CN103279545A - Method for preliminarily retrieving images - Google Patents
Method for preliminarily retrieving images Download PDFInfo
- Publication number
- CN103279545A CN103279545A CN2013102222238A CN201310222223A CN103279545A CN 103279545 A CN103279545 A CN 103279545A CN 2013102222238 A CN2013102222238 A CN 2013102222238A CN 201310222223 A CN201310222223 A CN 201310222223A CN 103279545 A CN103279545 A CN 103279545A
- Authority
- CN
- China
- Prior art keywords
- image
- approximate
- approximation
- function
- retrieval
- 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
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a method for preliminarily retrieving images. The method includes extracting attributes and attribute values of each object in a certain graph library or a certain image library, and creating a graph feature library or an image feature library; acquiring concurrence frequency values of all the attributes in the graph feature library or the image feature library; acquiring an uncertainty function according to the frequency values; acquiring a membership function according to the uncertainty function; acquiring lower approximation and upper approximation of each object according to the membership function; determining lower approximation and upper approximation of an inquiry; respectively roughly matching the lower approximation and the upper approximation of the inquiry with the lower approximation and the upper approximation of each object; and sequentially outputting matching results. The method has the advantages that graphs or images in a database are queued according to correlation degrees of the graphs or the images in the database to requirements of a user (on a target graph or a target image) and are provided for the user according to a certain sequence, so that the initial retrieval range is greatly reduced, the retrieval time is greatly shortened, and the retrieval efficiency is improved.
Description
Technical field
The present invention relates to image processing field, particularly a kind of image preindexing method.
Background technology
In recent years, the application of rough set theory (rough set) and research have obtained very fast development.Effective mathematical tool of rough set theory and uncertain problem ambiguous as a kind of new processing, not needing to be characterized in the quantity of some feature given in advance or attribute to describe, as the probability distribution in the statistics, degree of membership in the fuzzy set theory or subordinate function etc., but directly gather from the description of given problem, by can not differentiate the relation and can not differentiate the approximate territory that class is determined given problem, thereby find out the inherent law in this problem.
Current some scholar has been applied to rough set theory for example utilize rough set theory to carry out the methods such as research of image characteristics extraction and image enchancing method in treatment of picture and the understanding field.But early stage work all is to mate by the relation of equivalence in the domain, therefore is called as rough set model of equal value, is called for short ERSM.The main contribution of ERSM provides a kind of new method of semantic relation being carried out equivalence classification.It requires three attribute (reflexivitys: XRX of relation of equivalence; Symmetry: XRY → YRX; Transitivity: XRY ∧ YRZ → XRZ; ) all must satisfy simultaneously.But always can not satisfy in some field transitivity, especially in natural language understanding and information retrieval field.
For this reason, some expert has proposed to substitute with " compatibility relation " the new information retrieval concept of " relation of equivalence ", is called " compatible rough set model ", is called for short TRSM.TRSM has obtained gratifying result in the document information retrieval field, for example Tu Bao Ho and the Kaname Funakoshi of Japanese science and technology institute once utilized compatible rough set theory, all articles and paper in 1986 to these 10 years nineteen ninety-five to Japanese artificial intelligence association are retrieved, this database has 802 texts, wherein 725 texts have keyword, always have 1823 keywords, several keywords are wherein retrieved, consequently retrieve 95% related text altogether, its accuracy reaches 76%, and operate to retrieve with Boolean OP equally, only retrieving 60% related text, its accuracy only reaches 70%.
Yet the retrieval of existing content-based image information generally is a text based, because textual description often has ambiguity, and can not accurately express the information content abundant in the image, so often be difficult to retrieve exactly.Desirable content-based image information retrieval system shows to replace textual description with image, namely at first some images are presented to the user, determine that by the user has comprised the initial pictures of some information in the target image, though these information are incomplete concerning target image, but searching system can be accordingly, range of search is locked in the effective scope as much as possible, retrieves required target image as early as possible to help the user.If therefore can design a kind of model of dealing with problems, make the solution that under the incomplete situation of image information, can both give the one group of maximum possible that goes wrong as far as possible, image indexing system then is of practical significance beyond doubt very much.Consider from the efficient that improves image retrieval work on the other hand, if range of search can be locked at the very start target image around, can shorten time of retrieval greatly, improve effectiveness of retrieval.
Summary of the invention
The object of the present invention is to provide a kind of image preindexing method, solving existing image preindexing method under the incomplete prerequisite of initial graphics image information, retrieval time long, efficient difference and low problem.
For solving the problems of the technologies described above, the invention provides a kind of image preindexing method, comprising:
Step 1: extract attribute and the property value of each object in certain figure or the image library, set up a figure or characteristics of image storehouse;
Step 2: obtain the simultaneous frequency values f=C(ti of all properties in described figure or the characteristics of image storehouse, tj), C representative function, the property value of ti, ji indicated object;
Step 3: obtain nondeterministic function I(ti according to frequency values f)=tj|C(ti, tj) 〉=2} ∪ { ti};
Step 4: obtain membership function according to nondeterministic function
Step 5: according to membership function u(ti, X), obtain the following approximate of each object and go up approximate:
L（d，X）={ti∈J|v（I（ti），X）=1}
U（d，X）={ti∈J|v（I（ti），X）＞0}
Step 6: determine the following approximate L(q of inquiry, x) with last approximate U(q, x);
Step 7: the following approximate L(q that will inquire about, x) and last approximate U(q, x) respectively with the following approximate L(d of each object, X) and last approximate U(d, X) slightly mate;
Step 8: the result that will slightly mate exports successively.
In described image preindexing method, in step 3, if certain property value does not send simultaneously with any other property value, then its nondeterministic function is itself.
In described image preindexing method, the step that the result of mating is exported successively comprises:
If seven empty set A
11, A
21, A
22, A
31, A
32, A
41, A
42
The result of slightly mating is divided into seven grades;
Seven grades are put into seven empty sets;
With slightly the coupling the result according to A
11, A
21, A
22, A
31, A
32, A
41, A
42Order export successively.
A kind of image preindexing method provided by the invention, have following beneficial effect: with the figure in the database or image according to ranking with the strong and weak degree that is associated of customer requirements (targeted graphical or image), and offer the user sequentially, make the initial retrieval scope reduce greatly, shorten retrieval time greatly, improve effectiveness of retrieval.
Description of drawings
Fig. 1 is the schematic flow sheet of the image preindexing method of the embodiment of the invention.
Embodiment
Below in conjunction with the drawings and specific embodiments the image preindexing method that the present invention proposes is described in further detail.According to the following describes and claims, advantages and features of the invention will be clearer.It should be noted that accompanying drawing all adopts very the form of simplifying and all uses non-ratio accurately, only in order to convenient, the purpose of the aid illustration embodiment of the invention lucidly.
The retrieval of image is by the essential characteristic (attribute) of appointed object and the retrieval of their eigenwert (property value) are realized.An image indexing system can be described below by a quadruple attribute:
S=(J, D, Q, α); Wherein,
J={t1，t2，。。。, tm} is the set of attribute;
D={d1，d2，。。。, dn} is the set of object of being retrieved;
Q={q1，q2，。。。, qp} is the community set of target retrieval object;
α: Q*P → R is expression about the be retrieved function of correlation degree between the object dj of certain target retrieval Qi and certain.
For example given any two objects dj1 and dj2(dj1, dj2 ∈ D) and certain target retrieval q, if α (q, dj1)＞(q dj2) can think that then the relevance of target retrieval q and dj1 is better than dj2 to α.
Further, compatible rough set theory also can define with a quadruple attribute: R={U, I, v, P}.Wherein,
A domain of U indicated object;
I is a nondeterministic function, it can be any function that satisfies following condition: iff x ∈ I(y and y ∈ I(x x ∈ I(x))), for any x, y ∈ U, therefore can think I(x) be a compatible class, because it only satisfies reflexivity and symmetry, might not satisfy transitivity, it has comprised the object that all and x have analog information;
V is membership function, is used for measuring the degree that comprises between the subclass, determines especially whether compatible class is included in the domain in certain subset X;
P is a structure function, for the I(x of each x ∈ U) definable two kinds of P functions, a kind of is class-structure subclass P(I(x)=1), another kind is non-structure subclass P(I(x)=0).
Then the upper and lower approximation space for the relevant compatibility relation R of any subset X can be defined as follows:
L（R，X）={t∈U|v（I（ti），X）=1}
U（R，X）={t∈U|v（I（ti），X）＞0}
As shown in Figure 1, the present invention provides a kind of image preindexing method according to compatible rough set theory, comprising:
Step 1: extract attribute and the property value of each object in certain figure or the image library, set up a figure or characteristics of image storehouse;
Step 2: obtain the simultaneous frequency values f=C(ti of all properties in described figure or the characteristics of image storehouse, tj), C representative function, the property value of ti, ji indicated object;
Step 3: obtain nondeterministic function I(ti according to frequency values f)=tj|C(ti, tj) 〉=2} ∪ { ti};
Especially, if certain property value does not send simultaneously with any other property value, then its nondeterministic function is itself.
Step 4: obtain membership function according to nondeterministic function
Step 5: according to membership function u(ti, X), obtain the following approximate of each object and go up approximate:
L（d，X）={ti∈J|v（I（ti），X）=1}
U（d，X）={ti∈J|v（I（ti），X）＞0}；
Step 6: determine the following approximate L(q of inquiry, x) with last approximate U(q, x);
Step 7: the following approximate L(q that will inquire about, x) and last approximate U(q, x) respectively with the following approximate L(d of each object, X) and last approximate U(d, X) slightly mate;
Step 8: the result that will slightly mate exports successively.
Especially, the key point of utilizing compatible rough set theory to carry out image retrieval be each object in the specified data storehouse to the degree of support of target retrieval object, degree of support is more high, illustrates that the correlation degree of this object and target retrieval object is more high.
Therefore, the present invention has just adopted compatible thick coupling, specifically comprises:
1) establishes seven empty set A
11, A
21, A
22, A
31, A
32, A
41, A
42
2) result that will slightly mate is divided into seven grades;
Concrete, seven grades are respectively:
1) accurately of equal value:
if?Q=d
jthenA
11=A
11∪{d
j}
2) thick of equal value:
if?L(R,Q)=L(R,d
j)thenA
21=A
21∪{d
j}
if?U(R,Q)=U(R,d
j)thenA
22=A
22∪{d
j}
3) slightly comprise:
if?L(R,Q)ìL(R,d
j)thenA
31=A
31∪{d
j}
if?U(R,Q)ìU(R,d
j)thenA
32=A
32∪{d
j}
if?L(R,d
j)ìL(R,Q)thenA
41=A
41∪{d
j}
if?L(R,d
j)ìL(R,Q)thenA
42=A
42∪{d
j}
3) seven grades are put into seven empty sets;
4) result that will slightly mate is according to A
11, A
21, A
22, A
31, A
32, A
41, A
42Order export successively.
From definition approximate and that be similar to down as can be known, for searched targets, it is approximate approximate more important than last down, and thick equivalence is stronger than slightly comprising relevance.So can think A
11, A
21, A
22, A
31, A
32, A
41, A
42Relevance be to present decline trend about searched targets, so the present invention with this order object in the database is retrieved out successively.
Foregoing description only is the description to preferred embodiment of the present invention, is not any restriction to the scope of the invention, and any change, modification that the those of ordinary skill in field of the present invention is done according to above-mentioned disclosure all belong to the protection domain of claims.
Claims (3)
1. an image preindexing method is characterized in that, comprising:
Step 1: extract attribute and the property value of each object in certain figure or the image library, set up a figure or characteristics of image storehouse;
Step 2: obtain the simultaneous frequency values f=C(ti of all properties in described figure or the characteristics of image storehouse, tj), C representative function, the property value of ti, ji indicated object;
Step 3: obtain nondeterministic function I(ti according to frequency values f)=tj|C(ti, tj) 〉=2} ∪ { ti};
Step 4: obtain membership function according to nondeterministic function
Step 5: according to membership function u(ti, X), obtain the following approximate of each object and go up approximate:
L（d，X）={ti∈J|v（I（ti），X）=1}
U（d，X）={ti∈J|v（I（ti），X）＞0}
Step 6: determine the following approximate L(q of inquiry, x) with last approximate U(q, x);
Step 7: the following approximate L(q that will inquire about, x) and last approximate U(q, x) respectively with the following approximate L(d of each object, X) and last approximate U(d, X) slightly mate;
Step 8: the result that will slightly mate exports successively.
2. image preindexing method as claimed in claim 1 is characterized in that, in step 3, if certain property value does not send simultaneously with any other property value, then its nondeterministic function is itself.
3. image preindexing method as claimed in claim 1 is characterized in that, the step that the result of mating is exported successively comprises:
If seven empty set A
11, A
21, A
22, A
31, A
32, A
41, A
42
The result of slightly mating is divided into seven grades;
Seven grades are put into seven empty sets;
With slightly the coupling the result according to A
11, A
21, A
22, A
31, A
32, A
41, A
42Order export successively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013102222238A CN103279545A (en) | 2013-06-05 | 2013-06-05 | Method for preliminarily retrieving images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013102222238A CN103279545A (en) | 2013-06-05 | 2013-06-05 | Method for preliminarily retrieving images |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103279545A true CN103279545A (en) | 2013-09-04 |
Family
ID=49062064
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013102222238A Pending CN103279545A (en) | 2013-06-05 | 2013-06-05 | Method for preliminarily retrieving images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103279545A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104462466A (en) * | 2014-12-17 | 2015-03-25 | 北京百度网讯科技有限公司 | Method and device for inquiring mathematic calculation information |
| CN104636330A (en) * | 2013-11-06 | 2015-05-20 | 北京航天长峰科技工业集团有限公司 | Related video rapid searching method based on structural data |
| CN106169065A (en) * | 2016-06-30 | 2016-11-30 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN108491414A (en) * | 2018-02-05 | 2018-09-04 | 中国科学院信息工程研究所 | A kind of online abstracting method of news content and system of fusion topic feature |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101373518A (en) * | 2008-06-28 | 2009-02-25 | 合肥工业大学 | Method for constructing prototype vector and reconstructing sequence parameter based on semantic information in image comprehension |
-
2013
- 2013-06-05 CN CN2013102222238A patent/CN103279545A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101373518A (en) * | 2008-06-28 | 2009-02-25 | 合肥工业大学 | Method for constructing prototype vector and reconstructing sequence parameter based on semantic information in image comprehension |
Non-Patent Citations (1)
| Title |
|---|
| 胡静等: "基于相容粗糙集的图形图像信息预检索", 《计算机辅助设计与图形学学报》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104636330A (en) * | 2013-11-06 | 2015-05-20 | 北京航天长峰科技工业集团有限公司 | Related video rapid searching method based on structural data |
| CN104462466A (en) * | 2014-12-17 | 2015-03-25 | 北京百度网讯科技有限公司 | Method and device for inquiring mathematic calculation information |
| CN106169065A (en) * | 2016-06-30 | 2016-11-30 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
| CN106169065B (en) * | 2016-06-30 | 2019-12-24 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN108491414A (en) * | 2018-02-05 | 2018-09-04 | 中国科学院信息工程研究所 | A kind of online abstracting method of news content and system of fusion topic feature |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9104749B2 (en) | Semantically aggregated index in an indexer-agnostic index building system | |
| US8489650B2 (en) | System, implementation, application, and query language for a tetrahedral data model for unstructured data | |
| CN107590128B (en) | Paper homonymy author disambiguation method based on high-confidence characteristic attribute hierarchical clustering method | |
| US20120117051A1 (en) | Multi-modal approach to search query input | |
| US20060206483A1 (en) | Method for domain identification of documents in a document database | |
| Mahmoud et al. | Schema clustering and retrieval for multi-domain pay-as-you-go data integration systems | |
| EP2021959A2 (en) | Annotation by search | |
| CN107291895B (en) | Quick hierarchical document query method | |
| CN107844493B (en) | File association method and system | |
| CN111026710A (en) | Data set retrieval method and system | |
| US20180276244A1 (en) | Method and system for searching for similar images that is nearly independent of the scale of the collection of images | |
| CN102314464B (en) | Lyrics searching method and lyrics searching engine | |
| CN106055539A (en) | Name disambiguation method and apparatus | |
| US8700624B1 (en) | Collaborative search apps platform for web search | |
| CN115563313A (en) | Knowledge graph-based document book semantic retrieval system | |
| CN103279545A (en) | Method for preliminarily retrieving images | |
| CN110750599A (en) | Associated information extraction and display method based on entity modeling | |
| CN102915381B (en) | Visual network retrieval based on multi-dimensional semantic presents system and presents control method | |
| CN112836008B (en) | Index establishing method based on decentralized storage data | |
| Toba et al. | Enhanced unsupervised person name disambiguation to support alumni tracer study | |
| WO2012091539A1 (en) | A semantic similarity matching system and a method thereof | |
| CN115687773A (en) | Cross-environment metadata matching method and system based on knowledge graph | |
| Fize et al. | Matching heterogeneous textual data using spatial features | |
| WO2013177751A1 (en) | Cross-media retrieval method based on generalized linear regression model | |
| c Neethu et al. | Retrieval of images using data mining techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130904 |