CN104794183A - Picture labeling method based on multiple views and multiple labels - Google Patents
Picture labeling method based on multiple views and multiple labels Download PDFInfo
- Publication number
- CN104794183A CN104794183A CN201510169472.4A CN201510169472A CN104794183A CN 104794183 A CN104794183 A CN 104794183A CN 201510169472 A CN201510169472 A CN 201510169472A CN 104794183 A CN104794183 A CN 104794183A
- Authority
- CN
- China
- Prior art keywords
- picture
- text label
- label
- labels
- 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
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a picture labeling method based on multiple views and multiple labels. The method includes the steps that pictures and text labels are obtained from the internet, and a picture database and a text label database are established; picture features are extracted, picture views are established, and each picture view includes one picture feature; a text label relation tree is established, and the incidence relation among the text labels is excavated; for each picture view, the picture is labeled with the corresponding text label; similar label results are established in different picture views for the text labels with the incidence relation; each picture in the database is labeled with multiple text labels related to the picture. According to the picture labeling method based on multiple views and multiple labels, the relation among multiple views of pictures, the relation among the labels and the relation between the pictures and the labels are excavated at the same time, and study and popularization of the picture labeling technology are facilitated.
Description
Technical field
The present invention relates to the technical field of picture mask method, based on the picture mask method of the many labels of multi views.
Background technology
In recent years, along with the explosion type of digital camera in people's daily life is popularized, people are always submerged in a large amount of retrievable picture.But these pictures often major part do not comprise markup information.In order to effectively manage, obtain and retrieve these multi-medium datas, extensive adopted method is the content corresponding relationship by text label and picture.Had these text labels, the search problem of picture just can change into text retrieval problem, thus substantially increases the validity of calculating and the accuracy of retrieval.Always not only time-consuming but also require great effort owing to manually marking, so the effective ways naturally becoming and be applied to picture mark propagated by semi-supervised many labels.First user needs mark sub-fraction picture, and then the remaining picture that do not mark can mark picture with these and mutually works in coordination with, thus automatic learning and infer corresponding text marking information.
Generally speaking, key one step of automatic picture mark task is that the visual signature extracting picture is expressed as the machine of picture.But we but can extract the feature of not homology from picture, i.e. multi views feature.The visual characteristic that different feature interpretation picture is different, can help user to understand image content to some extent.Existing research has proposed the various picture mask method for multi views problem, but they have ignored the relevance between view and view.Although there are some methods based on sparse expression to be devoted to study the select permeability of not homology picture feature, they are just directly merged into a unified view different types of picture feature.
Second committed step of automatic picture mark task each is not marked picture to connect with some given text labels.But existing many label for labelling work is limited to substantially (or part is limited to) in the face of the propagation of many labels but still independently considers each label.
As far as we know, also do not have a kind of effective picture mask method can solve contact problem between the multi views problem of picture feature and many labels so far simultaneously.
Summary of the invention
The present invention will overcome the above-mentioned shortcoming of prior art, provides a kind of picture mask method based on the many labels of multi views, to solve multi views Characteristic Problem and many labels propagation problem simultaneously.
Based on a picture mask method for the many labels of multi views, comprise the steps:
1) obtain picture and text label from internet, set up picture database and text label database;
2) extract picture feature, set up picture view, each picture view comprises a kind of picture feature;
3) set up text label relational tree, excavate the incidence relation between text label;
4) for each picture view, be picture mark text label;
5) for relevant text label sets up similar annotation results in different picture view;
6) for the every pictures in database marks with it related multiple text label.
Step 2) described in picture feature, comprising:
1) extract the color histogram of picture, obtain the global characteristics of 256 dimensions, and form a view;
2) extract the SIFT feature point of picture, and cluster obtains the local feature of 500 dimensions, and form a view.
Step 3) described in text label relational tree, comprising: according to the classification of known text label, being labeled as subtab under class label by belonging to similar text label, traveling through all classes, setting up text label relational tree.
Step 5) described in relevant text label, in text label relational tree, namely belong to the label of same parent.
The present invention proposes a kind of transmission method of the many labels based on multi views completely newly, multi views Characteristic Problem and many labels propagation problem can be solved simultaneously.The core concept of algorithm comprises following two aspects: the label on the different picture feature views of (1) identical picture is propagated can not differ too many; (2) related label should have similar propagation.
Advantage of the present invention is: can solve multi views Characteristic Problem and many labels propagation problem simultaneously.
Accompanying drawing explanation
Fig. 1 is method flow diagram of the present invention.
Embodiment
Based on a picture mask method for the many labels of multi views, comprise the steps:
1) obtain picture and text label from internet, set up picture database and text label database;
2) extract picture feature, set up picture view, each picture view comprises a kind of picture feature;
3) set up text label relational tree, excavate the incidence relation between text label;
4) for each picture view, be picture mark text label;
5) for relevant text label sets up similar annotation results in different picture view;
6) for the every pictures in database marks with it related multiple text label.
Step 2) described in picture feature, comprising:
1) extract the color histogram of picture, obtain the global characteristics of 256 dimensions, and form a view;
2) extract the SIFT feature point of picture, and cluster obtains the local feature of 500 dimensions, and form a view.
Step 3) described in text label relational tree, comprising: according to the classification of known text label, being labeled as subtab under class label by belonging to similar text label, traveling through all classes, setting up text label relational tree.
Step 5) described in relevant text label, in text label relational tree, namely belong to the label of same parent.
Content described in this instructions embodiment is only enumerating the way of realization of inventive concept; should not being regarded as of protection scope of the present invention is only limitted to the concrete form that embodiment is stated, protection scope of the present invention also and conceive the equivalent technologies means that can expect according to the present invention in those skilled in the art.
Claims (4)
1., based on a picture mask method for the many labels of multi views, comprise the steps:
1) obtain picture and text label from internet, set up picture database and text label database;
2) extract picture feature, set up picture view, each picture view comprises a kind of picture feature;
3) set up text label relational tree, excavate the incidence relation between text label;
4) for each picture view, be picture mark text label;
5) for relevant text label sets up similar annotation results in different picture view;
6) for the every pictures in database marks with it related multiple text label.
2. a kind of picture mask method based on the many labels of multi views as claimed in claim 1, is characterized in that, step 2) described in picture feature, comprising:
21) extract the color histogram of picture, obtain the global characteristics of 256 dimensions, and form a view;
22) extract the SIFT feature point of picture, and cluster obtains the local feature of 500 dimensions, and form a view.
3. a kind of picture mask method based on the many labels of multi views as claimed in claim 1, it is characterized in that, step 3) described in text label relational tree, comprise: classify according to known text label, subtab under class label is labeled as by belonging to similar text label, travel through all classes, set up text label relational tree.
4. a kind of picture mask method based on the many labels of multi views as claimed in claim 1, is characterized in that, step 5) described in relevant text label, in text label relational tree, namely belong to the label of same parent.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510169472.4A CN104794183A (en) | 2015-04-10 | 2015-04-10 | Picture labeling method based on multiple views and multiple labels |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510169472.4A CN104794183A (en) | 2015-04-10 | 2015-04-10 | Picture labeling method based on multiple views and multiple labels |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104794183A true CN104794183A (en) | 2015-07-22 |
Family
ID=53558975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510169472.4A Pending CN104794183A (en) | 2015-04-10 | 2015-04-10 | Picture labeling method based on multiple views and multiple labels |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104794183A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292341A (en) * | 2017-06-20 | 2017-10-24 | 西安电子科技大学 | Adaptive multi views clustering method based on paired collaboration regularization and NMF |
| CN110727820A (en) * | 2019-10-22 | 2020-01-24 | 杭州数澜科技有限公司 | Method and system for obtaining label for picture |
| CN113536184A (en) * | 2021-07-15 | 2021-10-22 | 广东工业大学 | User division method and system based on multi-source information |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8027940B2 (en) * | 2007-06-13 | 2011-09-27 | Microsoft Corporation | Classification of images as advertisement images or non-advertisement images |
| CN103714178A (en) * | 2014-01-08 | 2014-04-09 | 北京京东尚科信息技术有限公司 | Automatic image marking method based on word correlation |
| CN103853792A (en) * | 2012-12-07 | 2014-06-11 | 中兴通讯股份有限公司 | Automatic image semantic annotation method and system |
| US8787683B1 (en) * | 2009-07-17 | 2014-07-22 | Google Inc. | Image classification |
| CN103955462A (en) * | 2014-03-21 | 2014-07-30 | 南京邮电大学 | Image marking method based on multi-view and semi-supervised learning mechanism |
-
2015
- 2015-04-10 CN CN201510169472.4A patent/CN104794183A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8027940B2 (en) * | 2007-06-13 | 2011-09-27 | Microsoft Corporation | Classification of images as advertisement images or non-advertisement images |
| US8787683B1 (en) * | 2009-07-17 | 2014-07-22 | Google Inc. | Image classification |
| CN103853792A (en) * | 2012-12-07 | 2014-06-11 | 中兴通讯股份有限公司 | Automatic image semantic annotation method and system |
| CN103714178A (en) * | 2014-01-08 | 2014-04-09 | 北京京东尚科信息技术有限公司 | Automatic image marking method based on word correlation |
| CN103955462A (en) * | 2014-03-21 | 2014-07-30 | 南京邮电大学 | Image marking method based on multi-view and semi-supervised learning mechanism |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292341A (en) * | 2017-06-20 | 2017-10-24 | 西安电子科技大学 | Adaptive multi views clustering method based on paired collaboration regularization and NMF |
| CN107292341B (en) * | 2017-06-20 | 2019-12-10 | 西安电子科技大学 | self-adaptive multi-view clustering method based on pair-wise collaborative regularization and NMF |
| CN110727820A (en) * | 2019-10-22 | 2020-01-24 | 杭州数澜科技有限公司 | Method and system for obtaining label for picture |
| CN113536184A (en) * | 2021-07-15 | 2021-10-22 | 广东工业大学 | User division method and system based on multi-source information |
| CN113536184B (en) * | 2021-07-15 | 2022-05-31 | 广东工业大学 | User division method and system based on multi-source information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102142089B (en) | Semantic binary tree-based image annotation method | |
| Workman et al. | Wide-area image geolocalization with aerial reference imagery | |
| CN104199972B (en) | A kind of name entity relation extraction and construction method based on deep learning | |
| CN101950284B (en) | Chinese word segmentation method and system | |
| CN102833638B (en) | Automatic video segmentation and annotation method and system based on caption information | |
| CN102982076B (en) | Based on the various dimensions content mask method in semantic label storehouse | |
| CN103049750B (en) | Character identifying method | |
| CN102136218B (en) | Automatic collocation method for map lettering during map making through computer | |
| CN102722709B (en) | Method and device for identifying garbage pictures | |
| CN103984738A (en) | Role labelling method based on search matching | |
| CN104035975B (en) | It is a kind of to realize the method that remote supervisory character relation is extracted using Chinese online resource | |
| CN104156352A (en) | Method and system for handling Chinese event | |
| CN106055667A (en) | Method for extracting core content of webpage based on text-tag density | |
| CN107967494B (en) | Image region labeling method based on visual semantic relation graph | |
| CN104598577A (en) | Extraction method for webpage text | |
| CN101957919A (en) | Character recognition method based on image local feature retrieval | |
| CN101685464A (en) | Method for automatically labeling images based on community potential subject excavation | |
| CN104572735A (en) | Photo tagging word recommending method and device | |
| CN103268330A (en) | User interest extraction method based on image content | |
| CN105138538A (en) | Cross-domain knowledge discovery-oriented topic mining method | |
| CN110705292B (en) | Entity name extraction method based on knowledge base and deep learning | |
| CN104794183A (en) | Picture labeling method based on multiple views and multiple labels | |
| Li et al. | A method based on an adaptive radius cylinder model for detecting pole-like objects in mobile laser scanning data | |
| CN103473275A (en) | Automatic image labeling method and automatic image labeling system by means of multi-feature fusion | |
| CN104063701B (en) | Fast electric television stations TV station symbol recognition system and its implementation based on SURF words trees and template matches |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150722 |
|
| WD01 | Invention patent application deemed withdrawn after publication |