CN104091303A - Robust image hashing method and device based on Radon transformation and invariant features - Google Patents
Robust image hashing method and device based on Radon transformation and invariant features Download PDFInfo
- Publication number
- CN104091303A CN104091303A CN201410330434.8A CN201410330434A CN104091303A CN 104091303 A CN104091303 A CN 104091303A CN 201410330434 A CN201410330434 A CN 201410330434A CN 104091303 A CN104091303 A CN 104091303A
- Authority
- CN
- China
- Prior art keywords
- image
- hash
- extract
- features
- utilize
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Editing Of Facsimile Originals (AREA)
Abstract
The invention relates to a robust image hashing method and device based on Radon transformation and invariant features, and belongs to the field of information safety. In terms of the problem that hashing cannot resist geometric attacks well, normalized preprocessing operation is carried out on images firstly, invariant feature points are generated by utilizing an unchanged centroid algorithm, the circular area around an unchanged centroid is selected, Radon transformation is carried out on the circular area to generate a coefficient matrix, multiple lines of coefficients are selected randomly from a transformation domain by utilizing a chaotic system, robust features of each line are extracted, the features of all lines are combined with the invariant moment features of the whole matrix to generate image hashing, and similarity comparison is carried out by utilizing Euclidean distance. By the adoption of the robust image hashing method and device based on the Radon transformation and invariant features, the problem that the false drop rate rises due to geometric attacks can be solved effectively; the problems that computation complexity is too high and hashing is too long can be solved according to hashing steps and hashing lengths. The method and device can be applied to the field of image content authentication, and can also be applied to image retrieval, image identification and other information safety fields.
Description
Technical field
The present invention relates to image processing field and image content authentication field, espespecially a kind of image hash method and device thereof based on Radon conversion and invariant features.
Background technology
The develop rapidly of network and digital information technology, impels digital information to be applied more and more widely, and this also impels many traditional media to start to change to digitizing, makes the information on network more and more diversified.Digital medium information can be replicated easily, compresses, stores and transmit, because it is easily acquired and energy true-time operation, so be widely used.The characteristic storing, process, copy and transmit of being easy to just because of multimedia messages, people can go to obtain by network the information oneself needing very easily, and can obtain very rapidly these information, this gives the possessory copyright interest of multimedia messages and security information also with a lot of potential threats.Even without information owner's license, anyone can copy and spreading digital information easily, and declares numerical information copyright to one's name, can also illegally use and distort others' information with favourable to oneself.The digitizing of multimedia messages has equally also facilitated distorting of lawless person; this not only makes the border between infringement and reasonable utilization thicken; brought very large difficulty also to traditional judicial expertise; who is the real copyright ownership person of digital content to be difficult to judgement, thereby cannot really protect copyright owner's legitimate interests.Lawless person utilizes various approach to obtain and distort the situation that multimedia messages obtains sudden huge profits often to occur, and for this situation, we should strengthen copyright protection and integrated authentication to multimedia messagess such as image, audio frequency, videos.
The technology of image being carried out to content authentication mainly contains digital watermark technology and image salted hash Salted.Digital watermark technology, as a kind of method of image content authentication and copyright protection, has been widely used in image field.Image itself has a large amount of redundant informations, and these redundant informations are modified and can not changed significantly the visual quality of image.Digital watermarking is by revising redundant information with embed watermark information, and this technology can be applied in multiple multimedia messages, such as text, image, audio frequency, video etc.Digital watermarking need to embed information in initial carrier, can carry out some to carrier unavoidably and revise, and we will guarantee that the initial carrier after embed watermark can not have very large quality distortion, and the image after embed watermark is watermarking images.But watermark embedding can cause the robustness of watermark and the contradiction between invisibility, when the watermark strength embedding is too high, can cause image quality distortion more serious, but robustness is relatively good; When the watermark strength embedding is too low, image watermark invisibility is relatively good, but robustness is poor.And these all can be modified and cause distortion in various degree carrier information unavoidably, in the situation that some requirements of the fidelitys for picture quality and content are stricter, as medical image etc., thisly lose genuine situation and do not allow, so the application of digital watermark technology in these fields is subject to certain restrictions.
