result carries out the classification of three kinds of situations, is no longer simple under local binary be larger and smaller than.Therefore the present invention carries
The character description method gone out can preferably characterize the details of grain distribution.
The characteristic vector { RT1 ..., RT3 } that a length is three is obtained in final pixel point (x0, y0) one.
To characteristic area Fn(x, y, z) all pixels, are all handled according to the method described above, obtain respective feature
Vector, all characteristic vectors is arranged together, it is final to obtain the characteristic vector that a length is 49 × 3=147
{ RT1 ..., RT147 }, namely complete to characteristic point (xn,yn) description.
According to the method described above, all characteristic points are described respectively, available length is 49 × 3=147's
Characteristic vector.
Following feature based vector matches to characteristic point, for set of characteristic points { (x1,y1),…,(xN,yN)}
Each characteristic point, the minimum characteristic point of distance is found inside set of characteristic points by bi-directional matching.
Appoint and take a characteristic point (xa,ya) exemplified by illustrate, characteristic vector corresponding to note is RTa, and it is special to calculate itself and other
The Euclidean distance of point is levied, then the minimum characteristic point of chosen distance is designated as (xb,yb) as its preliminary matches characteristic point.Then
With same method to characteristic point (xb,yb) handled, when the minimum characteristic point of distance is exactly (xa,ya) when, it is final to determine
Characteristic point (xa,ya) and characteristic point (xb,yb) it is a pair of similar features points, otherwise characteristic point (xa,ya) there is no similar features point.
Bi-directional matching is carried out to each characteristic point by the above method, a series of similar features points can be obtained, be designated as
{[(xai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t represents the number finally matched.
After similar features point set is obtained, every a pair of similar features point is connected with line segment on image P (x, y, z)
Come, calculate the direction of every line segment, namely slope：
{fi| i=1 ..., t } slope is represented, the difference of these slopes is counted, difference is interior in scope (- 0.05,0.05), I
Think that the direction of line segment is basically identical.It is basically identical if there is at least 50 slopes, namely the direction of at least 50 line segments
It is basically identical, it is then preliminary to judge that image P (x, y, z) be present has duplication tampered region；Otherwise directly judge image P (x, y,
Z) without duplication tampered region.
If after tentatively judging that image P (x, y, z) be present has duplication tampered region, retain that on image P (x, y, z)
The basically identical line segment information of a little slopes, deletes other line segments.Then the distance of these line segments is calculated, formula is as follows：
dij=| yai-yaj|+|xai-xaj| (9)
Wherein i, j represent the label information of line segment, dijRepresent the distance of two lines section.If there is between 40 line segments
Mutual distance is less than 100, then final to judge that image P (x, y, z) be present has duplication tampered region；Otherwise judge image P (x, y,
Z) without duplication tampered region.
The present invention proposes a kind of image zone duplicating and altering detecting method based on local invariant feature, first by changing
The SUSAN Corner Detection Algorithms entered calculate set of characteristic points, then carry out feature to characteristic point by a kind of local tertiary mode
Description, the result that color compares is divided into three classes, is allowed to the details of preferably reaction texture variations, is examined finally by characteristic matching
Altimetric image is with the presence or absence of duplication tampered region.The inventive method can be more careful description image grain distribution, improve figure
As the accuracy of region duplication tampering detection.
Those skilled in the art will be clear that the scope of the present invention is not restricted to example discussed above, it is possible to which it is carried out
Some changes and modification, the scope of the present invention limited without departing from appended claims.Although oneself is through in accompanying drawing and explanation
Illustrate and describe the present invention in book in detail, but such explanation and description are only explanations or schematical, and it is nonrestrictive.
The present invention is not limited to the disclosed embodiments.
By to accompanying drawing, the research of specification and claims, when implementing of the invention, those skilled in the art can be with
Understand and realize the deformation of the disclosed embodiments.In detail in the claims, term " comprising " is not excluded for other steps or element,
And indefinite article "one" or " one kind " be not excluded for it is multiple.The some measures quoted in mutually different dependent claims
The fact does not mean that the combination of these measures can not be advantageously used.Any reference marker in claims is not formed pair
The limitation of the scope of the present invention.
Claims (5)
1. a kind of image zone duplicating and altering detecting method based on local invariant feature, it is characterised in that comprise the following steps：
Step S100：Feature point detection is carried out to all pixels point in image to be detected using SUSAN algorithms, obtains characteristic point
Set, is designated as { (x1,y1),…,(xN,yN)}；
Step S200：The set of characteristic points is traveled through with partial 3 d pattern, obtained corresponding with each characteristic point for retouching
State the image feature vector of each characteristic point；
Step S300：Travel through set of characteristic points and carry out described image characteristic vector and the characteristic point bi-directional matching, obtain each institute
State the similar features point pair of characteristic point, the Euclidean distance between the characteristic point of each similar features point centering is minimum, and
The similar features point is designated as { [(x to setai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t represents what is finally matched
Number；
Step S400：By similar features point described in each pair to being connected with line segment in described image to be detected, and calculate every
The slope f of bar line segmentiAnd fi-fi+1, judge fi-fi+1Whether in -0.05~0.005, this line segment is designated as if meeting
Consistent line segment, is unsatisfactory for, and does not mark；
Step S500：The quantity of the consistent line segment is counted, as slope quantity, is judged described oblique in described image to be detected
Rate quantity whether≤50, if meet if described image to be detected be duplicating image, described image to be detected if being unsatisfactory for
In the absence of replicate tampered region,
Delete the remaining line segment in addition to the consistent line segment in the duplicating image, calculate the line segments of the wantonly two unanimously line segments away from
From, and the consistent line segment quantity of the line segment distance less than 100 is counted, and whether judge the consistent line segment quantity of gained >=40
Individual, described image to be detected, which exists, if meeting replicates tampered region, is otherwise distorted in described image to be detected in the absence of duplication
Region.
2. the image zone duplicating and altering detecting method according to claim 1 based on local invariant feature, its feature exist
In the step S100 comprises the following steps：
Step S110：Image to be detected is traveled through with circular shuttering, calculates the USAN at each pixel in described image to be detected
Value u (i, j)；
Step S120：To USAN values u (i, j) thresholding of all pixels in the testing image, obtain characteristic point response Rx (i,
j)；
Rx (i, j)=max (0, th2-u (i, j)) (4)
Wherein, th2 is gained threshold value；
Step S130：Travel through the point of all pixels in described image to be detected and judge whether USAN value u (i, j) meet u (i, j) ＜
The point is then labeled as preliminary characteristic point by th2, satisfaction, otherwise carries out step S120 to next pixel, and gained is all preliminary
Feature point set is combined into preliminary set of characteristic points；
Step S140：The preliminary set of characteristic points is traveled through with non-maxima suppression, acquired results are assembled into the characteristic point
Set, is designated as { (x1,y1),…,(xN,yN)}；
The circular shuttering is all satisfactions (x-x1) centered on (x1, y1) pixel2+(y-y1)2≤ 10 pixel (x,
Y) region of composition.
3. the image zone duplicating and altering detecting method according to claim 2 based on local invariant feature, its feature exist
In the step S200 comprises the following steps：Wait to appoint the characteristic point (x taken with describedn,yn) centered on, by image P (x, y, z)
7 × 7 image-region is designated as characteristic area Fn(x, y, z), traversal characteristic area FnEach pixel in (x, y, z), to each
Pixel sequentially carries out first Model Comparison, second Model Comparison and the 3rd Model Comparison, obtains characteristic vector
{ RT1 ..., RT3 }, image feature vector will be obtained after characteristic vector set obtained by each pixel.
It is 4. a kind of such as the image zone duplicating based on local invariant feature of method according to any one of claims 1 to 3
Tampering detection apparatus, it is characterised in that including：
Feature point detection module：For carrying out feature point detection to all pixels point in image to be detected using SUSAN algorithms,
Set of characteristic points is obtained, is designated as { (x1,y1),…,(xN,yN)}；
Feature vector module：For traveling through the set of characteristic points with partial 3 d pattern, obtain corresponding with each characteristic point
Be used for the image feature vector of each characteristic point is described；
Similar features point is to module：Described image characteristic vector and two-way of the characteristic point are carried out for traveling through set of characteristic points
Match somebody with somebody, obtain the similar features point pair of each characteristic point, it is European between the characteristic point of each similar features point centering
Distance is minimum, and the similar features point is designated as into { [(x to setai,yai),(xbi,ybi)] | i=1 ..., t }, wherein t is represented
The number finally matched；
Consistent line segment module：For in described image to be detected by similar features point described in each pair to being connected with line segment,
And calculate the slope f of every line segmentiAnd fi-fi+1, judge fi-fi+1Whether in -0.05~0.005, if meet if this
Line segment is designated as consistent line segment, is unsatisfactory for, and does not mark；
Judge module：For counting the quantity of the consistent line segment, as slope quantity, the institute in described image to be detected is judged
Whether state slope quantity≤50, described image to be detected is duplicating image if meeting, described to be detected if being unsatisfactory for
Image, which is not present, replicates tampered region,
Delete the remaining line segment in addition to the consistent line segment in the duplicating image, calculate the line segments of the wantonly two unanimously line segments away from
From, and the consistent line segment quantity of the line segment distance less than 100 is counted, and whether judge the consistent line segment quantity of gained >=40
Individual, described image to be detected, which exists, if meeting replicates tampered region, is otherwise distorted in described image to be detected in the absence of duplication
Region.
5. the image zone duplicating and altering detection means according to claim 4 based on local invariant feature, its feature exist
In the feature point detection module includes：
USAN value modules：For traveling through image to be detected with circular shuttering, calculate in described image to be detected at each pixel
USAN value u (i, j)；
Threshold module：For USAN values u (i, j) thresholding to all pixels in the testing image, characteristic point response is obtained
Rx(i,j)；
Rx (i, j)=max (0, th2-u (i, j)) (4)
Wherein, th2 is gained threshold value；
Preliminary feature point module：All pixels point for traveling through in described image to be detected judges whether USAN value u (i, j) are full
The point is then labeled as preliminary characteristic point, otherwise enters threshold module to next pixel, by institute by sufficient u (i, j) ＜ th2, satisfaction
Obtain all preliminary feature point sets and be combined into preliminary set of characteristic points；
Non-maxima suppression module：For traveling through the preliminary set of characteristic points with non-maxima suppression, by acquired results set
Into the set of characteristic points, { (x is designated as1,y1),…,(xN,yN)}；
The circular shuttering is all satisfactions (x-x1) centered on (x1, y1) pixel2+(y-y1)2≤ 10 pixel (x,
Y) region of composition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710579410.XA CN107392949B (en) | 2017-07-17 | 2017-07-17 | Image zone duplicating and altering detecting method and detection device based on local invariant feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710579410.XA CN107392949B (en) | 2017-07-17 | 2017-07-17 | Image zone duplicating and altering detecting method and detection device based on local invariant feature |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107392949A true CN107392949A (en) | 2017-11-24 |
| CN107392949B CN107392949B (en) | 2019-11-05 |
Family
ID=60339347
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710579410.XA Active CN107392949B (en) | 2017-07-17 | 2017-07-17 | Image zone duplicating and altering detecting method and detection device based on local invariant feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107392949B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521821A (en) * | 2011-10-24 | 2012-06-27 | 南开大学 | Automatic identification and tampered area positioning method in digital image |
| CN104537654A (en) * | 2014-12-19 | 2015-04-22 | 大连理工大学 | Printed image tampering forensic methods based on half-tone dot location distortion |
| CN104616297A (en) * | 2015-01-26 | 2015-05-13 | 山东省计算中心（国家超级计算济南中心） | Improved SIFI algorithm for image tampering forensics |
| US20160132985A1 (en) * | 2014-06-10 | 2016-05-12 | Sam Houston State University | Rich feature mining to combat anti-forensics and detect jpeg down-recompression and inpainting forgery on the same quantization |
| CN105631871A (en) * | 2015-12-28 | 2016-06-01 | 辽宁师范大学 | Color image duplicating and tampering detection method based on quaternion exponent moments |
| US20170091588A1 (en) * | 2015-09-02 | 2017-03-30 | Sam Houston State University | Exposing inpainting image forgery under combination attacks with hybrid large feature mining |
-
2017
- 2017-07-17 CN CN201710579410.XA patent/CN107392949B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521821A (en) * | 2011-10-24 | 2012-06-27 | 南开大学 | Automatic identification and tampered area positioning method in digital image |
| US20160132985A1 (en) * | 2014-06-10 | 2016-05-12 | Sam Houston State University | Rich feature mining to combat anti-forensics and detect jpeg down-recompression and inpainting forgery on the same quantization |
| CN104537654A (en) * | 2014-12-19 | 2015-04-22 | 大连理工大学 | Printed image tampering forensic methods based on half-tone dot location distortion |
| CN104616297A (en) * | 2015-01-26 | 2015-05-13 | 山东省计算中心（国家超级计算济南中心） | Improved SIFI algorithm for image tampering forensics |
| US20170091588A1 (en) * | 2015-09-02 | 2017-03-30 | Sam Houston State University | Exposing inpainting image forgery under combination attacks with hybrid large feature mining |
| CN105631871A (en) * | 2015-12-28 | 2016-06-01 | 辽宁师范大学 | Color image duplicating and tampering detection method based on quaternion exponent moments |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107392949B (en) | 2019-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TW393629B (en) | Hand gesture recognition system and method | |
| Rosin et al. | Evaluation of global image thresholding for change detection | |
| Heath et al. | A robust visual method for assessing the relative performance of edge-detection algorithms | |
| US7502496B2 (en) | Face image processing apparatus and method | |
| US8000505B2 (en) | Determining the age of a human subject in a digital image | |
| CN109409190A (en) | Pedestrian detection method based on histogram of gradients and Canny edge detector | |
| CN111160249A (en) | Multi-class target detection method in optical remote sensing images based on cross-scale feature fusion | |
| CN103294989A (en) | Method for discriminating between a real face and a two-dimensional image of the face in a biometric detection process | |
| CN101147159A (en) | Fast method of object detection by statistical template matching | |
| CN109800682B (en) | Driver attribute identification method and related products | |
| Liu et al. | Optimal matching problem in detection and recognition performance evaluation | |
| JP2000517452A (en) | Viewing method | |
| CN108765407A (en) | A kind of portrait picture quality determination method and device | |
| WO2020217812A1 (en) | Image processing device that recognizes state of subject and method for same | |
| CN107689039A (en) | Estimate the method and apparatus of image blur | |
| CN110288040A (en) | A kind of similar evaluation method of image based on validating topology and equipment | |
| Panetta et al. | Unrolling post-mortem 3D fingerprints using mosaicking pressure simulation technique | |
| CN112991159B (en) | Face illumination quality evaluation method, system, server and computer readable medium | |
| CN107392949B (en) | Image zone duplicating and altering detecting method and detection device based on local invariant feature | |
| Wu et al. | Detecting image forgeries using metrology | |
| CN103093204B (en) | Behavior monitoring method and device | |
| CN110348464A (en) | An Image Forgery Detection Algorithm Based on Local Brightness Sequence of Multiple Support Regions | |
| CN104732521B (en) | A kind of similar purpose dividing method based on weight group similar active skeleton pattern | |
| CN109690555A (en) | Face detector based on curvature | |
| CN107122714A (en) | A kind of real-time pedestrian detection method based on edge constraint |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |