CN107918624A - Image retrieving apparatus and method, electronic equipment - Google Patents
Image retrieving apparatus and method, electronic equipment Download PDFInfo
- Publication number
- CN107918624A CN107918624A CN201610886031.0A CN201610886031A CN107918624A CN 107918624 A CN107918624 A CN 107918624A CN 201610886031 A CN201610886031 A CN 201610886031A CN 107918624 A CN107918624 A CN 107918624A
- Authority
- CN
- China
- Prior art keywords
- input picture
- database images
- similar
- image
- grader
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Library & Information Science (AREA)
- Databases & Information Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Processing Or Creating Images (AREA)
Abstract
The embodiment of the present invention provides a kind of image retrieving apparatus and method, electronic equipment, by the way that input picture is used to train grader, and classified according to the grader to multiple database images, the database images similar to input picture can be retrieved, and, by determining the similar area similar to predetermined reference zone in input picture, and the reference zone and similar area are used to train grader, the classification accuracy of grader can be effectively improved, so as to improve the accuracy of retrieval.In addition, when needing to analyze input picture, by fast and accurately retrieving similar database images for reference, analysis efficiency and precision of analysis can be effectively improved.
Description
Technical field
The present invention relates to information technology field, more particularly to a kind of image retrieving apparatus and method, electronic equipment.
Background technology
With the continuous development of information technology, image procossing is also more and more extensive in the application of every field.In many feelings
, it is necessary to be marked and analyze based on some features in image under shape, and provide analysis result.At present, existing method one
As be the independent manual analysis for carrying out handmarking and feature based for needing each image for analyzing, and for each
Image to be analyzed independence provides analysis result.
It should be noted that the introduction to technical background above be intended merely to it is convenient technical scheme is carried out it is clear,
Complete explanation, and facilitate the understanding of those skilled in the art and illustrate.Cannot merely because these schemes the present invention
Background section is set forth and thinks that above-mentioned technical proposal is known to those skilled in the art.
The content of the invention
Above-mentioned existing method needs the carry out handmarking and analysis for each image independence to be analyzed, this method
The consuming time is longer, and depends on the experience and level of analysis personnel, is easy to cause the inaccuracy of analysis result.
The embodiment of the present invention provides a kind of image retrieving apparatus and method, electronic equipment, by the way that input picture to be used to instruct
Practice grader, and classified according to the grader to multiple database images, the number similar to input picture can be retrieved
According to storehouse image, also, by determining the similar area similar to predetermined reference zone in input picture, and this is referred to
Region and similar area are used to train grader, the classification accuracy of grader can be effectively improved, so as to improve retrieval
Accuracy.In addition, when needing to analyze input picture, by fast and accurately retrieving similar database images
For reference, analysis efficiency and precision of analysis can be effectively improved.
First aspect according to embodiments of the present invention, there is provided a kind of image retrieving apparatus, described device include：First determines
Unit, it is used to determine similar area similar to predetermined reference zone in the input picture in input picture；Instruction
Practice unit, it is used for the reference zone and the similar area in the input picture, training grader；Classification
Unit, it is used to classify to multiple database images using the grader that training obtains, and obtains each database images
Classification results；Second determination unit, it is used for the classification results according to each database images, determines the multiple database diagram
The database images similar to the input picture as in.
Second aspect according to embodiments of the present invention, there is provided a kind of electronic equipment, the electronic equipment are included according to this hair
Device described in the first aspect of bright embodiment.
The third aspect according to embodiments of the present invention, there is provided a kind of image search method, the described method includes：Determine input
The similar area similar to predetermined reference zone in the input picture in image；Institute in the input picture
State reference zone and the similar area, training grader；Using the obtained grader of training to multiple database images into
Row classification, obtains the classification results of each database images；According to the classification results of each database images, determine the multiple
The database images similar to the input picture in database images.
The beneficial effects of the present invention are：By the way that input picture is used to train grader, and according to the grader to more
A database images are classified, and can retrieve the database images similar to input picture, also, by determining and inputting
The similar similar area of predetermined reference zone in image, and the reference zone and similar area are used to training point
Class device, can effectively improve the classification accuracy of grader, so as to improve the accuracy of retrieval.In addition, when needs scheme input
As when being analyzed, by fast and accurately retrieving similar database images for reference, analysis efficiency can be effectively improved
And precision of analysis.
With reference to following explanation and attached drawing, only certain exemplary embodiments of this invention is disclose in detail, specifies the original of the present invention
Reason can be in a manner of adopted.It should be understood that embodiments of the present invention are not so limited in scope.In appended power
In the range of the spirit and terms that profit requires, embodiments of the present invention include many changes, modifications and are equal.
The feature for describing and/or showing for a kind of embodiment can be in a manner of same or similar one or more
Used in a other embodiment, it is combined with the feature in other embodiment, or substitute the feature in other embodiment.
It should be emphasized that term "comprises/comprising" refers to the presence of feature, one integral piece, step or component when being used herein, but simultaneously
It is not excluded for the presence or additional of one or more further features, one integral piece, step or component.
Brief description of the drawings
Included attached drawing is used for providing being further understood from the embodiment of the present invention, which constitutes one of specification
Point, come together explaination the principle of the present invention for illustrating embodiments of the present invention, and with word description.Under it should be evident that
Attached drawing in the description of face is only some embodiments of the present invention, for those of ordinary skill in the art, is not paying wound
On the premise of the property made is laborious, other attached drawings can also be obtained according to these attached drawings.In the accompanying drawings：
Fig. 1 is a schematic diagram of the image retrieving apparatus of the embodiment of the present invention 1；
Fig. 2 is a schematic diagram of the first determination unit 101 of the embodiment of the present invention 1；
Fig. 3 is a schematic diagram of the 3rd determination unit 105 of the embodiment of the present invention 1；
Fig. 4 is a schematic diagram of the matching degree computing unit 106 of the embodiment of the present invention 1；
Fig. 5 is a schematic diagram of the second determination unit 104 of the embodiment of the present invention 1；
Fig. 6 is a schematic diagram of the 4th determination unit 107 of the embodiment of the present invention 1；
Fig. 7 is a schematic diagram of the electronic equipment of the embodiment of the present invention 2；
Fig. 8 is the schematic block diagram that the system of the electronic equipment of the embodiment of the present invention 2 is formed；
Fig. 9 is a schematic diagram of the image search method of the embodiment of the present invention 3；
Figure 10 is another schematic diagram of the image search method of the embodiment of the present invention 3.
Embodiment
Referring to the drawings, will be apparent by following specification, foregoing and further feature of the invention.In specification
In attached drawing, only certain exemplary embodiments of this invention is specifically disclosed, which show wherein can be with the portion of principle using the present invention
Divide embodiment, it will thus be appreciated that the invention is not restricted to described embodiment, on the contrary, the present invention includes falling into appended power
Whole modification, modification and equivalents in the range of profit requirement.
Embodiment 1
The embodiment of the present invention provides a kind of image retrieving apparatus, and Fig. 1 is the one of the image retrieving apparatus of the embodiment of the present invention 1
Schematic diagram.As shown in Figure 1, detection device 100 includes：
First determination unit 101, it is used to determine in input picture and predetermined reference zone in the input picture
Similar similar area；
Training unit 102, it is used for according to reference zone and the similar area in the input picture, training classification
Device；
Taxon 103, it is used to classify to multiple database images using the grader that training obtains, obtained each
The classification results of a database images；
Second determination unit 104, it is used for the classification results according to each database images, determines the plurality of database diagram
The database images similar to the input picture as in.
From above-described embodiment, by the way that input picture is used to train grader, and according to the grader to more numbers
Classify according to storehouse image, the database images similar to input picture can be retrieved, also, pass through definite and input picture
In the similar similar area of predetermined reference zone, and the reference zone and similar area be used to train classification
Device, can effectively improve the classification accuracy of grader, so as to improve the accuracy of retrieval.In addition, working as needs to input picture
When being analyzed, by fast and accurately retrieving similar database images for reference, can effectively improve analysis efficiency and
Precision of analysis.
In the present embodiment, which can be any kind of image, for example, the input picture can be photography
Image, monitoring image and medical image etc..
In the present embodiment, in the input image with predetermined reference zone, for example, the reference zone can root
The region marked in advance according to actual needs.For example, the reference zone is user object region interested, alternatively,
The reference zone is typical object to be analyzed region.
In the present embodiment, the first determination unit 101 is used to determine similar to the reference zone similar in input picture
Region.Wherein, input picture can have a similar area, it is possible to have multiple similar areas.
The structure of the first determination unit 101 to the present embodiment and the method for definite similar area carry out exemplary below
Explanation.
Fig. 2 is a schematic diagram of the first determination unit 101 of the embodiment of the present invention 1.As shown in Fig. 2, the first determination unit
101 include：
First division unit 201, it is used to the input picture being divided into multiple first subregions；
Similarity calculated 202, it is used to calculate in the input picture each first sub-district beyond the reference zone
The similarity of domain and the reference zone, similarity is more than the region that the first subregion of first threshold forms, and to be determined as this similar
Region.
In the present embodiment, which is divided into multiple first subregions and can used by the first division unit 201
Existing method, it is for instance possible to use existing image partition method, irregular multiple sub-districts are divided into by the input picture
Domain, can also use Meshing Method, which is divided into multiple grid subregions of homalographic.
For example, input picture can be divided into the square net that size is 16 × 16 or 32 × 32, wherein, each side
Shape lattice are first subregion.
In the present embodiment, similarity calculated 202 calculates in the input picture each the beyond the reference zone
The similarity of one subregion and the reference zone, the region that similarity is more than to the first subregion composition of first threshold are determined as
The similar area.In the present embodiment, the similarity can be calculated using existing method.
For example, the feature in input picture in each first subregion can be extracted first, for example, local binary patterns
(Local binary pattern, LBP) feature, alternatively, feature based on Gabor filter etc..
Assuming that the feature of each first subregion is s in the reference zone extracted1,…,sn, n is sub-district in reference zone
The quantity in domain；The feature of each first subregion in the input picture extracted in addition to reference zone is t1,…,tm, m is
The quantity of first subregion in region in addition to reference zone.
It is, for example, possible to use following formula (1) calculate in region in addition to reference zone each first subregion with
The similarity of reference zone：
Wherein, sim (tj) represent the phases of j-th of first subregions and reference zone in region in addition to reference zone
Like degree, f (si,tj) represent j-th of first subregions in region in addition to reference zone and i-th first in reference zone
The similarity of feature in subregion, n represent the quantity of subregion in reference zone, and n is positive integer.
In the present embodiment, similarity calculated 202 in the region in addition to reference zone is calculated each
The similarity of one subregion and reference zone, the region that similarity is more than to the first subregion composition of first threshold are determined as this
Similar area, for example, the first threshold can be set according to actual needs.
In the present embodiment, obtained similar area can also be post-processed, be less than for example, removing isolated area
The similar area of predetermined threshold, fills the cavity etc. in similar area.It is artificial alternatively, it is also possible to be used by graphical interaction interface
Mode such as optimizes similar area at the post processing.
In the present embodiment, after similar area is determined, the reference area of training unit 102 in the input picture
Domain and the similar area, training grader.For example, using the reference zone in input picture and the similar area as positive sample
This, using the subregion of the remaining area beyond the reference zone in input picture and the similar area as negative sample, training
Grader.
In the present embodiment, by the way that the scope of positive sample to be extended to the similar area similar to reference zone, expand
The quantity of effective positive sample, so as to improve the classification accuracy of grader.
In the present embodiment, the negative sample roughly the same with the quantity of positive sample can be chosen, to train grader.
In the present embodiment, existing classifier type can be used, for example, support vector machines (Support Vector
Machine, SVM) grader, Bayes classifier etc..
In the present embodiment, taxon 103 is used to carry out multiple database images using the grader that training obtains
Classification, obtains the classification results of each database images.For example, each database images are separately input to trained classification
In device, using the output of grader as a result, being classification results of the score as the database images of the database images.
In this example, multiple database images can be obtained from the database pre-established, which can
Be according to it is various needs and establish.For example, for medical image, the data for being stored with a large amount of medical images can be established
Storehouse, wherein each medical image is the medical image for having clear and definite analytical conclusions.
In the present embodiment, by the second determination unit 104 for the classification results according to each database images, determining should
The database images similar to the input picture in multiple database images.For example, according to the descending order of classification results,
Corresponding database images are ranked up, the image of database is more forward, it is bigger with the similarity of input picture.
In the present embodiment, after being ranked up to multiple database images, it can choose and come above several
Database images are referred to.For example, when needing to analyze input picture, the higher data of these similarities are referred to
The existing analysis result of storehouse image.
In the present embodiment, which can also include：
3rd determination unit 105, it is used for the grader obtained using training, determines the candidate in each database images
Region；
Matching degree computing unit 106, it is used to calculate the candidate region and the input picture in each database images
The similar area spatial match degree.
At this time, the second determination unit 104 is determined according to the classification results and the spatial match degree of each database images
The database images similar to the input picture in the plurality of database images.
In this way, being combined spatial match degree with classification results to determine similar database images, can further carry
The accuracy of height retrieval.
In the present embodiment, the 3rd determination unit 105 and matching degree computing unit 106 are selectable unit (SU), in Fig. 1 with void
Wire frame representation.
Below for the 3rd determination unit 105, the structure of matching degree computing unit 106 and definite candidate region, calculating
The method of spatial match degree carries out exemplary explanation.
Fig. 3 is a schematic diagram of the 3rd determination unit 105 of the embodiment of the present invention 1.As shown in figure 3, the 3rd determination unit
105 include：
Second division unit 301, it is used to the database images being divided into multiple second subregions；
Score unit 302, it is used to score to each second subregion using the grader that training obtains, and will
The candidate region for dividing the region that the second subregion more than second threshold forms to be determined as in the database images.
In the present embodiment, the image segmentation side that the second division unit 301 can be identical with 201 use of the first division unit
Method, splits each database images respectively, obtains multiple second subregions of the database images.
In the present embodiment, the unit 302 that scores scores each second subregion using the grader that training obtains,
That is, each second subregion is input in the grader, it exports the score that result is second subregion.
In the present embodiment, the region for the second subregion composition that score is more than second threshold by the unit 302 that scores determines
For the candidate region in the database images, for example, the second threshold can be set according to actual needs.
In the present embodiment, each database images can have a candidate region, it is possible to have multiple candidate regions
Domain.
In the present embodiment, after the candidate region of each database images is determined, matching degree computing unit 106 is counted
Calculate similar area and the spatial match degree of the candidate region in each database images of the input picture.Fig. 4 is this hair
One schematic diagram of the matching degree computing unit 106 of bright embodiment 1.As shown in figure 4, matching degree computing unit 106 includes：
First computing unit 401, it is used to calculate each pixel and the data in the similar area of the input picture
The minimum range of the candidate region of storehouse image；
Second computing unit 402, each pixel that it is used to calculate in the candidate region of the database images are defeated with this
Enter the minimum range of the similar area of image；
3rd computing unit 403, it is used for the time for according to two minimum ranges being calculated, calculating the database images
Favored area and the spatial match degree of the similar area of the input picture.
In the present embodiment, matching degree computing unit 106 has multiple phases when calculating spatial match is spent in input picture
Like region, and each data graph image area in multiple similar areas can be chosen with the case of multiple candidate regions
The candidate region of area maximum is used to calculate spatial match degree in maximum similar area and multiple candidate regions, can also incite somebody to action
All candidate regions in all similar areas and database images in input picture are used to calculate spatial match degree.
In the present embodiment, the first computing unit 401 and the second computing unit 402 can be calculated using existing method this two
A minimum range, for example, the first computing unit 401 according to the coordinate of each pixel in the similar area of the input picture with
The coordinate of each boundary point in candidate region of the database images, calculates minimum range；Second computing unit 402 is according to the data
The coordinate of the coordinate of each pixel in the candidate region of storehouse image and each boundary point of similar area of the input picture, meter
Calculate minimum range.
In the present embodiment, the 3rd computing unit 403 calculates the database diagram according to two minimum ranges being calculated
The candidate region of picture and the spatial match degree of the similar area of the input picture.For example, it can be calculated according to following formula (2)
The spatial match degree：
Wherein, Dis represents the similar area of the input picture and the spatial match degree of the candidate region of the database images,
D(pi) represent input picture similar area in pixel piWith the minimum range of the candidate region of the database images, D
(qj) represent pixel q in the candidate regions of the database imagesjWith the minimum range of the similar area of the input picture, Np
Represent the quantity of the first subregion in the similar area, NqRepresent the quantity of the second subregion in the candidate region.
