CN104881668A - Method and system for extracting image fingerprints based on representative local mode - Google Patents
Method and system for extracting image fingerprints based on representative local mode Download PDFInfo
- Publication number
- CN104881668A CN104881668A CN201510241287.1A CN201510241287A CN104881668A CN 104881668 A CN104881668 A CN 104881668A CN 201510241287 A CN201510241287 A CN 201510241287A CN 104881668 A CN104881668 A CN 104881668A
- Authority
- CN
- China
- Prior art keywords
- image
- local mode
- storehouse
- representative
- finger
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Abstract
The invention discloses a method and system for extracting image fingerprints based on a representative local mode, and relates to the field of image processing. The method comprises: performing image attack analog processing of database images to generate new database images; extracting key points of the database images and the new database images; obtaining local blocks according to the key points; generating local modes and establishing a local mode database according to the local blocks; obtaining the representative local mode from the local mode database; establishing image fingerprints of the database images and the new database images according to the representative local mode; storing the image fingerprints into an image fingerprint database; obtaining a new image and extracting fingerprints of the new image; comparing the fingerprints of the new image with the image fingerprints in the image fingerprint database; and searching an image corresponding to the new image in the database images. According to the invention, the required memory is small; accelerated matching is performed by using an optimized machine instruction; the method and system are suitable for massive image copying and detection.
Description
Technical field
The present invention relates to image processing field, particularly a kind of image fingerprint extracting method based on representative local mode and system.
Background technology
It is one of computer vision hot research field that image copy detects all the time, and its general strategy is the feature extracting minority from image, and is integrated into the proper vector that can reflect picture material.From the constituted mode of characteristics of image, be generally divided into global characteristics and local feature, the current situation below for these two kinds of features carries out concise and to the point discussion.As far back as 1999, Naphade first extracted the histogram of image at LUV color space as the method for image signatures, and Mohan uses the method for piecemeal subsequently, extracts the average gray value of each piecemeal, according to the gray-scale value size sequence construction feature of these piecemeals; A more influential finger image feature is the GIST method that Oliva proposes, the method is extracted image 5 kinds of visually-perceptible features: naturalness, openness, roughness, degree of expansion and dangerously steep degree, the method embodies good accuracy under common image attack.In the last few years, in order to tackle the image retrieval under image local area change scene, there has been proposed local description.Method the most classical is the SIFT of Lowe, DoG peak response is chosen as key point in the three dimensions that first the method is formed in plane and yardstick, then centered by unique point, get a certain size region and carry out piecemeal, finally the gradient orientation histogram extracted in piecemeal being connected and forming final descriptor.Recently, binary local feature obtains everybody extensive concern, Leutenegger extracts key point by efficient AGAST Corner detector, then sample around key point, binary string BRISK is built by the comparison of gray-scale value, the method is fast at extraction rate, has certain scale invariability.
Existing global characteristics and local feature application detect in application to large-scale image copy and mainly contain following shortcoming and defect: poor robustness: existing global characteristics builds according to Global Information such as the color histogram of image, therefore block in image generating portion, cutting will lose efficacy; Feature extraction speed is slow: for large-scale image retrieval, and the extraction rate of characteristics of image requires very high, and especially emerging cell phone platform, its hardware performance is limited, so require that the complexity of feature extraction can not be too high.And the sub-GIST of main flow global description needs calculating 5 kinds of feature interpretation at present, local feature SIFT needs to calculate DOG operator consuming time, and these algorithms all cannot be applied in large-scale image copy detection and go; Memory consumption is large: characteristic matching complexity is high, existing feature interpretation is in order to ensure its robustness and distinction, often dimension is all higher, for SIFT descriptor, block around unique point is divided into 16 regions by this algorithm, each region calculates the gradient orientation histogram in 8 directions, totally 128 dimension floating type features, if each floating type accounts for 4 bytes, a SIFT feature need consume 500 byte of memorys, and a sub-picture has hundreds of SIFT feature, the feature of higher-dimension like this is given to store and mate and is all brought huge pressure.