Image hash algorithm is a kind of new technology that is more suitable for image authentication than Image Watermarking Technique, is also better than to a certain extent the hash function authentication that some conventional ciphers are learned.Traditional cryptographic Hash function mainly contains MD5 and SHA scheduling algorithm.Image hash algorithm be mainly some robust image features that represent perceived content by extracting image to generate Hash information, as long as great changes will take place for the apperceive characteristic of image, the Hash information extracting does not just have very large change.Hash information is transmitted by being attached on original image, receiving end has received after image information, can obtain and pass the image Hash information of coming, and then extract the Hash information that receives image, contrast with the Hash information receiving, just can confirm that the image receiving is real or forges.As can be seen here, image hash algorithm does not need to embed extra information, can not revise the information of original image, therefore can not cause the distortion of initial carrier information.The research of image salted hash Salted relates to a plurality of technical fields such as signal processing, image processing, human visual system, theory of probability, information theory, cryptography, research to image salted hash Salted, not only can realize multimedia messages content recognition and reliable content authentication accurately, but also the fusion of multidisciplinary theory and method is played to certain facilitation, contribute to find new technological growth point.No matter so from theory or the angle of applying, the research of image salted hash Salted is all significant.By the analysis to existing algorithm, find, these algorithms are very sensitive to the geometric attack of image, the rotation attack of wide-angle etc. particularly, though and some algorithms can resist some small angle rotations, but its property distinguished and robustness existing problems.
The present invention is directed to image Hash in the problem of anti-geometric distortion robustness and the property distinguished existence, concept in image processing field and technology are applied in the research of image Hash, have realized a kind of robust image hash method and device based on Radon conversion and invariant features.
Summary of the invention
The present invention is intended to propose the concept of image Hash and the general structure flow process of image hash method, and a kind of robust image hash method and device based on Radon conversion and invariant features is provided.For gray level image, excavate the robust validity feature making new advances; Solve the problem of error in judgement in geometric distortion, make image Hash can either resist preferably some normal images and attack, also can resist preferably geometric attack; And algorithm computation complexity is too high and the long problem of Hash length, and there is good differentiation for different perceived content images with through the image of maliciously distorting, reach the balance between the property distinguished and robustness.
Image Hash is different from traditional hash function and digital watermark technology, tradition hash function is very responsive to raw information, very little change also can cause the cryptographic hash of generation to produce very large change, and digital watermark technology can be modified with embed watermark information to initial carrier, cause certain distortion.Image Hash is mainly to extract some robust features to generate Hash information, is attached to and on original image, carries out image authentication.
The key that image Hash extracts is the extraction of robust features, and the proper vector that how to extract the robust of resist geometric attacks is very important.The present invention has utilized Radon conversion to carry out feature extraction, and Radon converts as shown in Figure 1, along different angle directions, calculates projection information, and the Radon coefficient after rotating is as Fig. 2.G (r, θ) is the spatial information after conversion, and the mathematic(al) representation of Radon conversion is as formula (1):
G (r, θ)=R{f (x, y) }=∫ ∫ f (x, y) δ (r-xcos θ-ysin θ) dxdy (1) wherein r=xcos θ+ysin θ represent the distance from initial point, along different parallel lines, do integration and just can form projection g (r, θ), θ represents the normal vector of projection ray and the angle of transverse axis x axle, and 0≤θ < π.Radon conversion has good resist geometric attacks characteristic, and the geometric distortion operations such as convergent-divergent, translation, rotation are had to good robustness, and detailed is analyzed as follows:
Translation: image f (x, y) translation (x
0, y
0) distance, g (r, θ) can correspondingly carry out translation along r direction of principal axis, as formula (2).
R{f(x-x
0,y-y
0)}=g(r-x
0cosθ-y
0sinθ,θ) (2)
Convergent-divergent: image f (x, y) carries out convergent-divergent with the factor of λ (λ >0), makes g (r, θ) that transformation of coefficient occur, as formula (3).
R{f(x/λ,y/λ)}=λg(r/λ,θ)=g
λ(r,θ) (3)
Rotation: image f (x, y) if rotate θ centered by initial point
rangle, g (r, θ) can carry out corresponding translation in horizontal ordinate θ angle, as formula (4).
R{f(xcosθ
r-ysinθ
r,xsinθ
r+ycosθ
r)}=g(r,θ-θ
r) (4)
By above-mentioned formula, can know that Radon conversion is converted into the rotation of original image the angle translation of Radon coefficient domain, be still same zoom factor after original image convergent-divergent is converted into Radon coefficient.
In Radon transform domain, by utilizing svd (SVD) to carry out eigenwert extraction, in conjunction with other local feature value, form local Hash sequence, by utilizing normalized center square η
pq,
, r is (p+q)/2+1, p+q>=2, and seven invariant moment features that combination extracts Hu square, as overall Hash sequence, generate final image Hash sequence in conjunction with local Hash sequence and overall Hash sequence.
The general steps that the method that the present invention proposes realizes is as follows:
The first step: image pre-service.First image will be through the preprocessing process of a standard, and the image obtaining by pre-service can have better robustness.
Second step: constant barycenter and border circular areas extract.Because constant barycenter extracts after image pre-service, image has carried out filtering operation, so robustness can be improved.Extracted after constant barycenter, take constant center of mass point as the center of circle, the self-adaptation R of take extracts a border circular areas as radius.
The 3rd step: feature extraction and Hash generate.The general frame that Hash generates as shown in Figure 3.
Border circular areas is carried out to Radon map function, choose at random the generation that 15 row carry out Hash sequence, every a line is extracted to robust features and form a proper vector, the combination of eigenvectors of every a line is got up to form local cryptographic hash.Extract seven moment characteristics of Hu square as overall cryptographic hash, then combine local cryptographic hash and overall cryptographic hash and obtain final Hash sequence.
The 4th step: image authentication.The general frame of image authentication as shown in Figure 4.
When carrying out image authentication, the method that obtains Hash is the same with Hash generation method, the method is mainly to have utilized Euclidean distance d to carry out the comparison of Hash similarity, by comparing Hash, apart from the magnitude relationship of d and setting threshold, determines whether image is real.
Advantage of the present invention and good effect
The present invention proposes a kind of robust image hash method and device based on Radon conversion and invariant features.The method utilizes Radon conversion to generate robust hashing sequence, and can resist certain geometric attack.First the method carries out standardized pre-service to original image, then extract the constant barycenter of image, around constant barycenter, extract border circular areas around, border circular areas is carried out to Radon conversion, can there is good anti-translation feature, by key, select at random 15 row of coefficient, extract the robust features vector of every row, in conjunction with overall invariant features square synthetic image Hash, the method can meet the robustness requirement that normal image is processed, also can resist certain geometric attack, and there is the preferably differentiation to malice tampered image and perception different images, also there is the security feature based on key simultaneously, its Hash calculation complexity is lower and Hash length is shorter.
The present invention makes robust image Hash practical, and it applies as follows substantially:
1. image authentication.Image authentication can detect original image and whether passed through conventional image attack or the different image of content and provided judgement.
2. image recognition.Image Hash is the intrinsic characteristic of presentation video preferably, can represent the general characteristic of piece image, therefore can be applied to preferably image recognition.
3. image retrieval.If image Hash is applied to image retrieval, to its robustness and the property distinguished, require meeting very high so, image retrieval can fast and accurately really be oriented the image that will search in image data base, and this requires higher to the computation complexity of hash algorithm and recall precision.
4. distort detection.Whether image salted hash Salted also can distort operation through malice for detection of original image, if there is malice, distorts, and the Hash sequence extracting has a great difference.
Accompanying drawing explanation
Fig. 1 is Radon conversion schematic diagram
Fig. 2 is Radon coefficient distribution schematic diagram after image rotation
Fig. 3 is image Hash product process figure
Fig. 4 is image Hash identifying procedure figure
Embodiment
For making object of the present invention, technical scheme more clear, below specific embodiments of the present invention is described in detail.
The concrete steps of the robust image hash method based on Radon conversion and invariant features are as follows:
The first step, image pre-service.First image will be through the preprocessing process of a standard, first original image is converted into the image of m * m size of standard through interpolation algorithm, this step is mainly in order to guarantee that the Hash sequence that different images generates has identical length, can resist convergent-divergent distortion to a certain extent.Then image is carried out to low-pass filtering operation one time, can filter out some and hold labile noise, retained the important content feature of image, the image drawing by pre-service has better robustness.
Second step, constant barycenter and border circular areas extract.The center algorithm that never degenerates refers to an invariant features point that extracts image, even if image is processed and geometric distortion by conventional image, this unique point still can remain unchanged in image.By extracting the constant barycenter of image, at it, extract the anti-geometric distortion characteristic that border circular areas just can guarantee image around.
The constant barycenter of image be can be after conventional attack and geometric attack constant point still, so extract the anti-geometrical property that constant barycenter concerns hash algorithm, we carry out the constant center of mass point of synthetic image by a kind of method of iteration.Suppose that the computing formula of barycenter of original image F (x, y) is as formula (5):
Suppose the x here, y ∈ M * N, belongs to whole image.The main step of algorithm is a barycenter that first calculates original image, is assumed to be C
0, as an initial value C of constant barycenter
b, i.e. C
b=C
0, point centered by this initial value then, the border circular areas that radius is r continues to extract barycenter C
rif, C
b=C
r, C
rbe exactly the constant barycenter of image, otherwise C has been set
b=C
r, continue with C
bcentered by point, radius is that r extracts barycenter, until the constant barycenter of extracted twice is samely just to finish, last center of mass point is exactly constant center of mass point.Because constant barycenter extracts after image pre-service, image has carried out filtering operation, thus the constant barycenter extracting be subject to some conventional attacks affect the corresponding reduction of meeting, robustness can be improved.
Extracted after constant barycenter, take constant center of mass point as the center of circle, the self-adaptation R of take extracts a border circular areas as radius, R can be less than the length of original image and wide, the region extracting like this can keep translation, invariable rotary, can generate Hash by extract important robust features in this border circular areas.
The 3rd step, feature extraction and Hash generate.
Extracting border circular areas and this region is being carried out after Radon conversion, again wavelet transform (DWT) is carried out in conversion Radon territory out, can access four corresponding wavelet coefficient sub-band images, this algorithm is mainly the low frequency sub-band coefficient that has utilized wavelet transformation, because the main energy that this sub-band coefficients has comprised Radon coefficient, can characterize Radon coefficient characteristics preferably.The processing that Radon coefficient is normalized, all Radon coefficients are divided by maximum Radon coefficient, mainly to utilize Radon coefficient original image to be converted into the characteristic of translation, logistic system can be used for generating pseudo-random sequence, so can utilize logistic system to select 15 Radon line of coefficients, this is based on secret key safety.Utilize logistic sequence to choose at random 15 row herein and carry out the generation of Hash sequence.These coefficients are mainly the conversion of having carried out some translations, and content very large change can not occur, so will extract the invariant features of robust on these row.When θ axle generation cycle spinning, suppose that the anglec of rotation is φ, define zeroth order moment function to be:
When rotating, there is following relation:
Therefore the zeroth order square of Radon conversion row can be used as an invariant features, and the variances sigma of corresponding line also can be used as an eigenwert:
I wherein, j represents corresponding line number and every row coefficient numbering, and len represents the length of every row, and m is average, g
i(j) coefficient value of representative row.By the analysis to row coefficient, consider that SVD decomposition has good resist geometric attacks performance, so extract maximum singular value that the SVD of every row decomposes as an eigenwert, finally utilize the anti-translation feature of Fourier transform, because the coefficient generating is constant, the real part of Fourier Transform Coefficients has extraordinary stability, and imaginary part may change to some extent because of some modifications, very responsive for conventional attack, so extract the absolute value of the real part of Fourier Transform Coefficients, generate another one matrix, this matrix has good anti-conventional attack ability, because be real part, be not easy to be subject to the change of conventional attack, again this matrix is carried out to dct transform one time, extract DC coefficient DC as an eigenwert, so just generated four eigenwerts, form a proper vector, the combination of eigenvectors of every a line is got up to form local feature Hash.
Being defined as follows of Hu square, supposes that the height of original image and width are respectively m and n, establish M
pqfor (p+q) rank square
When p and q get null value, just can obtain the zeroth order square M of image
00, be also called " quality " of image, and
the center-of-mass coordinate of image is so
wherein
The center square that defines so image is μ
pq
Suppose that normalized center square is η
pq, it can be defined as:
R is (p+q)/2+1, and p+q>=2 utilizes seven moment characteristics that square combination in center extracts Hu square as overall cryptographic hash.Be mainly to have utilized normalization center square to generate seven constant feature squares, concrete building method is as follows:
M
1=η
20+η
02 (13)
M
2=(η
20-η
02)
2+4η
11 2 (14)
M
3=(η
30-3η
12)
2+(3η
21-η
03)
2 (15)
M
4=(η
30+η
12)
2+(η
21+η
03)
2 (16)
M
5=(η
30-3η
12)(η
30+η
12)[(η
30+η
12)
2-3(η
21+η
03)
2]
(17)
+3(η
21-η
03)(η
21+η
03)[3(η
30+η
12)
2-(η
21+η
03)
2]
M
6=(η
20-η
02)[(η
30+η
12)
2-(η
21+η
03)
2]
(18)
+4η
11(η
30+η
12)(η
21+η
03)
M
7=(3η
21-η
03)(η
30+η
12)[(η
30+η
12)
2-3(η
21+η
03)
2]
(19)
-3(η
30-3η
12)(η
21+η
03)[3(η
30+η
12)
2-(η
21+η
03)
2]
Combination extract Hu square seven moment characteristics as overall cryptographic hash, global characteristics preferably represent images content and can resist geometric attack, there is good stability simultaneously.Then utilize the cryptographic hash of partial row and overall cryptographic hash to be combined into a final Hash sequence H.So the Hash length of this algorithm is 67 decimal numbers.
The 4th step, image authentication.
This chapter algorithm is mainly to have utilized Euclidean distance to carry out the comparison of Hash similarity, supposes that original Hash sequence is H, and the Hash sequence that certified image is extracted is H
1, the definition of Euclidean distance is as shown in formula (20).
Wherein H (i) and H
1(i) represent i cryptographic hash of original Hash sequence and extraction Hash.By comparing Hash, apart from d, determine whether image is real.The distance of general perception identical image is smaller, and the distance of perception different images is larger, can set by experiment a threshold value T, if this d<=T, think that image is real image, otherwise think that two images are that perception is different, image is untrue.
Claims (8)
1. robust image hash method and the device thereof based on radon conversion and invariant features, utilize the geometric distortion invariant features extracting in the anti-geometrical property of radon conversion and radon transform domain with synthetic image Hash, transmitting terminal first carries out pretreatment operation to image, the utilization center algorithm that never degenerates extracts constant barycenter in image and obtains and take surrounding's border circular areas that this barycenter is the center of circle, again border circular areas is carried out radon conversion and extracts local feature and global characteristics, and utilize chaos system to guarantee the security of extracting, both are combined and obtain image Hash sequence and be attached to image uploading to receiving end, when carrying out image authentication, receiving end profit uses the same method to extract and receives image Hash, and carry out Euclidean distance calculating with the Hash sequence receiving, by with setting threshold relatively carry out the authentication operation of image.
2. method according to claim 1, is characterized in that, the algorithm general steps that the present invention proposes is as follows:
A. image carries out normalized pretreatment operation;
B. utilize the center algorithm that never degenerates to generate invariant features point, then extract constant barycenter border circular areas around;
C. utilize chaos system to extract local Hash string at radon transform domain, then try to achieve whole matrix of coefficients overall situation Hash string, finally in conjunction with local Hash string and overall Hash, concatenate into final Hash sequence;
D. utilize Euclidean distance as the comparison of similarity, for image authentication.
3. method according to claim 1, is characterized in that, the pretreated operation steps of image is as follows:
First original image is converted into the image of m * m size of standard through interpolation algorithm, this step is mainly in order to guarantee that the Hash that different images generates has identical length, can resist convergent-divergent distortion to a certain extent.Then image is carried out to low-pass filtering operation one time, can filter out some and hold labile noise, retained the important content feature of image, the image obtaining after pre-service has better robustness.
4. method according to claim 1, is characterized in that, the process that constant barycenter and border circular areas extract is as follows:
Because constant barycenter extracts after image pre-service, image has carried out filtering operation, so robustness can be improved, recycle the constant barycenter that the center algorithm that never degenerates extracts image, take constant center of mass point as the center of circle, the self-adaptation R of take extracts a border circular areas as radius, and the region extracting like this can keep translation, invariable rotary, can generate Hash by extract important robust features from this border circular areas.
5. method according to claim 1, it is characterized in that, in feature extraction and Hash generative process, border circular areas is carried out to radon conversion, utilize the resist geometric attacks feature of radon transform domain, in this matrix of coefficients, choose at random 15 row and extract robust how much invariant features and form local Hash sequence, then from whole matrix of coefficients, extract invariant moment features and obtain overall Hash sequence, finally in conjunction with local Hash and overall Hash, generate final Hash sequence.
6. method according to claim 1, is characterized in that, after obtaining image Hash sequence, utilizes Euclidean distance to carry out the comparison of Hash similarity, supposes that original Hash sequence is H, and the Hash sequence that certified image is extracted is H
1, Euclidean distance is defined as follows:
Wherein H (i) and H
1(i) represent i cryptographic hash of original Hash sequence and the Hash sequence extracting, by comparing Hash, apart from d, determine that whether image is true, the distance of general perception identical image is smaller, and the distance of perception different images is larger, set by experiment a threshold value T, if d<=T, think that image is real image, two images are that vision is similar, otherwise think that two images are that perception is different, and image is untrue.
7. method according to claim 5, it is characterized in that, extract in local Hash procedure, utilize random 15 row that extract of logistic chaos system, by the analysis to row coefficient, zeroth order square be mainly ask every row coefficient and, there is invariant features, utilize the zeroth order square of every row as a cryptographic hash, then generate the variance of corresponding line as another one cryptographic hash, consider that SVD decomposition has good resist geometric attacks performance, so extract maximum singular value that the SVD of every row decomposes as an eigenwert, finally utilize the anti-translation feature of Fourier transform, because the coefficient generating is constant, the real part of Fourier Transform Coefficients has extraordinary stability, so extract the absolute value of the real part of Fourier Transform Coefficients, generate another one matrix, again this matrix is carried out to dct transform one time, extract DC coefficient DC as an eigenwert, so just generated four eigenwerts, form a proper vector, the combination of eigenvectors of every a line is got up to form local feature Hash.
8. method according to claim 5, it is characterized in that, extract in overall Hash procedure, utilize the normalization center square of Hu square, extract seven invariant moment features as overall Hash string, global characteristics can be expressed preferably the content of image and can be resisted geometric attack, has good stability simultaneously.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410330434.8A CN104091303A (en) | 2014-07-11 | 2014-07-11 | Robust image hashing method and device based on Radon transformation and invariant features |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410330434.8A CN104091303A (en) | 2014-07-11 | 2014-07-11 | Robust image hashing method and device based on Radon transformation and invariant features |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104091303A true CN104091303A (en) | 2014-10-08 |
Family
ID=51639018
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410330434.8A Pending CN104091303A (en) | 2014-07-11 | 2014-07-11 | Robust image hashing method and device based on Radon transformation and invariant features |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104091303A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104680473A (en) * | 2014-12-20 | 2015-06-03 | 辽宁师范大学 | Machine learning-based color image watermark embedding and detecting method |
CN104933182A (en) * | 2015-06-30 | 2015-09-23 | 孙建德 | Method for determining video hash effective length |
CN105844643A (en) * | 2016-03-30 | 2016-08-10 | 北京信息科技大学 | Image tamper detection method |
CN106650829A (en) * | 2017-01-04 | 2017-05-10 | 华南理工大学 | Picture similarity calculation method |
CN109332192A (en) * | 2018-08-03 | 2019-02-15 | 小黄狗环保科技有限公司 | A kind of image-recognizing method classified for pop can and beverage bottle |
CN109613462A (en) * | 2018-11-21 | 2019-04-12 | 河海大学 | A kind of scaling method of CT imaging |
CN110276772A (en) * | 2019-05-10 | 2019-09-24 | 深圳大学 | The automatic positioning method and system of structural element in a kind of musculature |
CN111104872A (en) * | 2019-11-29 | 2020-05-05 | 兰州交通大学 | GF-2 image integrity authentication method applying SIFT and SVD perceptual hashing |
CN111429337A (en) * | 2020-02-28 | 2020-07-17 | 上海电力大学 | Image hash acquisition method based on transform domain and shape characteristics |
CN113450247A (en) * | 2021-07-05 | 2021-09-28 | 重庆邮电大学 | Digital watermarking method based on binary matrix scrambling and chaos theory |
CN113960580A (en) * | 2021-10-14 | 2022-01-21 | 电子科技大学 | Transform domain invariant feature extraction method for true and false target one-dimensional range profile |
-
2014
- 2014-07-11 CN CN201410330434.8A patent/CN104091303A/en active Pending
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104680473A (en) * | 2014-12-20 | 2015-06-03 | 辽宁师范大学 | Machine learning-based color image watermark embedding and detecting method |
CN104933182A (en) * | 2015-06-30 | 2015-09-23 | 孙建德 | Method for determining video hash effective length |
CN104933182B (en) * | 2015-06-30 | 2018-05-11 | 山东管理学院 | A kind of method of definite video Hash effective length |
CN105844643B (en) * | 2016-03-30 | 2018-11-13 | 北京信息科技大学 | Distorted image detection method |
CN105844643A (en) * | 2016-03-30 | 2016-08-10 | 北京信息科技大学 | Image tamper detection method |
CN106650829B (en) * | 2017-01-04 | 2019-07-16 | 华南理工大学 | A kind of picture similarity calculating method |
CN106650829A (en) * | 2017-01-04 | 2017-05-10 | 华南理工大学 | Picture similarity calculation method |
CN109332192A (en) * | 2018-08-03 | 2019-02-15 | 小黄狗环保科技有限公司 | A kind of image-recognizing method classified for pop can and beverage bottle |
CN109613462A (en) * | 2018-11-21 | 2019-04-12 | 河海大学 | A kind of scaling method of CT imaging |
CN110276772A (en) * | 2019-05-10 | 2019-09-24 | 深圳大学 | The automatic positioning method and system of structural element in a kind of musculature |
CN110276772B (en) * | 2019-05-10 | 2022-10-18 | 深圳大学 | Automatic positioning method and system for structural elements in muscle tissue |
CN111104872A (en) * | 2019-11-29 | 2020-05-05 | 兰州交通大学 | GF-2 image integrity authentication method applying SIFT and SVD perceptual hashing |
CN111429337A (en) * | 2020-02-28 | 2020-07-17 | 上海电力大学 | Image hash acquisition method based on transform domain and shape characteristics |
CN111429337B (en) * | 2020-02-28 | 2022-06-21 | 上海电力大学 | Image hash acquisition method based on transform domain and shape characteristics |
CN113450247A (en) * | 2021-07-05 | 2021-09-28 | 重庆邮电大学 | Digital watermarking method based on binary matrix scrambling and chaos theory |
CN113450247B (en) * | 2021-07-05 | 2024-03-29 | 重庆邮电大学 | Digital watermarking method based on binary matrix scrambling and chaos theory |
CN113960580A (en) * | 2021-10-14 | 2022-01-21 | 电子科技大学 | Transform domain invariant feature extraction method for true and false target one-dimensional range profile |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104091303A (en) | Robust image hashing method and device based on Radon transformation and invariant features | |
Singh et al. | Image watermarking using soft computing techniques: A comprehensive survey | |
Wang et al. | A visual model-based perceptual image hash for content authentication | |
Davarzani et al. | Perceptual image hashing using center-symmetric local binary patterns | |
CN100361437C (en) | Digital image hash signature method | |
CN102306305B (en) | Method for authenticating safety identity based on organic characteristic watermark | |
Xuehua | Digital watermarking and its application in image copyright protection | |
Karsh et al. | Robust image hashing through DWT-SVD and spectral residual method | |
Fadl et al. | Robust copy–move forgery revealing in digital images using polar coordinate system | |
Bilal et al. | Single and multiple copy–move forgery detection and localization in digital images based on the sparsely encoded distinctive features and DBSCAN clustering | |
CN102903075B (en) | Robust watermarking method based on image feature point global correction | |
CN104166955B (en) | Based on the generation of conformal mapping image Hash and distorted image detection localization method | |
Feng et al. | A novel semi-fragile digital watermarking scheme for scrambled image authentication and restoration | |
Liu et al. | Robust blind image watermarking based on chaotic mixtures | |
CN104636764A (en) | Image steganography analysis method and device | |
CN102129655A (en) | Wavelet domain-based method for weighting fractional differential image digital watermark | |
Su | Color image watermarking: algorithms and technologies | |
CN102646259A (en) | Anti-attack robustness multiple zero watermark method | |
Davarzani et al. | Image authentication using LBP-based perceptual image hashing | |
CN101817268A (en) | Application of digital watermarking in RFID technology | |
Yang et al. | Robust Watermarking Algorithm for Medical Images Based on Non-Subsampled Shearlet Transform and Schur Decomposition. | |
Ito et al. | Access control of semantic segmentation models using encrypted feature maps | |
Tiankai et al. | A robust authentication algorithm for medical images based on fractal Brownian model and visual cryptography | |
Xia et al. | Perceptual image hashing using rotation invariant uniform local binary patterns and color feature | |
Rai | Robust Digital Watermarking based on Machine Learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20141008 |
|
WD01 | Invention patent application deemed withdrawn after publication |