In the case where the device 100 has the 3rd determination unit 105 and matching degree computing unit 106, second determines list
Member 104 according to the classification results and spatial match degree of each database images, determine in the plurality of database images with the input
The similar database images of image.
Fig. 5 is a schematic diagram of the second determination unit 104 of the embodiment of the present invention 1.As shown in figure 5, the second determination unit
104 include：
Synthesis unit 501, it is used to be synthesized the classification results of each database images and spatial match degree；
Sequencing unit 502, it is used to be ranked up multiple database images according to the result of synthesis.
In the present embodiment, synthesis unit 501 can be in various manners by the classification results of each database images and sky
Between matching degree synthesized.
For example, can be calculated according to following formula (3) classification results of each database images and spatial match degree into
The composite result of row synthesis, that is, calculate the final score of the database images：
NonSim=Dis* (1-Score) (3)
Wherein, NonSim represents the final score of the database images, Dis represent the candidate regions of the database images with
The spatial match degree of the similar area of the input picture, Score represent the classification results of the database images.
In the present embodiment, the classification results Score of the database images can be by the way that database images be directly inputted
Obtained into grader or in the database images candidate region classification results, for example, can be according to following public affairs
Formula (4) calculates Score：
Wherein, score (qi) represent candidate region in i-th of second subregion qiScore, NqRepresent in candidate region
The second subregion quantity, 1≤i≤Nq。
For example, it is also possible to classification results and spatial match degree are synthesized using linear weighted function average method.
In the present embodiment, sequencing unit 502 is used to be ranked up multiple database images according to the result of synthesis.Example
Such as, for the composite result being calculated according to above-mentioned formula (3), i.e., the final score of each database images, final score
Lower, then the database images and the similarity of input picture are higher, which sorts more before examination, and vice versa.
In the present embodiment, it is required for being handled for some input pictures, not all region.In this feelings
Under condition, which can also include：4th determination unit 107, it is used to determine the target area in the input picture.
In the present embodiment, the 4th determination unit 107 is selectable unit (SU), is indicated by the dashed box in Fig. 1.
In the present embodiment, which refers to region to be treated in input picture.
Fig. 6 is a schematic diagram of the 4th determination unit 107 of the embodiment of the present invention 1.As shown in fig. 6, the 4th determination unit
107 include：
Acquiring unit 601, it is used for the foreground image for obtaining the input picture；
Removal unit 602, it is used to remove the part being connected with image border in the foreground image；
Post-processing unit 603, it is used to post-process the foreground image after removal, and the prospect that will pass through post processing
The region that image intermediate value is 1 is determined as the target area.
In the present embodiment, existing method can be used to obtain foreground image for acquiring unit 601, for example, passing through binaryzation
Method obtains foreground image.For example, the pixel value that pixel value in input picture is more than to the pixel of some threshold value is set to 0, make
For background pixel point, the pixel value that pixel value is less than to the pixel of the threshold value is set to 1, as foreground pixel point.
In the present embodiment, removal unit 602 is used to remove the part being connected with image border in the foreground image, so that
Remove the useless region in part.
In the present embodiment, post-processing unit 603 is used to post-process the foreground image after removal, and will be through later
The region that the foreground image intermediate value of processing is 1 is determined as the target area.For example, the post processing can include：Utilize morphology
Closed operation, the profile of the foreground area is carried out it is smooth, and by the noise in foreground area, i.e., included in foreground area
The background block of very little is filled.
In the present embodiment, when the device 100 has four determination units 107, the first determination unit 101 is in the input
In the range of the target area of image determine similar area, that is, determine the input picture target area in in the input picture
The similar similar area of predetermined reference zone, wherein, the first division unit 201 draws the target area of the input picture
It is divided into multiple first subregions；Correspondingly, the 3rd determination unit 105 is additionally operable to determine the target area in each database images
The target area of the database images is divided into multiple second subregions by domain, the second division unit 301.
From above-described embodiment, by the way that input picture is used to train grader, and according to the grader to more numbers
Classify according to storehouse image, the database images similar to input picture can be retrieved, also, pass through definite and input picture
In the similar similar area of predetermined reference zone, and the reference zone and similar area be used to train classification
Device, can effectively improve the classification accuracy of grader, so as to improve the accuracy of retrieval.In addition, working as needs to input picture
When being analyzed, by fast and accurately retrieving similar database images for reference, can effectively improve analysis efficiency and
Precision of analysis.
Embodiment 2
The embodiment of the present invention additionally provides a kind of electronic equipment, and Fig. 7 is a signal of the electronic equipment of the embodiment of the present invention 2
Figure.As shown in fig. 7, electronic equipment 700 includes image retrieving apparatus 701, the 26S Proteasome Structure and Function of image retrieving apparatus 701 and implementation
Record in example 1 is identical, and details are not described herein again.
Fig. 8 is the schematic block diagram that the system of the electronic equipment of the embodiment of the present invention 2 is formed.As shown in figure 8, electronic equipment
800 can include central processing unit 801 and memory 802；Memory 802 is coupled to central processing unit 801.The figure is exemplary
's；Other types of structure is can also use, to supplement or instead of the structure, to realize telecommunications functions or other functions.
As shown in figure 8, the electronic equipment 800 can also include：Input unit 803, display 804, power supply 805.
In one embodiment, the function of the image retrieving apparatus described in embodiment 1 can be integrated into central processing
In device 801.Wherein, central processing unit 801 can be configured as：Determine in input picture with being predefined in the input picture
The similar similar area of reference zone；The reference zone and the similar area in the input picture, instruction
Practice grader；The grader obtained using training classifies multiple database images, obtains point of each database images
Class result；According to the classification results of each database images, determine in the multiple database images with the input picture phase
As database images.
The similar area similar to predetermined reference zone in the input picture in the definite input picture, bag
Include：The input picture is divided into multiple first subregions；Calculate each beyond reference zone described in the input picture
The similarity, is more than the first subregion composition of first threshold by the similarity of a first subregion and the reference zone
Region is determined as the similar area.
Wherein, the reference zone in the input picture and the similar area, training grader,
Including：Using the reference zone in the input picture and the similar area as positive sample, by the input picture
The subregion of remaining area beyond the reference zone and the similar area is as negative sample, training grader.
Wherein, central processing unit 801 can be additionally configured to：The grader obtained using training, determines each database
Candidate region in image；Calculate the candidate region in each database images and the similar area of the input picture
The spatial match degree in domain, the classification results according to each database images, determine in the multiple database images with institute
The similar database images of input picture are stated, including：According to the classification results of each database images and the space
With degree, database images similar to the input picture in the multiple database images are determined.
Wherein, the grader obtained using training, determines the candidate region in each database images, including：Will
The database images are divided into multiple second subregions；Using the obtained grader of training to each second subregion into
Row scoring, and by score be more than second threshold the second subregion form region be determined as in the database images described in
Candidate region.
Wherein, the similar area for calculating the input picture and the candidate regions in each database images
The spatial match degree in domain, including：Calculate each pixel in the similar area of the input picture and the database
The minimum range of the candidate region of image；Calculate each pixel in the candidate region of the database images with
The minimum range of the similar area of the input picture；According to two minimum ranges being calculated, the data are calculated
The candidate region of storehouse image and the spatial match degree of the similar area of the input picture.
Wherein, the classification results according to each database images, determine in the multiple database images with it is described
The similar database images of input picture, including：By the classification results of each database images and the spatial match degree
Synthesized；The multiple database images are ranked up according to the result of synthesis.
Wherein, central processing unit 801 can be additionally configured to：Determine the target area in the input picture；It is described true
Target area in the fixed input picture, including：Obtain the foreground image of the input picture；Remove in the foreground image
The part being connected with image border；Foreground image after removal is post-processed, and by the foreground image by post processing
The region being worth for 1 is determined as the target area.
Electronic equipment 800 is also not necessary to include all components shown in Fig. 8 in the present embodiment.
As shown in figure 8, central processing unit 801 be otherwise referred to as controller or operational controls, can include microprocessor or
Other processor devices and/or logic device, central processing unit 801 receive all parts of input and control electronics 800
Operation.
Memory 802, such as can be buffer, flash memory, hard disk driver, removable medium, volatile memory, non-volatile
One or more in memory or other appropriate devices.And central processing unit 801 can perform the memory 802 storage
The program, to realize information storage or processing etc..The function of other components is with existing similar, and details are not described herein again.Electronic equipment
800 each component can be realized by specialized hardware, firmware, software or its combination, be made without departing from the scope of the present invention.
From above-described embodiment, by the way that input picture is used to train grader, and according to the grader to more numbers
Classify according to storehouse image, the database images similar to input picture can be retrieved, also, pass through definite and input picture
In the similar similar area of predetermined reference zone, and the reference zone and similar area be used to train classification
Device, can effectively improve the classification accuracy of grader, so as to improve the accuracy of retrieval.In addition, working as needs to input picture
When being analyzed, by fast and accurately retrieving similar database images for reference, can effectively improve analysis efficiency and
Precision of analysis.
Embodiment 3
The embodiment of the present invention also provides a kind of image search method, it corresponds to the image retrieving apparatus of embodiment 1.Fig. 9
It is a schematic diagram of the image search method of the embodiment of the present invention 3.As shown in figure 9, this method includes：
Step 901：Determine similar area similar to predetermined reference zone in the input picture in input picture；
Step 902：Reference zone and similar area in the input picture, training grader；
Step 903：The grader obtained using training classifies multiple database images, obtains each database diagram
The classification results of picture；
Step 904：According to the classification results of each database images, determine in multiple database images with the input picture
Similar database images.
Figure 10 is another schematic diagram of the image search method of the embodiment of the present invention 3.As shown in Figure 10, this method includes：
Step 1001：Determine similar area similar to predetermined reference zone in the input picture in input picture
Domain；
Step 1002：Reference zone and similar area in the input picture, training grader；
Step 1003：The grader obtained using training classifies multiple database images, obtains each database
The classification results of image；
Step 1004：The grader obtained using training, determines the candidate region in each database images；
Step 1005：Calculate the candidate region in each database images and the space of the similar area of the input picture
With degree；
Step 1006：According to the classification results of each database images and spatial match degree, multiple database images are determined
In database images similar to the input picture.
In the present embodiment, determine similar area method, training grader method, to multiple database images carry out
The method of classification, the method for determining candidate region, calculate spatial match degree method and determine multiple database images in
The method of the similar database images of the input picture is identical with the record in embodiment 1, is not repeated herein.
From above-described embodiment, by the way that input picture is used to train grader, and according to the grader to more numbers
Classify according to storehouse image, the database images similar to input picture can be retrieved, also, pass through definite and input picture
In the similar similar area of predetermined reference zone, and the reference zone and similar area be used to train classification
Device, can effectively improve the classification accuracy of grader, so as to improve the accuracy of retrieval.In addition, working as needs to input picture
When being analyzed, by fast and accurately retrieving similar database images for reference, can effectively improve analysis efficiency and
Precision of analysis.
The embodiment of the present invention also provides a kind of computer-readable program, wherein when in image retrieving apparatus or electronic equipment
When performing described program, described program causes computer to perform the institute of embodiment 3 in described image retrieves device or electronic equipment
The image search method stated.
The embodiment of the present invention also provides a kind of storage medium for being stored with computer-readable program, wherein the computer can
Reader causes computer to perform the image search method described in embodiment 3 in image retrieving apparatus or electronic equipment.
With reference to what the embodiment of the present invention described image search method is performed in described image retrieves device or electronic equipment
Hardware, the software module performed by processor or the two combination can be embodied directly in.For example, in functional block diagram shown in Fig. 1
One or more and/or functional block diagram one or more combinations, both can correspond to each soft of computer program flow
Part module, can also correspond to each hardware module.These software modules, can correspond respectively to each shown in Fig. 9 or Figure 10
A step.These software modules are for example cured and realized by these hardware modules using field programmable gate array (FPGA).
Software module can be located at RAM memory, flash memory, ROM memory, eprom memory, eeprom memory, post
Storage, hard disk, mobile disk, the storage medium of CD-ROM or any other form known in the art.One kind can be deposited
Storage media is coupled to processor, so as to enable a processor to from the read information, and can be write to the storage medium
Information；Or the storage medium can be the part of processor.Pocessor and storage media can be located in ASIC.This is soft
Part module can store in a memory in the mobile terminal, can also be stored in the storage card of pluggable mobile terminal.For example,
If equipment (such as mobile terminal) is using the MEGA-SIM cards of larger capacity or the flash memory device of large capacity, the software
Module is storable in the flash memory device of the MEGA-SIM cards or large capacity.
The one or more of one or more of functional block diagram and/or functional block diagram for Fig. 1 descriptions combine, can be with
It is embodied as the general processor, digital signal processor (DSP), application-specific integrated circuit for performing function described herein
(ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete
Nextport hardware component NextPort or it is any appropriately combined.One or more of functional block diagram for Fig. 1 descriptions and/or functional block diagram
One or more combinations, the combination of computing device is also implemented as, for example, the combining of DSP and microprocessor, multiple micro- places
Reason device, communicate the one or more microprocessors combined or any other this configuration with DSP.
Above in association with specific embodiment, invention has been described, it will be appreciated by those skilled in the art that this
A little descriptions are all exemplary, and are not limiting the scope of the invention.Those skilled in the art can be according to the present invention
Spirit and principle various variants and modifications are made to the present invention, these variants and modifications are also within the scope of the invention.
On the embodiment including above example, following note is also disclosed：
Note 1, a kind of image retrieving apparatus, described device include：
First determination unit, it is used to determine in input picture and predetermined reference zone phase in the input picture
As similar area；
Training unit, it is used for the reference zone and the similar area in the input picture, training
Grader；
Taxon, it is used to classify to multiple database images using the grader that training obtains, obtained each
The classification results of database images；
Second determination unit, it is used for the classification results according to each database images, determines the multiple database diagram
The database images similar to the input picture as in.
Note 2, the device according to note 1, wherein, first determination unit includes：
First division unit, it is used to the input picture being divided into multiple first subregions；
Similarity calculated, it is used to calculate each first sub-district beyond reference zone described in the input picture
Domain and the similarity of the reference zone, the region that the similarity is more than to the first subregion composition of first threshold are determined as
The similar area.
Note 3, the device according to note 1, wherein, the training unit is used for described in the input picture
Reference zone and the similar area as positive sample, by reference zone described in the input picture and the similar area with
The subregion of outer remaining area is as negative sample, training grader.
Note 4, the device according to note 1, wherein, described device further includes：
3rd determination unit, it is used for the grader obtained using training, determines the candidate regions in each database images
Domain；
Matching degree computing unit, it is used to calculate the candidate region and the input picture in each database images
The similar area spatial match degree,
Second determination unit is used for the classification results and the spatial match degree according to each database images,
Determine database images similar to the input picture in the multiple database images.
Note 5, the device according to note 4, wherein, the 3rd determination unit includes：
Second division unit, it is used to the database images being divided into multiple second subregions；
Score unit, it is used to score to each second subregion using the grader that training obtains, and will
The region that the second subregion that score is more than second threshold forms is determined as the candidate region in the database images.
Note 6, the device according to note 4, wherein, the matching degree computing unit includes：
First computing unit, its be used to calculate each pixel in the similar area of the input picture with it is described
The minimum range of the candidate region of database images；
Second computing unit, it is used to calculate each pixel and institute in the candidate region of the database images
State the minimum range of the similar area of input picture；
3rd computing unit, it is used for the institute for according to two minimum ranges being calculated, calculating the database images
State candidate region and the spatial match degree of the similar area of the input picture.
Note 7, the device according to note 4, wherein, second determination unit includes：
Synthesis unit, it is used to be closed the classification results of each database images and the spatial match degree
Into；
Sequencing unit, it is used to be ranked up the multiple database images according to the result of synthesis.
Note 8, the device according to note 1, wherein, described device further includes：
4th determination unit, it is used to determine the target area in the input picture；
4th determination unit includes：
Acquiring unit, it is used for the foreground image for obtaining the input picture；
Removal unit, it is used to remove the part being connected with image border in the foreground image；
Post-processing unit, it is used to post-process the foreground image after removal, and the foreground picture that will pass through post processing
As the region that intermediate value is 1 is determined as the target area.
Note 9, a kind of electronic equipment, including the device according to any one of note 1-8.
Note 10, a kind of image search method, the described method includes：
Determine similar area similar to predetermined reference zone in the input picture in input picture；
The reference zone and the similar area in the input picture, training grader；
The grader obtained using training classifies multiple database images, obtains the classification of each database images
As a result；
According to the classification results of each database images, determine in the multiple database images with the input picture phase
As database images.
Note 11, the method according to note 10, wherein, in the definite input picture with it is pre- in the input picture
The first similar similar area of definite reference zone, including：
The input picture is divided into multiple first subregions；
Calculate the phase of each first subregion and the reference zone described in the input picture beyond reference zone
Like degree, the region that the similarity is more than to the first subregion composition of first threshold is determined as the similar area.
Note 12, the method according to note 10, wherein, the reference zone in the input picture
And the similar area, training grader, including：
Using the reference zone in the input picture and the similar area as positive sample, by the input picture
Described in remaining area beyond reference zone and the similar area subregion as negative sample, training grader.
Note 13, the method according to note 10, wherein, the method further includes：
The grader obtained using training, determines the candidate region in each database images；
Calculate the candidate region in each database images and the space of the similar area of the input picture
Matching degree,
The classification results according to each database images, determine in the multiple database images with the input figure
As similar database images, including：
According to the classification results of each database images and the spatial match degree, the multiple database diagram is determined
The database images similar to the input picture as in.
Note 14, the method according to note 13, wherein, the grader obtained using training, determines each number
According to the candidate region in the image of storehouse, including：
The database images are divided into multiple second subregions；
The grader obtained using training scores each second subregion, and score is more than second threshold
The region of the second subregion composition be determined as the candidate region in the database images.
Note 15, the method according to note 13, wherein, the similar area for calculating the input picture with
The spatial match degree of the candidate region in each database images, including：
Calculate the time of each pixel and the database images in the similar area of the input picture
The minimum range of favored area；
Calculate the phase of each pixel and the input picture in the candidate region of the database images
Like the minimum range in region；
According to two minimum ranges being calculated, the candidate region of the database images and the input are calculated
The spatial match degree of the similar area of image.
Note 16, the method according to note 13, wherein, the classification results according to each database images, really
The database images similar to the input picture in fixed the multiple database images, including：
The classification results of each database images and the spatial match degree are synthesized；
The multiple database images are ranked up according to the result of synthesis.
Note 17, the method according to note 10, wherein, the method further includes：
Determine the target area in the input picture；
The target area determined in the input picture, including：
Obtain the foreground image of the input picture；
Remove the part being connected in the foreground image with image border；
Foreground image after removal is post-processed, and the region that the foreground image intermediate value by post processing is 1 is true
It is set to the target area.
Claims (10)
1. a kind of image retrieving apparatus, described device include：
First determination unit, it is used to determine in input picture similar to predetermined reference zone in the input picture
Similar area；
Training unit, it is used for the reference zone and the similar area in the input picture, training classification
Device；
Taxon, it is used to classify to multiple database images using the grader that training obtains, and obtains each data
The classification results of storehouse image；
Second determination unit, it is used for the classification results according to each database images, determines in the multiple database images
The database images similar to the input picture.
2. device according to claim 1, wherein, first determination unit includes：
First division unit, it is used to the input picture being divided into multiple first subregions；
Similarity calculated, its be used to calculate each first subregion described in the input picture beyond reference zone with
The similarity of the reference zone, by the similarity be more than the region that the first subregion of first threshold forms be determined as it is described
Similar area.
3. device according to claim 1, wherein, the training unit is used for the reference in the input picture
Region and the similar area are as positive sample, beyond reference zone described in the input picture and the similar area
The subregion of remaining area is as negative sample, training grader.
4. device according to claim 1, wherein, described device further includes：
3rd determination unit, it is used for the grader obtained using training, determines the candidate region in each database images；
Matching degree computing unit, it is used to calculate the candidate region in each database images and the institute of the input picture
The spatial match degree of similar area is stated,
Second determination unit is used for the classification results and the spatial match degree according to each database images, determines
The database images similar to the input picture in the multiple database images.
5. device according to claim 4, wherein, the 3rd determination unit includes：
Second division unit, it is used to the database images being divided into multiple second subregions；
Score unit, it is used to score to each second subregion using the obtained described grader of training, and by score
The region formed more than the second subregion of second threshold is determined as the candidate region in the database images.
6. device according to claim 4, wherein, the matching degree computing unit includes：
First computing unit, it is used to calculate each pixel and the data in the similar area of the input picture
The minimum range of the candidate region of storehouse image；
Second computing unit, its be used to calculate each pixel in the candidate region of the database images with it is described defeated
Enter the minimum range of the similar area of image；
3rd computing unit, it is used for the time for according to two minimum ranges being calculated, calculating the database images
Favored area and the spatial match degree of the similar area of the input picture.
7. device according to claim 4, wherein, second determination unit includes：
Synthesis unit, it is used to be synthesized the classification results of each database images and the spatial match degree；
Sequencing unit, it is used to be ranked up the multiple database images according to the result of synthesis.
8. device according to claim 1, wherein, described device further includes：
4th determination unit, it is used to determine the target area in the input picture；
4th determination unit includes：
Acquiring unit, it is used for the foreground image for obtaining the input picture；
Removal unit, it is used to remove the part being connected with image border in the foreground image；
Post-processing unit, it is used to post-process the foreground image after removal, and by the foreground image by post processing
The region being worth for 1 is determined as the target area.
9. a kind of electronic equipment, including the device according to any one of claim 1-8.
10. a kind of image search method, the described method includes：
Determine similar area similar to predetermined reference zone in the input picture in input picture；
The reference zone and the similar area in the input picture, training grader；
The grader obtained using training classifies multiple database images, obtains the classification knot of each database images
Fruit；
According to the classification results of each database images, determine similar to the input picture in the multiple database images
Database images.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610886031.0A CN107918624B (en) | 2016-10-11 | 2016-10-11 | Image retrieval device and method and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610886031.0A CN107918624B (en) | 2016-10-11 | 2016-10-11 | Image retrieval device and method and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107918624A true CN107918624A (en) | 2018-04-17 |
| CN107918624B CN107918624B (en) | 2022-03-15 |
Family
ID=61891827
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610886031.0A Active CN107918624B (en) | 2016-10-11 | 2016-10-11 | Image retrieval device and method and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107918624B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003109009A (en) * | 2001-09-26 | 2003-04-11 | Communication Research Laboratory | Method and instrument for measuring similarity of image |
| CN101021870A (en) * | 2007-03-20 | 2007-08-22 | 北京中星微电子有限公司 | Picture inquiry method and system |
| CN101539930A (en) * | 2009-04-21 | 2009-09-23 | 武汉大学 | Search method of related feedback images |
| CN102542014A (en) * | 2011-12-16 | 2012-07-04 | 华中科技大学 | Image searching feedback method based on contents |
| CN102831434A (en) * | 2012-07-03 | 2012-12-19 | 天津师范大学 | Handwritten Chinese character recognizing method based on identification normalization |
| US20140294272A1 (en) * | 2013-03-29 | 2014-10-02 | Case Western Reserve University | Discriminatively Weighted Multi-Scale Local Binary Patterns |
| CN104217225A (en) * | 2014-09-02 | 2014-12-17 | 中国科学院自动化研究所 | A visual target detection and labeling method |
-
2016
- 2016-10-11 CN CN201610886031.0A patent/CN107918624B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003109009A (en) * | 2001-09-26 | 2003-04-11 | Communication Research Laboratory | Method and instrument for measuring similarity of image |
| CN101021870A (en) * | 2007-03-20 | 2007-08-22 | 北京中星微电子有限公司 | Picture inquiry method and system |
| CN101539930A (en) * | 2009-04-21 | 2009-09-23 | 武汉大学 | Search method of related feedback images |
| CN102542014A (en) * | 2011-12-16 | 2012-07-04 | 华中科技大学 | Image searching feedback method based on contents |
| CN102831434A (en) * | 2012-07-03 | 2012-12-19 | 天津师范大学 | Handwritten Chinese character recognizing method based on identification normalization |
| US20140294272A1 (en) * | 2013-03-29 | 2014-10-02 | Case Western Reserve University | Discriminatively Weighted Multi-Scale Local Binary Patterns |
| CN104217225A (en) * | 2014-09-02 | 2014-12-17 | 中国科学院自动化研究所 | A visual target detection and labeling method |
Non-Patent Citations (2)
| Title |
|---|
| REZA FAKOURI ET AL: "Region-Based Image Clustering and Retrieval Using Fuzzy Similarity and Relevance Feedback", 《2008 INTERNATIONAL CONFERENCE ON COMPUTER AND ELECTRICAL ENGINEERING》 * |
| 赵明: "基于机器学习的图像检索方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107918624B (en) | 2022-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5134628B2 (en) | Media material analysis of consecutive articles | |
| CN103678504B (en) | Similarity-based breast image matching image searching method and system | |
| CN110175613A (en) | Street view image semantic segmentation method based on Analysis On Multi-scale Features and codec models | |
| EP2570970A1 (en) | Method and system for the automatic analysis of an image of a biological sample | |
| CN105261109A (en) | Identification method of prefix letter of banknote | |
| CN101065762A (en) | Automated radiograph classification using anatomy information | |
| Chen et al. | Local feature based mammographic tissue pattern modelling and breast density classification | |
| CN109800781A (en) | A kind of image processing method, device and computer readable storage medium | |
| CN110222582B (en) | Image processing method and camera | |
| CN103679195A (en) | Method and system for classifying texture images on basis of local edge pattern | |
| Russakovsky et al. | A steiner tree approach to efficient object detection | |
| CN113689436A (en) | Image semantic segmentation method, device, equipment and storage medium | |
| Taschwer et al. | Automatic separation of compound figures in scientific articles | |
| KR101981284B1 (en) | Apparatus Processing Image and Method thereof | |
| CN115424053A (en) | Small sample image identification method, device and equipment and storage medium | |
| Rampun et al. | Breast density classification using local ternary patterns in mammograms | |
| RU2732895C1 (en) | Method for isolating and classifying blood cell types using deep convolution neural networks | |
| Stojnić et al. | Detection of pollen bearing honey bees in hive entrance images | |
| Rampun et al. | Breast density classification using multiresolution local quinary patterns in mammograms | |
| KR20100108778A (en) | Image information classification method and apparatus | |
| Xu et al. | Robust seed localization and growing with deep convolutional features for scene text detection | |
| Jeyavathana et al. | An efficient feature extraction method for tuberculosis detection using chest radiographs | |
| Chenni et al. | Patch clustering for representation of histopathology images | |
| CN115984588A (en) | Image background similarity analysis method and device, electronic equipment and storage medium | |
| CN105930515A (en) | Image retrieval system and method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |