CN107491456A - Image ranking method and device - Google Patents

Image ranking method and device Download PDF

Info

Publication number
CN107491456A
CN107491456A CN201610412993.2A CN201610412993A CN107491456A CN 107491456 A CN107491456 A CN 107491456A CN 201610412993 A CN201610412993 A CN 201610412993A CN 107491456 A CN107491456 A CN 107491456A
Authority
CN
China
Prior art keywords
image
text message
score
adjoint
text
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
Application number
CN201610412993.2A
Other languages
Chinese (zh)
Inventor
余宙
潘攀
华先胜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Alibaba Group Holding Ltd
Original Assignee
Alibaba Group Holding Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Alibaba Group Holding Ltd filed Critical Alibaba Group Holding Ltd
Priority to CN201610412993.2A priority Critical patent/CN107491456A/en
Publication of CN107491456A publication Critical patent/CN107491456A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/5846—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using extracted text

Landscapes

  • Engineering & Computer Science (AREA)
  • Library & Information Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The application proposes that a kind of image ranking method and device, the image ranking method include：It is determined that multiple images to be sorted；Obtain the adjoint text message of each image；According to the adjoint text message, the score of each image is calculated using figure sort algorithm；According to the score, the multiple image is ranked up.This method can be ranked up according to semantic similarity to image, lift Consumer's Experience.

Description

Image ranking method and device
Technical field
The application is related to technical field of image processing, more particularly to a kind of image ranking method and device.
Background technology
In correlation technique, picture search is normally based on the picture search of content, i.e., the given image of user, which is used as, looks into The input of inquiry, search engine extract feature after carrying out content analysis to image, several are returned in Large image database most Related image result.In addition, being also based on content when image sorts, it is ranked up according to content relevance.
But due to people to the distinguishing rule of image similarity and computer between the distinguishing rule of similitude not Together, the difference of " semantic similar " that people understood between " vision is similar " of computer understanding is caused.Here it is image to search " semantic gap " problem that rope field is widely present.
The problem of " semantic gap " would generally be faced due to the picture search based on content and sort algorithm, can not be from result In filter out these semantically incoherent images, reduce the usage experience of user.
The content of the invention
The application is intended to one of technical problem at least solving in correlation technique to a certain extent.
Therefore, the purpose of the application is to propose a kind of image ranking method, this method can be when image sorts Semantic information between reference picture, meet the needs of people is similar to semanteme, avoid semantic gap problem, so as to according to semantic Similarity is ranked up to image, lifts Consumer's Experience.
Further object is to propose a kind of image collator.
To reach above-mentioned purpose, image ranking method that the application first aspect embodiment proposes, including：It is determined that wait to sort Multiple images；Obtain the adjoint text message of each image；According to the adjoint text message, calculated using figure sort algorithm The score of each image；According to the score, the multiple image is ranked up.
The image ranking method that the application first aspect embodiment proposes, by the adjoint text message and evidence that obtain image This carries out successive image sequence, due to including more semantic informations relative to picture material with text message, therefore, is scheming Semantic information during as sequence between reference picture, meet the needs of people is similar to semanteme, avoid semantic gap problem, so as to Image is ranked up according to semantic similarity, lifts Consumer's Experience.
To reach above-mentioned purpose, image collator that the application second aspect embodiment proposes, including：Determining module, For determining multiple images to be sorted；Acquisition module, for obtaining the adjoint text message of each image；Computing module, use In the score that each image according to the adjoint text message, is calculated using figure sort algorithm；Order module, for according to Score, the multiple image is ranked up.
The image collator that the application second aspect embodiment proposes, by the adjoint text message and evidence that obtain image This carries out successive image sequence, due to including more semantic informations relative to picture material with text message, therefore, is scheming Semantic information during as sequence between reference picture, meet the needs of people is similar to semanteme, avoid semantic gap problem, so as to Image is ranked up according to semantic similarity, lifts Consumer's Experience.
The aspect and advantage that the application adds will be set forth in part in the description, and will partly become from the following description Obtain substantially, or recognized by the practice of the application.
Brief description of the drawings
The above-mentioned and/or additional aspect of the application and advantage will become from the following description of the accompanying drawings of embodiments Substantially and it is readily appreciated that, wherein：
Fig. 1 is the schematic flow sheet for the image ranking method that the application one embodiment proposes；
Fig. 2 is because " semantic gap " problem being shown by computer discriminant for similar two images in the embodiment of the present application It is intended to；
Fig. 3 is the schematic flow sheet of the image ranking method of the application another embodiment proposition；
Fig. 4 is the structural representation for the image collator that the application one embodiment proposes；
Fig. 5 is the structural representation of the image collator of the application another embodiment proposition.
Embodiment
Embodiments herein is described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end Same or similar label represents same or similar module or the module with same or like function.Below with reference to attached The embodiment of figure description is exemplary, is only used for explaining the application, and it is not intended that limitation to the application.On the contrary, this All changes that the embodiment of application includes falling into the range of the spirit and intension of attached claims, modification and equivalent Thing.
Fig. 1 is the schematic flow sheet for the image ranking method that the application one embodiment proposes.
Referring to Fig. 1, the method for the present embodiment includes：
S11：It is determined that multiple images to be sorted.
Wherein, image sequence can apply to many scenes, and therefore, multiple images to be sorted can be according to scene not Determine in different ways together.
For example, in picture search, query image can be scanned for based on the content of query image, will be obtained after search To the content degree of correlation meet the multiple images of requirement condition, be defined as multiple images to be sorted.
S12：Obtain the adjoint text message of each image.
Wherein, image when stored, generally not only can preserve image in itself, will also maintain the adjoint text envelope of image Breath, with text message for example including：Title or label etc..
More semantic informations can be being expressed with text message, for example, this image is people, animal or plant etc..
S13：According to the adjoint text message, the score of each image is calculated using figure sort algorithm.
Wherein, can be literary according to corresponding to being determined with text message after the adjoint text message of each image is obtained Eigen vector.After Text eigenvector corresponding to each image is obtained, text feature can be corresponded to according to different images Vector calculates the incidence matrix between image, and figure sort algorithm is used further according to incidence matrix, and obtaining for each image is calculated Point.
Specific calculation may refer to the description of subsequent embodiment.
S14：According to the score, the multiple image is ranked up.
, can be according to the order of score from high to low, for example, after the score of each image is obtained to image from front to back It is ranked up.
During common picture search based on content, computer when differentiating similitude using " vision is similar ", with " semantic similar " that people understands is different.For example, two image discriminatings shown in Fig. 2 can be similar by computer, and people can be with bright Aobvious determines dissmilarity, and here it is " semantic gap " problem.
The reason for above mentioned problem is present be, based on content search or during sequence, can only be arranged according to the content of image Sequence, without with reference to semantic information, and in the present embodiment, by obtaining with text message and being referenced to text message In image sequence, semantic information is may be referred to, avoids obvious semantic dissimilar content being determined as similar.
In the present embodiment, by obtaining the adjoint text message of image and carrying out successive image sequence accordingly, due to adjoint Text message includes more semantic informations relative to picture material, therefore, the semantic letter when image sorts between reference picture Breath, meet the needs of people is similar to semanteme, avoid semantic gap problem, so as to be arranged according to semantic similarity image Sequence, lift Consumer's Experience.
Fig. 3 is the schematic flow sheet of the image ranking method of the application another embodiment proposition.
The image sort algorithm of the present embodiment is exemplified by applying in picture search flow.
Referring to Fig. 3, the method for the present embodiment includes：
S31：The query image of user's input is received, passes through picture search acquisition multiple images corresponding with query image.
Wherein it is possible to using image search engine, picture search is carried out based on content, to obtain and query image content phase The multiple images of pass.
Can be specifically that be carried out in itself based on image specifically, when image search engine is based on content progress picture search Search, further, picture search is specifically carried out based on the visual signature that image extracts in itself, above-mentioned visual signature is for example Including：Scale invariant features transform (Scale-invariant feature transform, SIFT) feature, convolutional Neural net Network (Convolutional Neural Network, CNN) feature etc..
The multiple images of above-mentioned acquisition can form a set, referred to as Candidate Set, and the size of Candidate Set can use K tables Show, show that Candidate Set includes K image.
S32：Obtain the adjoint text message of each image in Candidate Set.
Wherein, with text message for example including：Title or label etc..
Specifically, can be with pre-recorded image and corresponding adjoint text message, so as to from data in database The adjoint text message of each image is got in storehouse.
These text messages can be with the semanteme of response diagram picture.
S33：Calculate each with Text eigenvector corresponding to text message.
For example, being title with text message, then first title can be segmented, be further according to predetermined size D dictionary, title is expressed as to the Text eigenvector of D dimensions.
Wherein it is possible in existing corpus, D higher word of word frequency is selected to form above-mentioned dictionary, D's is specific Value can be set.
After the dictionary that size is D is obtained, it can obtain what D corresponding to title was tieed up according to the title after dictionary and participle Text eigenvector.
Wherein, each text feature in the Text eigenvector of D dimensions can be specifically the TF- of each word in dictionary IDF values.
For example, the dictionary that size is D is expressed as：[T1, T2 ..., TD], wherein Ti (i=1,2 ..., D) represents one Word after participle；To each text, the Text eigenvector of D dimensions is expressed as corresponding to it：[V1, V2 ..., VD], wherein Vi is Equivalent Ti TF-IDF values in dictionary.
Ti TF-IDF values are that Ti TF values are multiplied by Ti IDF values.
TF is word frequency (Term Frequency), and Ti TF values can be removed with the number occurred in titles of the Ti after participle Obtained with the word sum in the title after participle.
IDF is reverse document-frequency (inverse document frequency, IDF), and Ti IDF values can use language material General act number divided by the number of the file comprising Ti in storehouse, then obtained business is taken the logarithm to obtain.
Because each adjoint text message can obtain a Text eigenvector, therefore K image in corresponding Candidate Set, K Text eigenvector can be obtained.
S34：According to Text eigenvector, the incidence matrix between image is obtained.
Wherein, each element in incidence matrix is the metric range between Text eigenvector two-by-two.
For example, sharing K Text eigenvector, then the incidence matrix S that K*K is tieed up can be obtained, it is each in incidence matrix Element S [i] [j] is the metric range between i-th of Text eigenvector and j-th of Text eigenvector, wherein, i=1, 2nd ..., K, j=1,2 ..., K.
Metric range between two above-mentioned vectors can be：COS distance between two vectors either Euclidean away from From.
S35：Using figure sort algorithm, according to above-mentioned incidence matrix, the score of each image in Candidate Set is calculated.
Wherein, figure sort algorithm includes：PageRank, HITS, TrustRank, HillTop etc..
In the present embodiment, by taking PageRank algorithms as an example.
PageRank algorithms are a kind of iterative algorithms, and iterative formula can be expressed as：
γt=α × S × γt-1+(1-α)×γ0
Wherein, S is the above-mentioned incidence matrix being calculated；γ0The initial value of interative computation, be K dimension column vector, example Such as, γ0=[1/K, 1/K ..., 1/K]T；α is PageRank algorithm parameters, for example, α=0.85.
The termination condition of above-mentioned interative computation is：γt=γt-1。
In γt=γt-1When, by γtIn the value of each element be defined as the score of correspondence image in Candidate Set, for example, γt In first element be first image in Candidate Set score, γtIn second element be second image in Candidate Set Score, wherein, γtPutting in order for the image in Candidate Set of putting in order during with starting computing of middle element is consistent, such as is It is defined as first, second etc. according to vertical order.
S36：The image in Candidate Set is ranked up according to the score of each image.
, can be according to the order of score from high to low, for example, after the score of each image is obtained to image from front to back It is ranked up.
In the present embodiment, by obtaining the adjoint text message of image and carrying out successive image sequence accordingly, due to adjoint Text message includes more semantic informations relative to picture material, therefore, the semantic letter when image sorts between reference picture Breath, meet the needs of people is similar to semanteme, avoid semantic gap problem, so as to be arranged according to semantic similarity image Sequence, lift Consumer's Experience.Further, the image sequence of the present embodiment can apply to picture search scene, with to user's exhibition Show and more meet image of the people to semantic similarity demand.
Fig. 4 is the structural representation for the image collator that the application one embodiment proposes.
Referring to Fig. 4, the device 40 of the present embodiment includes：Determining module 41, acquisition module 42, computing module 43 and sequence mould Block 44.
Determining module 41, for determining multiple images to be sorted.
Wherein, image sequence can apply to many scenes, and therefore, multiple images to be sorted can be according to scene not Determine in different ways together.
For example, in picture search, query image can be scanned for based on the content of query image, will be obtained after search To the content degree of correlation meet the multiple images of requirement condition, be defined as multiple images to be sorted.
Acquisition module 42, for obtaining the adjoint text message of each image.
Wherein, image when stored, generally not only can preserve image in itself, will also maintain the adjoint text envelope of image Breath, with text message for example including：Title or label etc..
More semantic informations can be being expressed with text message, for example, this image is people, animal or plant etc..
Computing module 43, for according to the adjoint text message, calculating the score of each image using figure sort algorithm.
Wherein, can be literary according to corresponding to being determined with text message after the adjoint text message of each image is obtained Eigen vector.After Text eigenvector corresponding to each image is obtained, text feature can be corresponded to according to different images Vector calculates the incidence matrix between image, and figure sort algorithm is used further according to incidence matrix, and obtaining for each image is calculated Point.
Order module 44, for according to the score, being ranked up to the multiple image.
, can be according to the order of score from high to low, for example, after the score of each image is obtained to image from front to back It is ranked up.
In some embodiments, referring to Fig. 5, computing module 43 includes：
First computing unit 431, for calculating each adjoint Text eigenvector corresponding to text message.
For example, being title with text message, then first title can be segmented, be further according to predetermined size D dictionary, title is expressed as to the Text eigenvector of D dimensions.
Wherein it is possible in existing corpus, D higher word of word frequency is selected to form above-mentioned dictionary, D's is specific Value can be set.
After the dictionary that size is D is obtained, it can obtain what D corresponding to title was tieed up according to the title after dictionary and participle Text eigenvector.
Wherein, each text feature in the Text eigenvector of D dimensions can be specifically the TF- of each word in dictionary IDF values.
For example, the dictionary that size is D is expressed as：[T1, T2 ..., TD], wherein Ti (i=1,2 ..., D) represents one Word after participle；To each text, the Text eigenvector of D dimensions is expressed as corresponding to it：[V1, V2 ..., VD], wherein Vi is Equivalent Ti TF-IDF values in dictionary.
Ti TF-IDF values are that Ti TF values are multiplied by Ti IDF values.
TF is word frequency (Term Frequency), and Ti TF values can be removed with the number occurred in titles of the Ti after participle Obtained with the word sum in the title after participle.
IDF is reverse document-frequency (inverse document frequency, IDF), and Ti IDF values can use language material General act number divided by the number of the file comprising Ti in storehouse, then obtained business is taken the logarithm to obtain.
Because each adjoint text message can obtain a Text eigenvector, therefore K image in corresponding Candidate Set, K Text eigenvector can be obtained.
Second computing unit 432, for according to the Text eigenvector, obtaining the incidence matrix between image, wherein, close Each element in connection matrix is the metric range between Text eigenvector two-by-two.
For example, sharing K Text eigenvector, then the incidence matrix S that K*K is tieed up can be obtained, it is each in incidence matrix Element S [i] [j] is the metric range between i-th of Text eigenvector and j-th of Text eigenvector, wherein, i=1, 2nd ..., K, j=1,2 ..., K.
Metric range between two above-mentioned vectors can be：COS distance between two vectors either Euclidean away from From.
3rd computing unit 433, for using figure sort algorithm, according to the incidence matrix, each image to be calculated Score.
Wherein, figure sort algorithm includes：PageRank, HITS, TrustRank, HillTop etc..
In the present embodiment, by taking PageRank algorithms as an example.
PageRank algorithms are a kind of iterative algorithms, and iterative formula can be expressed as：
γt=α × S × γt-1+(1-α)×γ0
Wherein, S is the above-mentioned incidence matrix being calculated；γ0The initial value of interative computation, be K dimension column vector, example Such as, γ0=[1/K, 1/K ..., 1/K]T；α is PageRank algorithm parameters, for example, α=0.85.
The termination condition of above-mentioned interative computation is：γt=γt-1。
In γt=γt-1When, by γtIn the value of each element be defined as the score of correspondence image in Candidate Set, for example, γt In first element be first image in Candidate Set score, γtIn second element be second image in Candidate Set Score, wherein, γtPutting in order for the image in Candidate Set of putting in order during with starting computing of middle element is consistent, such as is It is defined as first, second etc. according to vertical order.
In some embodiments, referring to Fig. 5, determining module 41 includes：
Receiving unit 411, for receiving the query image of user's input.
Search unit 412, will be described more for obtaining multiple images corresponding with the query image by picture search Individual image is defined as multiple images to be sorted.
Wherein it is possible to using image search engine, picture search is carried out based on content, to obtain and query image content phase The multiple images of pass.
The multiple images of above-mentioned acquisition can form a set, referred to as Candidate Set, and the size of Candidate Set can use K tables Show, show that Candidate Set includes K image.
It is understood that the device in above-described embodiment is corresponding with the method for above-described embodiment, on above-described embodiment In device, wherein modules perform operation concrete mode carried out retouching in detail in the embodiment about this method State, explanation will be not set forth in detail herein.
In the present embodiment, by obtaining the adjoint text message of image and carrying out successive image sequence accordingly, due to adjoint Text message includes more semantic informations relative to picture material, therefore, the semantic letter when image sorts between reference picture Breath, meet the needs of people is similar to semanteme, avoid semantic gap problem, so as to be arranged according to semantic similarity image Sequence, lift Consumer's Experience.Further, the image sequence of the present embodiment can apply to picture search scene, with to user's exhibition Show and more meet image of the people to semantic similarity demand.
It should be noted that in the description of the present application, term " first ", " second " etc. are only used for describing purpose, without It is understood that to indicate or implying relative importance.In addition, in the description of the present application, unless otherwise indicated, the implication of " multiple " Refer at least two.
Any process or method described otherwise above description in flow chart or herein is construed as, and represents to include Module, fragment or the portion of the code of the executable instruction of one or more the step of being used to realize specific logical function or process Point, and the scope of the preferred embodiment of the application includes other realization, wherein can not press shown or discuss suitable Sequence, including according to involved function by it is basic simultaneously in the way of or in the opposite order, carry out perform function, this should be by the application Embodiment person of ordinary skill in the field understood.
It should be appreciated that each several part of the application can be realized with hardware, software, firmware or combinations thereof.Above-mentioned In embodiment, software that multiple steps or method can be performed in memory and by suitable instruction execution system with storage Or firmware is realized.If, and in another embodiment, can be with well known in the art for example, realized with hardware Any one of row technology or their combination are realized：With the logic gates for realizing logic function to data-signal Discrete logic, have suitable combinational logic gate circuit application specific integrated circuit, programmable gate array (PGA), scene Programmable gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method carries Suddenly it is that by program the hardware of correlation can be instructed to complete, described program can be stored in a kind of computer-readable storage medium In matter, the program upon execution, including one or a combination set of the step of embodiment of the method.
In addition, each functional unit in each embodiment of the application can be integrated in a processing module, can also That unit is individually physically present, can also two or more units be integrated in a module.Above-mentioned integrated mould Block can both be realized in the form of hardware, can also be realized in the form of software function module.The integrated module is such as Fruit is realized in the form of software function module and as independent production marketing or in use, can also be stored in a computer In read/write memory medium.
Storage medium mentioned above can be read-only storage, disk or CD etc..
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show The description of example " or " some examples " etc. means specific features, structure, material or the spy for combining the embodiment or example description Point is contained at least one embodiment or example of the application.In this manual, to the schematic representation of above-mentioned term not Necessarily refer to identical embodiment or example.Moreover, specific features, structure, material or the feature of description can be any One or more embodiments or example in combine in an appropriate manner.
Although embodiments herein has been shown and described above, it is to be understood that above-described embodiment is example Property, it is impossible to the limitation to the application is interpreted as, one of ordinary skill in the art within the scope of application can be to above-mentioned Embodiment is changed, changed, replacing and modification.

Claims (10)

  1. A kind of 1. image ranking method, it is characterised in that including：
    It is determined that multiple images to be sorted；
    Obtain the adjoint text message of each image；
    According to the adjoint text message, the score of each image is calculated using figure sort algorithm；
    According to the score, the multiple image is ranked up.
  2. 2. according to the method for claim 1, it is characterised in that it is described according to the adjoint text message, sorted using figure Algorithm calculates the score of each image, including：
    Calculate each with Text eigenvector corresponding to text message；
    According to the Text eigenvector, the incidence matrix between image is obtained, wherein, each element in incidence matrix is two-by-two Metric range between Text eigenvector；
    Using figure sort algorithm, according to the incidence matrix, the score of each image is calculated.
  3. 3. according to the method for claim 2, it is characterised in that the metric range includes：COS distance or Euclidean away from From.
  4. 4. according to the method described in claim any one of 1-3, it is characterised in that described to determine multiple images to be sorted, bag Include：
    Receive the query image of user's input；
    Multiple images corresponding with the query image are obtained by picture search, the multiple image are defined as to be sorted Multiple images.
  5. 5. according to the method described in claim any one of 1-3, it is characterised in that the adjoint text message includes：Title or Person's label.
  6. A kind of 6. image collator, it is characterised in that including：
    Determining module, for determining multiple images to be sorted；
    Acquisition module, for obtaining the adjoint text message of each image；
    Computing module, for according to the adjoint text message, calculating the score of each image using figure sort algorithm；
    Order module, for according to the score, being ranked up to the multiple image.
  7. 7. device according to claim 6, it is characterised in that the computing module includes：
    First computing unit, for calculating each adjoint Text eigenvector corresponding to text message；
    Second computing unit, for according to the Text eigenvector, obtaining the incidence matrix between image, wherein, incidence matrix In each element be the metric range between Text eigenvector two-by-two；
    3rd computing unit, for using figure sort algorithm, according to the incidence matrix, the score of each image is calculated.
  8. 8. device according to claim 7, it is characterised in that the metric range bag that second computing unit calculates Include：COS distance or Euclidean distance.
  9. 9. according to the device described in claim any one of 6-8, it is characterised in that the determining module includes：
    Receiving unit, for receiving the query image of user's input；
    Search unit, for obtaining multiple images corresponding with the query image by picture search, by the multiple image It is defined as multiple images to be sorted.
  10. 10. according to the device described in claim any one of 6-8, it is characterised in that the acquisition module obtains described adjoint Text message includes：Title or label.
CN201610412993.2A 2016-06-13 2016-06-13 Image ranking method and device Pending CN107491456A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610412993.2A CN107491456A (en) 2016-06-13 2016-06-13 Image ranking method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610412993.2A CN107491456A (en) 2016-06-13 2016-06-13 Image ranking method and device

Publications (1)

Publication Number Publication Date
CN107491456A true CN107491456A (en) 2017-12-19

Family

ID=60642982

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610412993.2A Pending CN107491456A (en) 2016-06-13 2016-06-13 Image ranking method and device

Country Status (1)

Country Link
CN (1) CN107491456A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108363781A (en) * 2018-02-11 2018-08-03 维沃移动通信有限公司 A kind of picture sending method, terminal and computer readable storage medium
CN111801703A (en) * 2018-04-17 2020-10-20 赫尔实验室有限公司 Hardware and system for bounding box generation for an image processing pipeline

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101320375A (en) * 2008-07-04 2008-12-10 浙江大学 Digital book search method based on user click action
CN101901249A (en) * 2009-05-26 2010-12-01 复旦大学 Text-based query expansion and sort method in image retrieval
CN102193934A (en) * 2010-03-11 2011-09-21 株式会社理光 System and method for searching representative image of image set
CN102236655A (en) * 2010-04-26 2011-11-09 北京大学 New web page recommendation method
US20130301938A1 (en) * 2012-05-11 2013-11-14 National Taiwan University Human photo search system
CN103473327A (en) * 2013-09-13 2013-12-25 广东图图搜网络科技有限公司 Image retrieval method and image retrieval system
CN103514276A (en) * 2013-09-22 2014-01-15 西安交通大学 Graphic target retrieval positioning method based on center estimation
CN104572651A (en) * 2013-10-11 2015-04-29 华为技术有限公司 Picture ordering method and device
EP2890109A1 (en) * 2013-12-27 2015-07-01 Thomson Licensing Method for sorting a group of images of a database and method for color correcting an image, corresponding devices, computer program and non-transitory computer-readable medium

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101320375A (en) * 2008-07-04 2008-12-10 浙江大学 Digital book search method based on user click action
CN101901249A (en) * 2009-05-26 2010-12-01 复旦大学 Text-based query expansion and sort method in image retrieval
CN102193934A (en) * 2010-03-11 2011-09-21 株式会社理光 System and method for searching representative image of image set
CN102236655A (en) * 2010-04-26 2011-11-09 北京大学 New web page recommendation method
US20130301938A1 (en) * 2012-05-11 2013-11-14 National Taiwan University Human photo search system
CN103473327A (en) * 2013-09-13 2013-12-25 广东图图搜网络科技有限公司 Image retrieval method and image retrieval system
CN103514276A (en) * 2013-09-22 2014-01-15 西安交通大学 Graphic target retrieval positioning method based on center estimation
CN104572651A (en) * 2013-10-11 2015-04-29 华为技术有限公司 Picture ordering method and device
EP2890109A1 (en) * 2013-12-27 2015-07-01 Thomson Licensing Method for sorting a group of images of a database and method for color correcting an image, corresponding devices, computer program and non-transitory computer-readable medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108363781A (en) * 2018-02-11 2018-08-03 维沃移动通信有限公司 A kind of picture sending method, terminal and computer readable storage medium
CN111801703A (en) * 2018-04-17 2020-10-20 赫尔实验室有限公司 Hardware and system for bounding box generation for an image processing pipeline

Similar Documents

Publication Publication Date Title
US8930288B2 (en) Learning tags for video annotation using latent subtags
CN112131350A (en) Text label determination method, text label determination device, terminal and readable storage medium
CN103988232A (en) IMAGE MATCHING by USING MOTION MANIFOLDS
Gupta et al. Vico: Word embeddings from visual co-occurrences
Zhao et al. Multi-modal microblog classification via multi-task learning
Wang et al. Task-generic semantic convolutional neural network for web text-aided image classification
EP2874102A2 (en) Generating models for identifying thumbnail images
CN111984824A (en) Multi-mode-based video recommendation method
WO2022188844A1 (en) Video classification method and apparatus, device, and medium
Camargo et al. Multimodal latent topic analysis for image collection summarization
Wang et al. Cluster ensemble-based image segmentation
US11250299B2 (en) Learning representations of generalized cross-modal entailment tasks
CN114419391A (en) Target image identification method and device, electronic equipment and readable storage medium
Khan et al. Food items detection and recognition via multiple deep models
Kastner et al. Estimating the imageability of words by mining visual characteristics from crawled image data
Guadarrama et al. Understanding object descriptions in robotics by open-vocabulary object retrieval and detection
Sharma et al. Index point detection for text summarization using cosine similarity in educational videos
CN107491456A (en) Image ranking method and device
Vishwakarma et al. Hybrid system for MPAA ratings of movie clips using support vector machine
Mery Face analysis: state of the art and ethical challenges
Vaca-Castano et al. Semantic image search from multiple query images
Alzu'Bi et al. Robust fusion of color and local descriptors for image retrieval and classification
CN111506754A (en) Picture retrieval method and device, storage medium and processor
Bhairnallykar et al. Content based medical image retrieval with SVM classification and relevance feedback
Souvannavong et al. Latent semantic indexing for semantic content detection of video shots

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
REG Reference to a national code

Ref country code: HK

Ref legal event code: DE

Ref document number: 1248334

Country of ref document: HK

RJ01 Rejection of invention patent application after publication

Application publication date: 20171219

RJ01 Rejection of invention patent application after publication