CN108959338A - A kind of image search method and system - Google Patents
A kind of image search method and system Download PDFInfo
- Publication number
- CN108959338A CN108959338A CN201810275778.1A CN201810275778A CN108959338A CN 108959338 A CN108959338 A CN 108959338A CN 201810275778 A CN201810275778 A CN 201810275778A CN 108959338 A CN108959338 A CN 108959338A
- Authority
- CN
- China
- Prior art keywords
- image
- roi
- feature vector
- color
- library
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Abstract
The present invention relates to a kind of image search method and systems.This method comprises: receiving operational order, and region of interest ROI is selected on target image according to operational order;Extract the characteristics of image of ROI;The characteristics of image of ROI is matched with the characteristics of image of image in pre-established image library, obtains at least two candidate images.Pass through the above method, designer is in design work, oneself desired image product can be quickly found out, avoid the drawbacks of scanning in face of large nuber of images rule of thumb or memory, the shortcomings that also avoiding traditional approach (such as by artificial typing field) search bring inaccuracy and omitting, the workload for greatly alleviating designer, has liberated the thinking of designer, and design is allowed to return itself.
Description
Technical field
The present invention relates to data processing field more particularly to a kind of image search method and systems.
Background technique
Furnishing fields indoors, wallpaper are also known as wallpaper, are a kind of quite extensive house decorative materials of application, are chiefly used in
Mounting metope recklessly.Wallpaper has abundant color multiplicity, pattern, safety and environmental protection, easy for construction, the strong, affordable of tactile sensation etc. a variety of
Feature.Interior decorator chooses wallpaper in design, according to different decorated styles.However, facing many kinds of, Cheng Qianshang
The wallpaper of ten thousand different colours and texture, designer by virtue of experience can only inquire inventory with memory, great effort spent in and is selected
In lookup.It is time-consuming and laborious.
Although there is also certain search methods.Traditional wallpaper search method is exactly the people in wall paper products data input
Work fills in field information, and then designer is when in use by the understanding to effect picture, according to all wall paper products of field, still
This method usually to there is clearly defined information effect relatively good, such as: price, manufacturer, material.And relatively feel for some
Property the fuzzyyer information of definition such as feature, such as color, decorative pattern, the understanding of different people and record be it is different, this is to wallpaper
Retrieval produces very big difficulty.
Summary of the invention
In order to solve the above technical problems, the present invention provides a kind of image search method and systems.
In a first aspect, the present invention provides one kind: receiving operational order, and according to the operational order on target image
It selects area-of-interest (region of interest, abbreviation ROI);
Extract the characteristics of image of ROI;
The characteristics of image of ROI is matched with the characteristics of image of image in pre-established image library, obtains at least two
Candidate image.
The beneficial effects of the present invention are: designer or user are in the case where known effect figure, it is specified to choose ROI or pass through
Algorithm adaptivenon-uniform sampling obtains target ROI.ROI image is obtained, feature extraction is carried out to ROI image;Then by the image of extraction
Feature is matched with the characteristics of image of image in pre-established image library, finds out immediate, the image conduct of specified quantity
Output is as a result, this method can be applied to wallpaper, floor or ceramic tile etc. arbitrarily convenient in the scene of image retrieval.By above-mentioned
Method, designer can be quickly found out oneself desired image product in design work, avoid in face of large nuber of images only according to
It is inaccurate to also avoid traditional approach (such as by artificial typing field) search bring for the drawbacks of experience or memory scan for
The shortcomings that true property and omission, the workload of designer is greatly alleviated, the thinking of designer has been liberated, design is allowed to return itself.
Second aspect, the present invention provides a kind of image indexing system, which includes:
Receiving unit selects area-of-interest for receiving operational order, and according to operational order on target image
ROI;
Extraction unit, for extracting the characteristics of image of ROI;
Matching unit, for the characteristics of image of ROI to be matched with the characteristics of image of image in pre-established image library,
Obtain at least two candidate images
The beneficial effects of the present invention are: designer or user are in the case where known effect figure, it is specified to choose ROI or pass through
Algorithm adaptivenon-uniform sampling obtains target ROI.ROI image is obtained, feature extraction is carried out to ROI image;Then by the image of extraction
Feature is matched with the characteristics of image of image in pre-established image library, finds out immediate, the image conduct of specified quantity
Output is as a result, the system can be applied to wallpaper, floor or ceramic tile etc. arbitrarily convenient in the scene of image retrieval.By above-mentioned
System, designer can be quickly found out oneself desired image product in design work, avoid in face of large nuber of images only according to
It is inaccurate to also avoid traditional approach (such as by artificial typing field) search bring for the drawbacks of experience or memory scan for
The shortcomings that true property and omission, the workload of designer is greatly alleviated, the thinking of designer has been liberated, design is allowed to return itself.
Detailed description of the invention
Fig. 1 is a kind of image search method flow diagram provided in an embodiment of the present invention;
Fig. 2 is a kind of image indexing system structural schematic diagram provided in an embodiment of the present invention.
Specific embodiment
In being described below, for illustration and not for limitation, propose such as specific system structure, interface, technology it
The detail of class understands the present invention to cut thoroughly.However, it will be clear to one skilled in the art that there is no these specific
The present invention also may be implemented in the other embodiments of details.In other situations, omit to well-known system, circuit and
The detailed description of method, in case unnecessary details interferes description of the invention.
Fig. 1 is a kind of image search method flow diagram provided in an embodiment of the present invention.Before introducing the above method,
Firstly the need of establishing an image library.When establishing image library, only for establishing image library relevant to wallpaper, other images
Library establishment process is similar, and which is not described herein again.
It specifically includes: all wallpapers in traversal wallpaper library.It is then based on the characteristic vector pickup and texture of color space
The characteristic vector pickup in space, obtains the feature vector of the corresponding color space of each wallpaper respectively, and is stored in color characteristic number
According to library, and the feature vector of the corresponding texture space of each wallpaper is obtained, is stored in textural characteristics database.And color characteristic number
The same database can be belonged to simultaneously according to library and textural characteristics database, but be stored in different storage locations.For narration
Convenient, above-mentioned database is collectively termed as image library.
When extracting the color feature vector of each wallpaper respectively, specifically comprise the following steps:
Step 1, (reader is also understood that from wallpaper image library, to select a wallpaper, wallpaper figure input RGB image
As being RGB image), RGB image is converted into hsv color spatial image using following formula, wherein RGB respectively indicates red green
Color component on blue three channels.
Step 2, non-uniform quantizing is carried out to hsv color space.
For example, being respectively 8,12 and 4 etc. three ranks to H, S and V space quantization using mapping table set in advance.To wall
(new/old) HSV value table of each pixel in paper image is searched, and new HSV value is obtained.
It is specific as shown in table 1:
New H value | Original H value | New S value | Original S value | New V value | Original V value |
0 | 0-45 | 0 | 0-0.083 | 0 | 0-0.25 |
1 | 46-90 | 1 | 0.084-0.166 | 1 | 0.26-0.50 |
2 | 91-135 | 2 | 0.167-0.249 | 2 | 0.51-0.75 |
3 | 136-180 | 3 | 0.250-0.332 | 3 | 0.76-1.0 |
4 | 181-225 | 4 | 0.333-0.415 | ||
5 | 226-270 | 5 | 0.416-0.498 | ||
6 | 271-315 | 6 | 0.499-0.581 | ||
7 | 316-360 | 7 | 0.582-0.664 | ||
8 | 0.665-0.747 | ||||
9 | 0.748-0.830 | ||||
10 | 0.831-0.913 | ||||
11 | 0.914-1.0 |
Table 1
Step 3, after carrying out non-homogeneous amount to hsv color space, hsv color space shares 8 × 12 × 4, i.e., 384 kinds take
Value.It needs to be 384 dimension color histogram vectors, the i.e. vector that every kind of wallpaper can be tieed up with one 384 by 3 dimension color space conversions
It indicates.
Step 4, the unique number of every kind of wallpaper and corresponding color feature vector are established into mapping relations, and stored
Into image library.
And the feature vector of the corresponding texture space of each wallpaper is obtained, specifically comprise the following steps:
Step 1, (reader is also understood that from wallpaper image library, to select a wallpaper, wallpaper figure input RGB image
As being RGB image), RGB image is converted into gray level image.
Can specifically following formula be used:
Gray=0.299R+0.587G+0.114B (formula 2)
Step 2, feature point extraction is carried out to gray level image, obtains N number of characteristic point.
Specifically, the extraction of characteristic point can be carried out to gray level image using SIFT algorithm, N number of characteristic point is found, and it is every
The feature vector that a characteristic point can be tieed up with one 128 indicates that wherein N is the positive integer more than or equal to 2.
Step 3, it can be obtained after handling all wallpapers in wallpaper library using the method for step 1 and step 2
N number of characteristic point corresponding to each wallpaper image and the corresponding feature vector of N number of characteristic point is taken to indicate.
The mapping relations between each characteristic point and feature vector are established respectively.
Step 4, the cluster centre in the corresponding characteristic point of all wallpapers where each characteristic point is predicted respectively, and is obtained
Residual error between each characteristic point and cluster centre.
Specifically, using Kmeans clustering algorithm, predicting each wallpaper according to the corresponding feature vector of each characteristic point
Cluster centre where corresponding key point, wherein the quantity of cluster centre cluster be K (such as: K can choose 192, theoretically K
Value more large texture space is more detailed), K is the positive integer more than or equal to 1.And find out each key point and cluster centre it
Between residual error.
Step 5, obtain respectively belong in each wallpaper same cluster centre all characteristic points and the cluster centre it
Between residual error and value.
All characteristic points for belonging to the same cluster centre in each wallpaper are counted, all characteristic point difference are then calculated
Residual error between the cluster centre and value.
After the residual error of all K cluster centres is connected, the dimensional vector of K × 128 is obtained.
Step 6, to being normalized with value and dimension-reduction treatment, the feature vector of each wallpaper is obtained.
Specifically, being exactly the feature vector of each wallpaper after carrying out the normalization of L2 norm to the feature vector that K × 128 is tieed up.
But when establishing textural feature space index to each wallpaper, since the feature vector dimension of wallpaper is excessively high, Billy is fast in data
Speed retrieval and storage etc., so in order to achieve the purpose that reduce memory space and real-time retrieval, it usually needs to features described above to
Amount does dimension-reduction treatment.It can make the reduction of the dimensional vector of K × 128 in the case where not reducing precision as far as possible for example, by using PCA method
To 128 dimensional vectors, the final texture feature vector of wallpaper is obtained.
Step 7, the unique encodings of each wallpaper and texture feature vector are established and maps and stores into image library.
More than, then the detailed process to establish image library.
Certainly, optionally, establish in the detailed process of image library, it can also be by the color feature vector finally obtained and line
Reason feature vector is connected, the feature vector after obtaining series connection, then by the spy after the unique number and series connection of every kind of wallpaper
Sign vector establishes mapping relations, stores into image library.
Both the above mode can choose in any one or two kinds of while storage image library.And subsequent user is then according to
The image library being set up selects final search method.
Certainly, establishing image library is only to execute primary preparation.In the following, in the above-mentioned image library established in advance
On the basis of, it will be described in detail image search method and system.
As shown in Figure 1, this method comprises:
Step 110, operational order is received, and region of interest ROI is selected on target image according to operational order.
Specifically, user or designer can specify on target image in the case where known effect figure and choose ROI
Or target ROI is obtained by inputting algorithm adaptivenon-uniform sampling, and system can then receive the operational order of user's input, be referred to according to operation
Order selects ROI on target image.
Step 120, the characteristics of image of ROI is extracted.
Optionally, the characteristics of image of ROI may include color characteristic and textural characteristics.Therefore, it needs to mention respectively when extraction
Take the color feature vector and texture feature vector of ROI.
It may include: the color feature vector that ROI is extracted according to preset color feature extracted algorithm when extraction;
And the texture feature vector of ROI is extracted according to preset Texture Segmentation Algorithm.
It is further alternative, it is specific to wrap when extracting the color feature vector of ROI according to preset color feature extracted algorithm
Include step 1201~step 1204.
Step 1201, hsv color spatial image is converted to the RGB image of ROI.
Step 1202, non-uniform quantizing and to hsv color spatial image is carried out, and is assigned every in hsv color spatial image
The new HSV value of one pixel.
Step 1203, respectively by the new HSV value of each pixel in hsv color spatial image be converted to color characteristic to
Amount.
Step 1204, according to the color feature vector of each pixel in hsv color spatial image, the color of ROI is obtained
Feature vector.
It is further alternative, the texture feature vector of ROI is extracted according to preset Texture Segmentation Algorithm, it specifically can be with
Include:
Step 1205, the image of ROI is converted into gray level image.
Step 1206, feature point extraction is carried out to gray level image, obtains at least two characteristic points.
Step 1207, it predicts the cluster centre at least two characteristic points where each characteristic point, and obtains each
Residual error between characteristic point and cluster centre.
Step 1208, obtain belong to residual error between all characteristic points of same cluster centre and the cluster centre and
Value.
Step 1209, and to being normalized with value and dimension-reduction treatment, the texture feature vector of ROI is obtained.
In fact, specific method step performed by step 1201~step 1209, hereinbefore establishes image
Corresponding introduction is done when library, the difference only dealt with objects.In step 1201~step 1209, handled object is
The image of ROI.Therefore, which is not described herein again for specific implementation procedure.
Step 130, the characteristics of image of ROI is matched with the characteristics of image of image in pre-established image library, is obtained
At least two candidate images.
Optionally, step 130 may include following two steps:
Step 1301, the color feature vector of image in the color feature vector of ROI and pre-established image library is carried out
Matching, obtains at least one candidate image.
Step 1302, the texture feature vector of image in the texture feature vector of ROI and pre-established image library is carried out
Matching, obtains at least one candidate image.
No matter can be used European in carrying out color feature vector matching either texture feature vector matching process
Furthest Neighbor realization, specific formula is as follows:
Wherein, dis is the difference between different wallpapers, and fq, i are target image ROI color (texture) feature vectors, and fd, i are
Color (texture) feature vector of a certain wallpaper in image library.
The difference between two vectors can also be calculated with COS distance formula, specific formula is as follows:
Wherein, n is the number of wallpaper in image library, AiFor target image ROI color (texture) feature vector, BiFor image
Color (texture) feature vector of a certain wallpaper in library, n are the positive integer more than or equal to 2.
Further alternative, after step 130, this method can also include step 140:
According to preset weight proportion, at least one the candidate's figure obtained after being matched according to color feature vector is shown
The number of picture, and obtained after being matched according to texture feature vector only a candidate image number.
That is, available m candidate image after being matched by color feature vector, and matched by texture feature vector
Afterwards, available j candidate image.So, it when these candidate images being showed user, can carry out according to a certain percentage
Displaying, m and j are the positive integer more than or equal to 1.
It for example, the weight ratio of m:j can be preset, for example is 7:3.Certainly, the weight ratio of this setting
Value can pass through the sets itselfs such as personal experience or different application scenarios for designer.
Further, then on the basis of the above method, the fields such as the price, brand and material of product can also be subjected to band
There is the retrieval of alternative condition to recommend.
Further alternative, as another alternative scheme, step 130 can also be accomplished in that
Step a connects the color feature vector of ROI and texture feature vector, the feature vector after obtaining series connection;
Step b matches the feature vector after series connection with the feature vector of connecting of image in pre-established image library,
Obtain at least two candidate images.
Certainly, the realization of step a and step b are that the face of each wallpaper in all wallpapers is implicitly present in image library
It just may be implemented in the concatenated situation of color characteristic vector sum texture feature vector.
A kind of image search method provided in an embodiment of the present invention, by the color feature vector for extracting target image simultaneously
And texture feature vector, then the two feature vectors are matched with feature vector corresponding in image library respectively again.Make
Designer can be quickly found out oneself desired image product in design work, avoid in face of large nuber of images, rule of thumb
With the worries of memory search, traditional approach (such as by artificial typing field) search bring inaccuracy and something lost are also avoided
The shortcomings that leakage, greatly alleviates the workload of designer, has liberated the thinking of designer, and design is allowed to return itself.It is time saving and energy saving
While, it is difficult to reduce retrieval.
Correspondingly, the embodiment of the invention also provides a kind of image indexing system structural schematic diagrams.It is specific as shown in Fig. 2,
The system includes: receiving unit 201, extraction unit 202 and matching unit 203.
Receiving unit 201 selects area-of-interest for receiving operational order, and according to operational order on target image
ROI;
Extraction unit 202, for extracting the characteristics of image of ROI;
A matching unit 203, for carrying out the characteristics of image of image in the characteristics of image of ROI and pre-established image library
Match, obtains at least two candidate images.
Optionally, the characteristics of image of ROI includes: color characteristic and/or textural characteristics.
Extraction unit 202 is specifically used for: the color feature vector of ROI is extracted according to preset color feature extracted algorithm.
And the texture feature vector of ROI is extracted according to preset Texture Segmentation Algorithm.
Specifically, extraction unit 202 extracts the color feature vector of ROI, tool according to preset color feature extracted algorithm
Body includes: to be converted to hsv color spatial image to the RGB image of ROI;
And non-uniform quantizing is carried out to hsv color spatial image, and assign each pixel in hsv color spatial image
New HSV value;
The new HSV value of each pixel in hsv color spatial image is converted into color feature vector respectively;
According to the color feature vector of each pixel in hsv color spatial image, the color feature vector of ROI is obtained.
Extraction unit 202 extracts the texture feature vector of ROI according to preset Texture Segmentation Algorithm, specifically includes:
The image of ROI is converted into gray level image;
Feature point extraction is carried out to gray level image, obtains at least two characteristic points;
It predicts the cluster centre at least two characteristic points where each characteristic point, and obtains each characteristic point and gather
Residual error between class center;
Obtain belong to residual error between all characteristic points of same cluster centre and the cluster centre and value;
And to being normalized with value and dimension-reduction treatment, the texture feature vector of ROI is obtained.
Optionally, matching unit 203 is specifically used for: by image in the color feature vector of ROI and pre-established image library
Color feature vector matched, obtain at least one candidate image;
And match the texture feature vector of ROI with the texture feature vector of image in pre-established image library,
Obtain at least one candidate image.
Optionally, which can also include: display unit 204, for showing according to face according to preset weight proportion
The number at least one candidate image that color characteristic vector obtains after being matched, and matched according to texture feature vector
The number of the only candidate image obtained afterwards.
Further alternative, matching unit 203 can be also used for, by the color feature vector and texture feature vector of ROI
It connects, the feature vector after obtaining series connection;
Feature vector after series connection is matched with the feature vector of connecting of image in pre-established image library, is obtained extremely
Few two candidate images.
Function performed by each component in the system has been done in a kind of image search method of above-described embodiment
Detailed introduction, which is not described herein again.
A kind of image indexing system provided in an embodiment of the present invention, by the color feature vector for extracting target image simultaneously
And texture feature vector, then the two feature vectors are matched with feature vector corresponding in image library respectively again.Make
Designer can be quickly found out oneself desired image product in design work, avoid in face of large nuber of images, rule of thumb
With the worries of memory search, traditional approach (such as by artificial typing field) search bring inaccuracy and something lost are also avoided
The shortcomings that leakage, greatly alleviates the workload of designer, has liberated the thinking of designer, and design is allowed to return itself.It is time saving and energy saving
While, it is difficult to reduce retrieval.
Reader should be understood that in the description of this specification reference term " one embodiment ", " is shown " some embodiments "
The description of example ", " specific example " or " some examples " etc. mean specific features described in conjunction with this embodiment or example, structure,
Material or feature are included at least one embodiment or example of the invention.In the present specification, above-mentioned term is shown
The statement of meaning property need not be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described
It may be combined in any suitable manner in any one or more of the embodiments or examples.In addition, without conflicting with each other, this
The technical staff in field can be by the spy of different embodiments or examples described in this specification and different embodiments or examples
Sign is combined.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is not considered as limiting the invention, those skilled in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, modifies, replacement and variant.
Claims (10)
1. a kind of image search method, which is characterized in that the described method includes:
Operational order is received, and region of interest ROI is selected on target image according to the operational order;
Extract the characteristics of image of the ROI;
The characteristics of image of the ROI is matched with the characteristics of image of image in pre-established image library, obtains at least two
Candidate image.
2. the method according to claim 1, wherein the characteristics of image of the ROI includes: color characteristic and texture
Feature;
The characteristics of image for extracting the ROI, specifically includes:
The color feature vector of the ROI is extracted according to preset color feature extracted algorithm;
And the texture feature vector of the ROI is extracted according to preset Texture Segmentation Algorithm.
3. according to the method described in claim 2, it is characterized in that, described extract institute according to preset color feature extracted algorithm
The color feature vector for stating ROI, specifically includes:
Hsv color spatial image is converted to the RGB image of the ROI;
And non-uniform quantizing is carried out to the hsv color spatial image, and assign each picture in the hsv color spatial image
The new HSV value of element;
The new HSV value of each pixel in the hsv color spatial image is converted into color feature vector respectively;
According to the color feature vector of each pixel in the hsv color spatial image, obtain the color characteristic of the ROI to
Amount.
4. according to the method described in claim 3, it is characterized in that, described extract institute according to preset Texture Segmentation Algorithm
The texture feature vector for stating ROI, specifically includes:
The image of the ROI is converted into gray level image;
Feature point extraction is carried out to the gray level image, obtains at least two characteristic points;
It predicts the cluster centre at least two characteristic point where each characteristic point, and obtains each characteristic point and institute
State the residual error between cluster centre;
Obtain belong to residual error between all characteristic points of same cluster centre and the cluster centre and value;
And described and value is normalized and dimension-reduction treatment, obtain the texture feature vector of the ROI.
5. according to the method described in claim 4, it is characterized in that, the characteristics of image by the ROI and pre-established figure
As the characteristics of image of image in library are matched, at least two candidate images are obtained, are specifically included:
The color feature vector of ROI is matched with the color feature vector of image in the pre-established image library, is obtained
At least one candidate image;
And the texture feature vector of image in the texture feature vector of the ROI and the pre-established image library is carried out
Matching, obtains at least one candidate image.
6. according to the method described in claim 5, it is characterized in that, the characteristics of image by the ROI and pre-established figure
As the characteristics of image of image in library is matched, after obtaining at least two candidate images, the method also includes:
According to preset weight proportion, show it is described matched according to color feature vector after at least one candidate's figure for obtaining
The number of picture and it is described matched according to texture feature vector after obtain only a candidate image number.
7. according to the method described in claim 4, it is characterized in that, the characteristics of image by the ROI and pre-established figure
As the characteristics of image of image in library are matched, at least two candidate images are obtained, are specifically included:
The color feature vector of the ROI and texture feature vector are connected, the feature vector after obtaining series connection;
Feature vector after the series connection is matched with the feature vector of connecting of image in the pre-established image library, is obtained
Take at least two candidate images.
8. a kind of image indexing system, which is characterized in that the system comprises:
Receiving unit selects area-of-interest for receiving operational order, and according to the operational order on target image
ROI;
Extraction unit, for extracting the characteristics of image of the ROI;
Matching unit, for the characteristics of image of the ROI to be matched with the characteristics of image of image in pre-established image library,
Obtain at least two candidate images.
9. system according to claim 8, which is characterized in that the characteristics of image of the ROI include: color characteristic and/or
Textural characteristics;
The extraction unit is specifically used for: being converted to hsv color spatial image to the RGB image of the ROI;
And non-uniform quantizing is carried out to the hsv color spatial image, and assign each picture in the hsv color spatial image
The new HSV value of element;
The new HSV value of each pixel in the hsv color spatial image is converted into color feature vector respectively;
According to the color feature vector of each pixel in the hsv color spatial image, obtain the color characteristic of the ROI to
Amount.
10. system according to claim 9, which is characterized in that the extraction unit is also used to, by the image of the ROI
Be converted to gray level image;
Feature point extraction is carried out to the gray level image, obtains at least two characteristic points;
It predicts the cluster centre at least two characteristic point where each characteristic point, and obtains each characteristic point and institute
State the residual error between cluster centre;
Obtain belong to residual error between all characteristic points of same cluster centre and the cluster centre and value;
And described and value is normalized and dimension-reduction treatment, obtain the texture feature vector of the ROI.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810275778.1A CN108959338A (en) | 2018-03-30 | 2018-03-30 | A kind of image search method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810275778.1A CN108959338A (en) | 2018-03-30 | 2018-03-30 | A kind of image search method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108959338A true CN108959338A (en) | 2018-12-07 |
Family
ID=64498554
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810275778.1A Pending CN108959338A (en) | 2018-03-30 | 2018-03-30 | A kind of image search method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108959338A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101211355A (en) * | 2006-12-30 | 2008-07-02 | 中国科学院计算技术研究所 | Image inquiry method based on clustering |
JP2008216066A (en) * | 2007-03-05 | 2008-09-18 | Kddi Corp | Analogous image retrieval device |
CN101639858A (en) * | 2009-08-21 | 2010-02-03 | 深圳创维数字技术股份有限公司 | Image search method based on target area matching |
CN105631001A (en) * | 2015-12-28 | 2016-06-01 | 辽宁师范大学 | Multi-feature fusion based color image retrieval method for HSV space image retrieval |
-
2018
- 2018-03-30 CN CN201810275778.1A patent/CN108959338A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101211355A (en) * | 2006-12-30 | 2008-07-02 | 中国科学院计算技术研究所 | Image inquiry method based on clustering |
JP2008216066A (en) * | 2007-03-05 | 2008-09-18 | Kddi Corp | Analogous image retrieval device |
CN101639858A (en) * | 2009-08-21 | 2010-02-03 | 深圳创维数字技术股份有限公司 | Image search method based on target area matching |
CN105631001A (en) * | 2015-12-28 | 2016-06-01 | 辽宁师范大学 | Multi-feature fusion based color image retrieval method for HSV space image retrieval |
Non-Patent Citations (2)
Title |
---|
刘铝: "基于内容的图像检索方法的研究与实现", 《万方数据库》 * |
达达MFZ: "图像检索——VLAD", 《博客园》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5797789B2 (en) | Method and system for quasi-duplicate image retrieval | |
Han et al. | Fuzzy color histogram and its use in color image retrieval | |
CN112766291B (en) | Matching method for specific target object in scene image | |
Kong | Image retrieval using both color and texture features | |
Kalist et al. | Possiblistic-fuzzy C-means clustering approach for the segmentation of satellite images in HSL color space | |
WO2005024662A3 (en) | Method and apparatus for indexing and searching graphic elements | |
JP2002319024A (en) | Image retrieval method based on combination of color and material feeling | |
Celebi et al. | An effective real-time color quantization method based on divisive hierarchical clustering | |
Kavitha et al. | An efficient content based image retrieval using color and texture of image sub blocks | |
US11347792B2 (en) | Video abstract generating method, apparatus, and storage medium | |
CN103236081A (en) | Registration method for color point cloud | |
Saad et al. | Image retrieval based on integration between YC b C r color histogram and shape feature | |
Suhasini et al. | Content based image retrieval based on different global and local color histogram methods: a survey | |
CN112561976A (en) | Image dominant color feature extraction method, image retrieval method, storage medium and device | |
CN114187215A (en) | Local recoloring algorithm based on image segmentation | |
KR20030005908A (en) | apparatus and method for representative feature extraction of objects and, its application to content-based image retrieval | |
CN113223159A (en) | Single remote sensing image three-dimensional modeling method based on target texture virtualization processing | |
CN108959338A (en) | A kind of image search method and system | |
Li et al. | A Student's t‐based density peaks clustering with superpixel segmentation (tDPCSS) method for image color clustering | |
JP2003216948A (en) | Representative color extraction device and representative color extraction program | |
Mansoori et al. | Content based image retrieval using the knowledge of texture, color and binary tree structure | |
Rejeb et al. | Image retrieval using spatial dominant color descriptor | |
Liu et al. | Palette-based recoloring of natural images under different illumination | |
Reddy et al. | Color sketch based image retrieval | |
Nalini et al. | Performances of Different Color Representations in Image Retrieval and Classification: A Comparative Analysis |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181207 |
|
RJ01 | Rejection of invention patent application after publication |