Invention " synthetic image fingerprint and carry out the method for retrieving similar images based on this ", the invention provides a kind of synthetic image fingerprint and carry out retrieving similar images based on the method and usurp knowledge method for distinguishing for image, described image has M capable N row pixel, comprises the steps: a) to each pixel X in described image
ijwherein i=2,3 ... M, j=2,3, N, the gray-scale value size of more described pixel pixel left with it and upper pixel, wherein, left pixel represents the pixel on the left side in the adjacent orientation of a described pixel four pixels up and down, and upper pixel represents the pixel of top in the adjacent orientation of a described pixel four pixels up and down; B) be described pixel X according to different comparative results
ijgive fingerprint value Z
ij; C) by the fingerprint value Z of described each pixel
ijbe combined as the fingerprint that sequence obtains the described image that length is (M-1) × (N-1).This invention adopts each pixel structure fingerprint value, its calculating, storage and Matching power flow are large, and the present invention excavates the representative pattern of the key point in image, every width image only constructs binary finger image of 16 bytes, in calculating, stores and matches and be better than this invention.
Invention " a kind of image copy detection method based on local digital fingerprint ", a kind of image copy detection method based on local digital fingerprint of this disclosure of the invention.The SIFT feature of these higher-dimensions vector to the every width image zooming-out local SIFT feature in test pattern storehouse, and is carried out the frequency that the conversion of local digital fingerprint and statistics fingerprint occur in every width image, to set up digital fingerprint data storehouse by the method; When image is inquired about, first SIFT feature is extracted to query image, then the information obtaining unreliable position in its digital finger-print and conversion process is transformed, inquire about in the inverted index structure in test fingerprint storehouse in conjunction with unreliable positional information again, thus obtain the test pattern set that is associated with the local digital fingerprint of query image fast, for query image carries out similarity measurement with the test pattern be associated, to determine whether copy.The method that this invention adopts SIFT descriptor to extract as finger image, calculate and memory cost larger, and adopt FAST algorithm as detection of image key points herein, and excavate representative local mode by statistics and set up binary picture fingerprint characteristic, the machine instruction optimized is used to carry out coupling retrieval, significantly decrease memory cost, improve the speed of images match.
Invention " image fingerprint extracting method and equipment, information filtering method and system thereof ", this invention provides a kind of image fingerprint extracting method and carries out the method and system of information filtering based on finger image.Image fingerprint extracting method wherein comprises the steps: S1: carry out interpolation processing to original image; S2: by the image block after interpolation processing, and carry out dct transform; S3: the picture after conversion is carried out RGB gray proces; S4: the image result of gray proces is quantized; S5: carry out Lossless Compression to image, exports a binary sequence; S6: above-mentioned binary sequence is carried out serializing restructuring, obtains finger image.This invention adopts dct transform and employs harmless compression of images mode further to obtain finger image, and method of the present invention adopts efficient FAST algorithm, and directly constructs binary picture fingerprint characteristic, and method is more efficient.
Summary of the invention
For the deficiencies in the prior art, the present invention proposes a kind of image fingerprint extracting method based on representative local mode and system.
The present invention proposes a kind of image fingerprint extracting method based on representative local mode, comprising:
Step 1, storehouse image is carried out image attack simulation process, generate new storehouse image, and extract the key point of described storehouse image and described new storehouse image, according to described key point, obtain partial block, and according to described partial block, generate local mode and set up local mode storehouse, from described local mode storehouse, obtaining representative local mode;
Step 2, according to described representative local mode, sets up the finger image of described storehouse image and described new storehouse image, and by described finger image stored in finger image storehouse;
Step 3, obtains new images, extracts the new images fingerprint of described new images, compared by the finger image in described new images fingerprint and described finger image storehouse, to search in the image of storehouse with described new images to corresponding image.
The described image fingerprint extracting method based on representative local mode, described in described step 1, representative local mode meets the following conditions:
Quantity of information in described representative local mode accounts for more than 80% of gross information content in image;
Described representative local mode, after image change, has high robust.
The described image fingerprint extracting method based on representative local mode, in described step 1, local mode P is:
The average gray in region centered by C, V
nbe the average gray of the n-th partial block, wherein n=1,2 ..., 7, b
nwhat be the n-th partial block with center block on gray-scale value compares zone bit.
The described image fingerprint extracting method based on representative local mode, described step 2 also comprises carries out pre-service to image, and extracts the key point of described image.
The described image fingerprint extracting method based on representative local mode, described step 3 comprises asks for Hamming distances by each finger image in described new images fingerprint and described finger image storehouse, candidate image is sorted from small to large according to described Hamming distances, gets front K and return as Query Result.
The present invention also proposes a kind of finger image extraction system based on representative local mode, comprising:
Set up local mode storehouse and obtain representative local mode module, for storehouse image is carried out image attack simulation process, generate new storehouse image, and extract the key point of described storehouse image and described new storehouse image, according to described key point, obtain partial block, and according to described partial block, generate local mode and set up local mode storehouse, from described local mode storehouse, obtaining representative local mode;
Set up finger image library module, for according to described representative local mode, set up the finger image of described storehouse image and described new storehouse image, and by described finger image stored in finger image storehouse;
Fingerprint comparison module, for obtaining new images, extracts the new images fingerprint of described new images, is compared by the finger image in described new images fingerprint and described finger image storehouse, to search in the image of storehouse with described new images to corresponding image.
The described finger image extraction system based on representative local mode, described set up local mode storehouse and obtain representative local mode described in representative local mode module meet the following conditions:
Quantity of information in described representative local mode accounts for more than 80% of gross information content in image;
Described representative local mode, after image change, has high robust.
The described finger image extraction system based on representative local mode, described set up local mode storehouse and obtain local mode P in representative local mode module be:
The average gray in region centered by C, V
nbe the average gray of the n-th partial block, wherein n=1,2 ..., 7, b
nwhat be the n-th partial block with center block on gray-scale value compares zone bit.
The described finger image extraction system based on representative local mode, described finger image library module of setting up also comprises and carries out pre-service to image, and extracts the key point of described image.
The described finger image extraction system based on representative local mode, described fingerprint comparison module comprises asks for Hamming distances by each finger image in described new images fingerprint and described finger image storehouse, candidate image is sorted from small to large according to described Hamming distances, gets front K and return as Query Result.From above scheme, the invention has the advantages that:
The present invention by experiment data statistics finds that image has certain stability in the local mode distribution of its unique point after copy-attack, and excavates by statistical study the basis that pattern that in these local modes, most is representative, robustness is high constructs as finger image; Organically the method for local detection method and the overall situation is merged; Technique effect: it is little that this method has global characteristics committed memory, the feature that matching speed is fast, possesses again the robustness of local feature method in the situations such as certain cutting, partial occlusion simultaneously; The relation that the representative local mode excavated by statistics is distributed in image four quadrants builds binary feature.Finger image feature committed memory of the present invention is few, the machine instruction of optimization can be used to carry out acceleration coupling, be suitable for large-scale image copy and detect.
Accompanying drawing explanation
Fig. 1 is overview flow chart of the present invention;
Fig. 2 a is original image in embodiment;
Original image is carried out the image after cutting process in embodiment by Fig. 2 b;
Original image is carried out brightness to change the image after processing by Fig. 2 c in embodiment;
Original image is carried out the image after dimensional variation process in embodiment by Fig. 2 d;
Fig. 2 e is the image after original image being carried out in embodiment logo and captions embedding process;
Fig. 2 f is for being the first five local representative mode in embodiment;
Fig. 3 is local mode figure;
Fig. 4 is design sketch of the present invention.
Embodiment
The present invention proposes a kind of image fingerprint extracting method based on representative local mode, as shown in Figure 1, main process of the present invention is divided into three phases to its concrete flow process: the training stage, build storehouse stage and online query stage.Wherein three phases all comprises the Image semantic classification mainly size of original image is unitized and smoothing processing, introduces the key step of three phases below respectively:
Training stage: storehouse image is mainly carried out various image attack simulation process by this stage, mainly comprise ratio change, captions or Logo embed, cutting, fuzzy etc., then the key point in FAST algorithm extraction image is used, (" key point " refers to the partial structurtes containing important information in image, the structures such as the eyes of such as face and nose are just more remarkable than forehead, containing more information, the object extracting " key point " from image, extracts the information that can represent this image vision content train), according to new image copy after former figure and attack, (the representational local mode that the present invention selects is exactly that those had both appeared in the important partial structurtes of image, and the visual signature therefrom extracted to remain unchanged again before and after attacking, using these primitives as image) the Model Establishment local mode storehouse of partial block, select representative local mode and mainly consider 2 points: one, shared by the representative local mode kind selected, amount of new information will account for more than 80% of gross information content, two, the representative local mode kind selected wants high through image change robustness.Pick out representative local mode according to these two principles, and apply it to and build storehouse stage and online query stage;
Build the storehouse stage, in storehouse, first each image carries out Image semantic classification (Image semantic classification comprises: longer sides is proportionally adjusted to 320 by input picture, then the Gaussian convolution of 5 × 5 is used to carry out image smoothing), then the key point in FAST algorithm extraction image is used, set up finger image according to the distribution of the representative local mode of block around these key points and (first image is divided into 4 quadrants, then in 4 quadrants, add up the histogram distribution that the training stage draws local representative mode respectively, finally represent pattern histogram quantity comparison spatially according to local and set up finger image), and mate for the online query stage stored in finger image storehouse,
The online query stage, this stage accepts a secondary sample image, the process extracting finger image feature is identical with building the storehouse stage, after obtaining the binary string of finger image feature, the instruction of machine optimization is used to ask for Hamming distances with each finger image of planting modes on sink characteristic respectively, Hamming distances larger explanation image similarity is less, and therefore candidate image is sorted from small to large according to Hamming distances, gets front K and returns as Query Result.
Be below one embodiment of the invention, as follows:
The stability of representative local mode, as shown in Fig. 2 a, 2b, 2c, 2d, 2e, 2f, Fig. 2 a is original image, 5 kinds of image attacks are simulated respectively from original image, and on corresponding image, extract the distribution histogram of the representative local mode of Top5, although find that image have passed through various attack, the histogram distribution trend of its local mode is substantially constant.
Local mode defines, and suppose that in image, certain block as shown in Figure 3, the point of black is the key point that FAST algorithm detects, the sampled point of its Green point representative for detecting FAST angle point.The present invention gets the block of 9 × 9 sizes around this key point, and this region is evenly divided into 9 blocks, makes C represent the average gray of central area, makes V
nrepresent the average gray of the n-th block, wherein n=1,2 ..., 7, b
nwhat be the n-th partial block with center block on gray-scale value compares zone bit, then the local mode P of this key point is defined as:
Through above-mentioned definition, the scope of local mode P is [0,255].
The present invention also proposes a kind of finger image extraction system based on representative local mode, comprising:
Set up local mode storehouse and obtain representative local mode module, for storehouse image is carried out image attack simulation process, generate new storehouse image, and extract the key point of described storehouse image and described new storehouse image, according to described key point, obtain partial block, and according to described partial block, generate local mode and set up local mode storehouse, from described local mode storehouse, obtain representative local mode, described representative local mode meets the following conditions: the quantity of information in described representative local mode accounts for more than 80% of gross information content in image; Described representative local mode, after image change, has high robust;
Set up finger image library module, for according to described representative local mode, set up the finger image of described storehouse image and described new storehouse image, and by described finger image stored in finger image storehouse;
Fingerprint comparison module, for obtaining new images, extracts the new images fingerprint of described new images, is compared by the finger image in described new images fingerprint and described finger image storehouse, to search in the image of storehouse with described new images to corresponding image.
Described set up local mode storehouse and obtain local mode P in representative local mode module be:
The average gray in region centered by C, V
nbe the average gray of the n-th partial block, wherein n=1,2 ..., 7, b
nwhat be the n-th partial block with center block on gray-scale value compares zone bit.
Described finger image library module of setting up also comprises and carries out pre-service to image, and extracts the key point of described image.
Described fingerprint comparison module comprises asks for Hamming distances by each finger image in described new images fingerprint and described finger image storehouse, sorts from small to large according to described Hamming distances to candidate image, gets front K and returns as Query Result.
The total technique effect of the present invention is: the present invention proposes a kind of image fingerprint extracting method based on representative local mode and system, and test on the data set of disclosed data set INRIA Holiday, with GIST (representational global image feature), SIFT (representational local image characteristics) and BRISK (binary features proposed in the recent period) algorithm contrasts, and the index of measurement is mAP (Average Accuracy).First experiment tests the present invention and the performance of other three kinds of methods under various image attack, as shown in Figure 4.
The present invention of second experiment test and other three kinds of methods are in feature extraction speed, contrast between EMS memory occupation and matching speed, because local image characteristics cannot support king-sized image library, therefore experiment is extra adds 10,000 interfering pictures downloaded from Flickr, and each performance index statistics is as shown in table 1:
Table 1
As can be seen from Table 1, all finger image features are loaded into the internal memory only consuming 0.16MB in internal memory by the present invention, and GIST, SIFT, BRISK consume 38.4MB, 1536MB and 128MB respectively, simultaneously finger image extract and coupling time on, than the fastest method also fast 2-3 the order of magnitude, therefore prove that the present invention is suitable with current the best way on accurately, but use at internal memory and in matching speed, have significantly advantage.
Claims (10)
1. based on an image fingerprint extracting method for representative local mode, it is characterized in that, comprising:
Step 1, storehouse image is carried out image attack simulation process, generate new storehouse image, and extract the key point of described storehouse image and described new storehouse image, according to described key point, obtain partial block, and according to described partial block, generate local mode and set up local mode storehouse, from described local mode storehouse, obtaining representative local mode;
Step 2, according to described representative local mode, sets up the finger image of described storehouse image and described new storehouse image, and by described finger image stored in finger image storehouse;
Step 3, obtains new images, extracts the new images fingerprint of described new images, compared by the finger image in described new images fingerprint and described finger image storehouse, to search in the image of storehouse with described new images to corresponding image.
2., as claimed in claim 1 based on the image fingerprint extracting method of representative local mode, it is characterized in that, described in described step 1, representative local mode meets the following conditions:
Quantity of information in described representative local mode accounts for more than 80% of gross information content in image;
Described representative local mode, after image change, has high robust.
3., as claimed in claim 1 based on the image fingerprint extracting method of representative local mode, it is characterized in that, in described step 1, local mode P is:
The average gray in region centered by C, V
nbe the average gray of the n-th partial block, wherein n=1,2 ..., 7, b
nwhat be the n-th partial block with center block on gray-scale value compares zone bit.
4., as claimed in claim 1 based on the image fingerprint extracting method of representative local mode, it is characterized in that, described step 2 also comprises carries out pre-service to image, and extracts the key point of described image.
5. as claimed in claim 1 based on the image fingerprint extracting method of representative local mode, it is characterized in that, described step 3 comprises asks for Hamming distances by each finger image in described new images fingerprint and described finger image storehouse, candidate image is sorted from small to large according to described Hamming distances, gets front K and return as Query Result.
6., based on a finger image extraction system for representative local mode, it is characterized in that, comprising:
Set up local mode storehouse and obtain representative local mode module, for storehouse image is carried out image attack simulation process, generate new storehouse image, and extract the key point of described storehouse image and described new storehouse image, according to described key point, obtain partial block, and according to described partial block, generate local mode and set up local mode storehouse, from described local mode storehouse, obtaining representative local mode;
Set up finger image library module, for according to described representative local mode, set up the finger image of described storehouse image and described new storehouse image, and by described finger image stored in finger image storehouse;
Fingerprint comparison module, for obtaining new images, extracts the new images fingerprint of described new images, is compared by the finger image in described new images fingerprint and described finger image storehouse, to search in the image of storehouse with described new images to corresponding image.
7., as claimed in claim 6 based on the finger image extraction system of representative local mode, it is characterized in that, described set up local mode storehouse and obtain representative local mode described in representative local mode module meet the following conditions:
Quantity of information in described representative local mode accounts for more than 80% of gross information content in image;
Described representative local mode, after image change, has high robust.
8., as claimed in claim 6 based on the finger image extraction system of representative local mode, it is characterized in that, described set up local mode storehouse and obtain local mode P in representative local mode module be:
The average gray in region centered by C, V
nbe the average gray of the n-th partial block, wherein n=1,2 ..., 7, b
nwhat be the n-th partial block with center block on gray-scale value compares zone bit.
9. as claimed in claim 6 based on the finger image extraction system of representative local mode, it is characterized in that, described finger image library module of setting up also comprises and carries out pre-service to image, and extracts the key point of described image.
10. as claimed in claim 6 based on the finger image extraction system of representative local mode, it is characterized in that, described fingerprint comparison module comprises asks for Hamming distances by each finger image in described new images fingerprint and described finger image storehouse, candidate image is sorted from small to large according to described Hamming distances, gets front K and return as Query Result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510241287.1A CN104881668B (en) | 2015-05-13 | 2015-05-13 | A kind of image fingerprint extracting method and system based on representative local mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510241287.1A CN104881668B (en) | 2015-05-13 | 2015-05-13 | A kind of image fingerprint extracting method and system based on representative local mode |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104881668A true CN104881668A (en) | 2015-09-02 |
| CN104881668B CN104881668B (en) | 2018-08-10 |
Family
ID=53949155
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510241287.1A Active CN104881668B (en) | 2015-05-13 | 2015-05-13 | A kind of image fingerprint extracting method and system based on representative local mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104881668B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105740858A (en) * | 2016-01-26 | 2016-07-06 | 南京风力舰信息技术有限公司 | Region-of-interest extraction based image copy detection method |
| CN106201290A (en) * | 2016-07-13 | 2016-12-07 | 南昌欧菲生物识别技术有限公司 | A kind of control method based on fingerprint, device and terminal |
| WO2017054442A1 (en) * | 2015-09-30 | 2017-04-06 | 腾讯科技（深圳）有限公司 | Image information recognition processing method and device, and computer storage medium |
| CN111177071A (en) * | 2019-12-12 | 2020-05-19 | 广州地理研究所 | Picture downloading method and device of Flickr platform |
| CN111461091A (en) * | 2020-06-19 | 2020-07-28 | 腾讯科技（深圳）有限公司 | Universal fingerprint generation method and device, storage medium and electronic device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1514409A (en) * | 2003-07-28 | 2004-07-21 | 西安电子科技大学 | Small wave region digital water marking mathod based on image target region |
| US20090213241A1 (en) * | 2008-02-26 | 2009-08-27 | Olympus Corporation | Image processing apparatus, image processing method, and recording medium |
| CN101853486A (en) * | 2010-06-08 | 2010-10-06 | 华中科技大学 | Image copying detection method based on local digital fingerprint |
-
2015
- 2015-05-13 CN CN201510241287.1A patent/CN104881668B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1514409A (en) * | 2003-07-28 | 2004-07-21 | 西安电子科技大学 | Small wave region digital water marking mathod based on image target region |
| US20090213241A1 (en) * | 2008-02-26 | 2009-08-27 | Olympus Corporation | Image processing apparatus, image processing method, and recording medium |
| CN101853486A (en) * | 2010-06-08 | 2010-10-06 | 华中科技大学 | Image copying detection method based on local digital fingerprint |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017054442A1 (en) * | 2015-09-30 | 2017-04-06 | 腾讯科技（深圳）有限公司 | Image information recognition processing method and device, and computer storage medium |
| CN106560840A (en) * | 2015-09-30 | 2017-04-12 | 腾讯科技（深圳）有限公司 | Recognition processing method and device of image information |
| CN106560840B (en) * | 2015-09-30 | 2019-08-13 | 腾讯科技（深圳）有限公司 | A kind of image information identifying processing method and device |
| US10438086B2 (en) | 2015-09-30 | 2019-10-08 | Tencent Technology (Shenzhen) Company Limited | Image information recognition processing method and device, and computer storage medium |
| CN105740858A (en) * | 2016-01-26 | 2016-07-06 | 南京风力舰信息技术有限公司 | Region-of-interest extraction based image copy detection method |
| CN105740858B (en) * | 2016-01-26 | 2018-12-25 | 南京风力舰信息技术有限公司 | A kind of image copy detection method based on region of interesting extraction |
| CN106201290A (en) * | 2016-07-13 | 2016-12-07 | 南昌欧菲生物识别技术有限公司 | A kind of control method based on fingerprint, device and terminal |
| CN111177071A (en) * | 2019-12-12 | 2020-05-19 | 广州地理研究所 | Picture downloading method and device of Flickr platform |
| CN111177071B (en) * | 2019-12-12 | 2023-07-07 | 广州地理研究所 | Picture downloading method and device of Flickr platform |
| CN111461091A (en) * | 2020-06-19 | 2020-07-28 | 腾讯科技（深圳）有限公司 | Universal fingerprint generation method and device, storage medium and electronic device |
| CN111461091B (en) * | 2020-06-19 | 2020-09-29 | 腾讯科技（深圳）有限公司 | Universal fingerprint generation method and device, storage medium and electronic device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104881668B (en) | 2018-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021115123A1 (en) | Method for footprint image retrieval | |
| CN104881668A (en) | Method and system for extracting image fingerprints based on representative local mode | |
| Kadam et al. | Detection and localization of multiple image splicing using MobileNet V1 | |
| Chu et al. | Image retrieval based on a multi-integration features model | |
| CN103366178A (en) | Method and device for carrying out color classification on target image | |
| CN109101981B (en) | Loop detection method based on global image stripe code in streetscape scene | |
| CN104036012A (en) | Dictionary learning method, visual word bag characteristic extracting method and retrieval system | |
| KR20010055492A (en) | Image retrival system and method using image histogram | |
| CN105404657A (en) | CEDD feature and PHOG feature based image retrieval method | |
| CN112561976A (en) | Image dominant color feature extraction method, image retrieval method, storage medium and device | |
| Reta et al. | Color uniformity descriptor: An efficient contextual color representation for image indexing and retrieval | |
| CN110334628B (en) | Outdoor monocular image depth estimation method based on structured random forest | |
| CN107977964A (en) | Slit cropping evidence collecting method based on LBP and extension Markov feature | |
| Zhou | Image retrieval: feature primitives, feature representation, and relevance feedback | |
| CN110110120B (en) | Image retrieval method and device based on deep learning | |
| CN108304588B (en) | Image retrieval method and system based on k neighbor and fuzzy pattern recognition | |
| CN112232249B (en) | Remote sensing image change detection method and device based on depth characteristics | |
| CN111414958B (en) | Multi-feature image classification method and system for visual word bag pyramid | |
| CN109299295B (en) | Blue printing layout database searching method | |
| CN110674334A (en) | Near-repetitive image retrieval method based on consistency region deep learning features | |
| CN116415210A (en) | Image infringement detection method, device and storage medium | |
| Kothyari et al. | Content based image retrieval using statistical feature and shape extraction | |
| CN107194930B (en) | Cultural relic surface texture feature extraction method based on cellular automaton | |
| CN109977849B (en) | Image texture feature fusion extraction method based on trace transformation | |
| Shrinivasacharya et al. | Content based image retrieval system using texture and modified block truncation coding |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |