CN104217006A - Method and device for searching image - Google Patents
Method and device for searching image Download PDFInfo
- Publication number
- CN104217006A CN104217006A CN201410470435.2A CN201410470435A CN104217006A CN 104217006 A CN104217006 A CN 104217006A CN 201410470435 A CN201410470435 A CN 201410470435A CN 104217006 A CN104217006 A CN 104217006A
- Authority
- CN
- China
- Prior art keywords
- image
- stored image
- obtains
- feature vector
- matching result
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Library & Information Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for searching an image, which are used for more accurately and effectively searching the image. The method comprises the following steps of extracting a scale invariant feature transform feature point of a specified image; describing the scale invariant feature transform feature point of the specified image to obtain a first feature vector of the specified image; performing binaryzation on the first feature vector of the specified image to obtain a second feature vector of the specified image; according to the second feature vector of the specified image, selecting pre-stored images matched with the specified image from the second feature vectors of pre-stored images to obtain a first matching result. The optimal matching result is obtained by using the characteristics of insensitive illumination, scale invariance and irradiation invariance of the scale invariant feature transform feature point, extracting the scale invariant feature transform feature point of the image, and selecting the optimal matched image from the pre-stored images.
Description
Technical field
The present invention relates to Internet technical field, particularly a kind of method of image retrieval and device.
Background technology
Based on image characteristics extraction to realize the process of image retrieval, be widely used in various research field.Based on image characteristics extraction to realize the process of image retrieval namely for image zooming-out unique point, and unique point is mated with the unique point in feature database, determine the most similar matching result, thus realize the object of image retrieval.In prior art, LBP (the Local Binary Patterns of image is extracted in usual employing, local binary patterns) method of unique point or ORB (oriented FAST and rotated BRIEF) unique point mates with feature database, but, this method is only mated unique point, and the imaging accuracy retrieved is lower.
Summary of the invention
The invention provides a kind of method and device of image retrieval, in order to realize more accurately, more efficiently retrieving images.
The invention provides a kind of method of image retrieval, comprising:
Extract the scale invariant feature converting characteristic point of specifying image;
The scale invariant feature converting characteristic point of described appointment image is described, obtains the first eigenvector of described appointment image;
Binaryzation is carried out to the first eigenvector of described appointment image, obtains the second feature vector of described appointment image;
Second feature vector according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image, obtains the first matching result.
Some beneficial effects of the embodiment of the present invention can comprise: the feature of the illumination-insensitive utilizing scale invariant feature converting characteristic point to have, yardstick, radiation unchangeability, from pre-stored image, screen matching image by the scale invariant feature converting characteristic point extracting image, thus obtain best matching result.
Optionally, described method also comprises:
Extract the scale invariant feature converting characteristic point of described pre-stored image;
The scale invariant feature converting characteristic point of described pre-stored image is described, obtains the first eigenvector of described pre-stored image;
Binaryzation is carried out to the first eigenvector of described pre-stored image, obtains the second feature vector of described pre-stored image;
Store the second feature vector of described pre-stored image.
In possibility, extract the scale invariant feature converting characteristic point of pre-stored image in advance, and the scale invariant feature converting characteristic point of pre-stored image is described, binaryzation, storage, image can be mated with pre-stored image by scale invariant feature converting characteristic point, thus obtain best matching result.
Optionally, the described vector of the second feature according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image, obtains the first matching result, comprising:
Calculate the first matching degree of the second feature vector of the second feature vector sum pre-stored image of described appointment image;
From the second feature vector of described pre-stored image, filter out the pre-stored image of described first matching degree higher than the first predetermined threshold value, obtain the first matching result.
In possibility, the matching degree according to the second feature vector of specifying image and pre-stored image carries out optical sieving, makes image retrieval more accurate, to obtain best matching result.
Optionally, the described vector of the second feature according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image, and after obtaining the first matching result, described method also comprises:
Extract the geological information of the scale invariant feature converting characteristic point of described appointment image;
Geological information according to the scale invariant feature converting characteristic point of described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result.
In possibility, after carrying out images match according to scale invariant feature converting characteristic point, carry out images match according to the geological information of scale invariant feature converting characteristic point further again, thus filter out the image mated the most, make image retrieval more accurate.
Optionally, described geological information comprises principal direction and the dimensional information of described scale invariant feature converting characteristic point.
Optionally, the geological information of the described scale invariant feature converting characteristic point according to described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result, comprising:
Calculate the second matching degree of the geological information in the geological information of the scale invariant feature converting characteristic point of described appointment image and described first matching result;
From the geological information described first matching result, filter out the pre-stored image of described second matching degree higher than the second predetermined threshold value, obtain the second matching result.
In possibility, the matching degree according to the geological information of specifying the scale invariant feature converting characteristic point of image and pre-stored image carries out optical sieving, makes image retrieval more accurate, to obtain best matching result.
Optionally, described method also comprises:
Pre-stored image in described second matching result is sorted from high to low according to described second matching degree;
Obtain a predetermined number pre-stored image before in described sequence, obtain the 3rd matching result.
In possibility, according to matching degree, the pre-stored image filtered out is sorted, and the preceding image that will sort is as result for retrieval, to obtain the highest pre-stored image of matching degree, obtains best matching result.
Optionally, the dimension of described first eigenvector is 128 dimensions;
The described first eigenvector to described appointment image carries out binaryzation, obtains the second feature vector of described appointment image, comprising:
The first eigenvector of described 128 dimensions is projected in 64 dimensional feature space and carries out dimension-reduction treatment, obtain the first eigenvector of 64 dimensions;
Binaryzation is carried out to the first eigenvector of described 64 dimensions, obtains the second feature vector of 64 dimensions.
In possibility, the proper vector that 128 tie up is carried out dimension-reduction treatment, reduce data volume during matching image, improve the efficiency of image retrieval.
Optionally, the second feature vector of described pre-stored image is, extract the scale invariant feature converting characteristic point of described pre-stored image, the scale invariant feature converting characteristic of described pre-stored image o'clock is weighted according to the first preset weights, scale invariant feature converting characteristic point after weighting is described, obtain the first eigenvector of described pre-stored image, the first eigenvector of described pre-stored image is carried out binaryzation, obtain the second feature vector of described pre-stored image;
The described scale invariant feature converting characteristic point to described appointment image is described, and after obtaining the first eigenvector of described appointment image, described method also comprises:
The first eigenvector of described appointment image is weighted according to described first preset weights, obtains the third feature vector of described appointment image;
Binaryzation is carried out to the third feature vector of described appointment image, obtains the fourth feature vector of described appointment image;
Fourth feature vector according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image.
In possibility, process is weighted to specifying the first eigenvector of image and pre-stored image, mate the first eigenvector after weighting again, with the unique point of outbalance in outstanding first eigenvector, thus filter out the pre-stored image for Important Characteristic Points.
Optionally, geological information in described first matching result is, obtain the geological information of the scale invariant feature converting characteristic point of described pre-stored image, the geological information of the scale invariant feature converting characteristic point of described pre-stored image is weighted according to the second preset weights;
The geological information of the described scale invariant feature converting characteristic point according to described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result, comprising:
The geological information of the scale invariant feature converting characteristic point of described appointment image is weighted according to described second preset weights;
From the geological information described first matching result, filter out the pre-stored image with described appointment images match according to the geological information after the weighting of the scale invariant feature converting characteristic point of described appointment image, obtain the second matching result.
In possibility, process is weighted to the geological information of the scale invariant feature converting characteristic point of specifying image and pre-stored image, mate the geological information after weighting again, with the information of outbalance in outstanding geological information, thus filter out the pre-stored image for important information.
The embodiment of the present invention also provides a kind of device of image retrieval, comprising:
First extraction module, for extracting the scale invariant feature converting characteristic point of specifying image;
First describing module, for being described the scale invariant feature converting characteristic point of described appointment image, obtains the first eigenvector of described appointment image;
First binarization block, for carrying out binaryzation to the first eigenvector of described appointment image, obtains the second feature vector of described appointment image;
First screening module, filters out the pre-stored image with described appointment images match for the second feature vector according to described appointment image, obtains the first matching result from the second feature vector of pre-stored image.
Optionally, described device also comprises:
Second extraction module, for extracting the scale invariant feature converting characteristic point of described pre-stored image;
Second describing module, for being described the scale invariant feature converting characteristic point of described pre-stored image, obtains the first eigenvector of described pre-stored image;
Second binarization block, for carrying out binaryzation to the first eigenvector of described pre-stored image, obtains the second feature vector of described pre-stored image;
Memory module, for storing the second feature vector of described pre-stored image.
Optionally, described first screening module comprises:
First computing unit, for calculating the first matching degree of the second feature vector of the second feature vector sum pre-stored image of described appointment image;
First screening unit, for filtering out the pre-stored image of described first matching degree higher than the first predetermined threshold value from the second feature vector of described pre-stored image, obtains the first matching result.
Optionally, described device also comprises:
3rd extraction module, for extracting the geological information of the scale invariant feature converting characteristic point of described appointment image;
Second screening module, the geological information for the scale invariant feature converting characteristic point according to described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result.
Optionally, described second screening module comprises:
Second computing unit, for calculating the second matching degree of the geological information in the geological information of the scale invariant feature converting characteristic point of described appointment image and described first matching result;
Second screening unit, for filtering out the pre-stored image of described second matching degree higher than the second predetermined threshold value from the geological information in described first matching result, obtains the second matching result.
Optionally, described device also comprises:
Order module, for sorting the pre-stored image in described second matching result from high to low according to described second matching degree;
Acquisition module, for obtaining a predetermined number pre-stored image before in described sequence, obtains the 3rd matching result.
Optionally, described first binarization block comprises:
Dimensionality reduction unit, carrying out dimension-reduction treatment for being projected in 64 dimensional feature space by the first eigenvector of described 128 dimensions, obtaining the first eigenvector of 64 dimensions;
Binarization unit, for carrying out binaryzation to the first eigenvector of described 64 dimensions, obtains the second feature vector of 64 dimensions.
Optionally, described device also comprises:
Weighting block, for being weighted according to described first preset weights by the first eigenvector of described appointment image, obtains the third feature vector of described appointment image;
3rd binarization block, for carrying out binaryzation to the third feature vector of described appointment image, obtains the fourth feature vector of described appointment image;
Three screening module, filters out the pre-stored image with described appointment images match for the fourth feature vector according to described appointment image from the second feature vector of pre-stored image.
Optionally, described second screening module comprises:
Weighted units, the geological information for the scale invariant feature converting characteristic point by described appointment image is weighted according to described second preset weights;
Three screening unit, from the geological information described first matching result, filter out the pre-stored image with described appointment images match for the geological information after the weighting of the scale invariant feature converting characteristic point according to described appointment image, obtain the second matching result.
Other features and advantages of the present invention will be set forth in the following description, and, partly become apparent from instructions, or understand by implementing the present invention.Object of the present invention and other advantages realize by structure specifically noted in write instructions, claims and accompanying drawing and obtain.
Below by drawings and Examples, technical scheme of the present invention is described in further detail.
Accompanying drawing explanation
Accompanying drawing is used to provide a further understanding of the present invention, and forms a part for instructions, together with embodiments of the present invention for explaining the present invention, is not construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the process flow diagram of the method for a kind of image retrieval in the embodiment of the present invention;
Fig. 2 is the process flow diagram of the method for a kind of image retrieval in the embodiment of the present invention;
Fig. 3 is the block diagram of the device of a kind of image retrieval in the embodiment of the present invention;
Fig. 4 is the block diagram of the device of a kind of image retrieval in the embodiment of the present invention;
Fig. 5 is the block diagram of the first screening module in the device of a kind of image retrieval in the embodiment of the present invention;
Fig. 6 is the block diagram of the second screening module in the device of a kind of image retrieval in the embodiment of the present invention;
Fig. 7 is the block diagram of the first binarization block in the device of a kind of image retrieval in the embodiment of the present invention.
Embodiment
Below in conjunction with accompanying drawing, the preferred embodiments of the present invention are described, should be appreciated that preferred embodiment described herein is only for instruction and explanation of the present invention, is not intended to limit the present invention.
Fig. 1 is the process flow diagram of the method for a kind of image retrieval in the embodiment of the present invention.As shown in Figure 1, S11-S14 is comprised the following steps:
Step S11, extracts the scale invariant feature converting characteristic point of specifying image;
Step S12, to specifying the scale invariant feature converting characteristic point of image to be described, obtains the first eigenvector of specifying image;
Step S13, to specifying the first eigenvector of image to carry out binaryzation, obtains the second feature vector of specifying image;
Step S14, according to specifying the second feature of image vector to filter out and the pre-stored image of specifying images match from the second feature vector of pre-stored image, obtains the first matching result.
Some beneficial effects of the present embodiment can comprise: the feature of the illumination-insensitive utilizing scale invariant feature converting characteristic point to have, yardstick, radiation unchangeability, from pre-stored image, screen matching image by the scale invariant feature converting characteristic point extracting image, thus obtain best matching result.
In step s 12, to scale invariant feature converting characteristic point (i.e. Scale-invariant feature transform unique point, hereinafter referred to as SIFT feature point) be described, namely feature interpretation is carried out to SIFT feature point, comprise various conversion and statistical computation, finally obtain proper vector.
In step s 12, the dimension describing the first eigenvector that SIFT feature point obtains is 128 dimensions, and dimension is higher, therefore can carry out dimension-reduction treatment to first eigenvector.Step S13 can be embodied as following steps S131-S132:
Step S131, projects to the first eigenvector that 128 tie up in 64 dimensional feature space and carries out dimension-reduction treatment, obtain the first eigenvector of 64 dimensions;
Step S132, carries out binaryzation to the first eigenvector of 64 dimensions, obtains the second feature vector of 64 dimensions.
Before retrieving images, need to do identical process to pre-stored image, as shown in Figure 2, specifically can be embodied as following steps S21-S24:
Step S21, extracts the SIFT feature point of pre-stored image;
Step S22, is described the SIFT feature point of pre-stored image, obtains the first eigenvector of pre-stored image;
Step S23, carries out binaryzation to the first eigenvector of pre-stored image, obtains the second feature vector of pre-stored image;
Step S24, stores the second feature vector of pre-stored image.
Said method, step S14 can be embodied as following steps S141-S142:
Step S141, calculates the first matching degree of the second feature vector of the second feature vector sum pre-stored image of specifying image;
Step S142, filters out the pre-stored image of the first matching degree higher than the first predetermined threshold value, obtains the first matching result from the second feature vector of pre-stored image.
After implementation step S142, the further comprising the steps of S143-S144 of said method:
Step S143, extracts the geological information of the SIFT feature point of specifying image;
Step S144, the geological information according to the SIFT feature point of specifying image filters out and the pre-stored image of specifying images match from the geological information the first matching result, obtains the second matching result.
Wherein, geological information comprises principal direction and the dimensional information of SIFT feature point.
In said method, step S144 can be embodied as following steps S1441-S1442:
Step S1441, calculates the second matching degree of the geological information in the geological information of SIFT feature point and the first matching result of specifying image;
Step S1442, filters out the pre-stored image of the second matching degree higher than the second predetermined threshold value, obtains the second matching result from the geological information the first matching result.
Adopt said method to carry out image retrieval, after carrying out images match according to SIFT feature point, then carry out images match according to the geological information of SIFT feature point further, thus filter out the image mated the most, make image retrieval more accurate.
In the second matching result retrieved, pre-stored image may have multiple, therefore also obtains the high image of matching degree further by implementing following steps S145-S146:
Step S145, sorts according to the second matching degree from high to low by the pre-stored image in the second matching result;
Step S146, obtains a predetermined number pre-stored image before in sequence, obtains the 3rd matching result.
In addition, can be weighted specifying image and the SIFT feature point of pre-stored image or the geological information of SIFT feature point, with outstanding key character.Comprise following two kinds of modes:
Mode one
When the second feature vector of pre-stored image obtains in the following manner: the SIFT feature point extracting pre-stored image; The SIFT feature of pre-stored image o'clock is weighted according to the first preset weights; Describe the SIFT feature point after weighting, obtain the first eigenvector of pre-stored image; The first eigenvector of pre-stored image is carried out binaryzation, obtains the second feature vector of pre-stored image.After step S12, the further comprising the steps of A11-A13 of said method:
Steps A 11, will specify the first eigenvector of image to be weighted according to the first preset weights, obtains the third feature vector of specifying image;
Steps A 12, to specifying the third feature vector of image to carry out binaryzation, obtains the fourth feature vector of specifying image;
Steps A 13, filters out and the pre-stored image of specifying images match from the second feature vector of pre-stored image according to specifying the fourth feature of image vector.
Mode two
When the geological information in the first matching result gone out according to SIFT feature point retrieval obtains as follows: the geological information obtaining the SIFT feature point of pre-stored image; The geological information of the SIFT feature of pre-stored image point is weighted according to the second preset weights.After step S144, the further comprising the steps of B11-B12 of said method:
Step B11, is weighted the geological information of the SIFT feature point of specifying image according to the second preset weights;
Step B12, filters out and the pre-stored image of specifying images match according to the geological information after specifying the weighting of the SIFT feature point of image, obtains the second matching result from the geological information the first matching result.
Fig. 3 is the block diagram of the device of a kind of image retrieval in the embodiment of the present invention.As shown in Figure 3, the device of image retrieval comprises:
First extraction module 301, for extracting the scale invariant feature converting characteristic point of specifying image;
First describing module 302, for specifying the scale invariant feature converting characteristic point of image to be described, obtains the first eigenvector of specifying image;
First binarization block 303, for specifying the first eigenvector of image to carry out binaryzation, obtains the second feature vector of specifying image;
First screening module 304, for according to specifying the second feature of image vector to filter out and the pre-stored image of specifying images match from the second feature vector of pre-stored image, obtains the first matching result.
Optionally, as shown in Figure 4, said apparatus also comprises:
Second extraction module 305, for extracting the scale invariant feature converting characteristic point of pre-stored image;
Second describing module 306, for being described the scale invariant feature converting characteristic point of pre-stored image, obtains the first eigenvector of described pre-stored image;
Second binarization block 307, for carrying out binaryzation to the first eigenvector of pre-stored image, obtains the second feature vector of pre-stored image;
Memory module 308, for storing the second feature vector of pre-stored image;
3rd extraction module 309, for extracting the geological information of the scale invariant feature converting characteristic point of specifying image;
Second screening module 310, for filtering out from the geological information the first matching result and the pre-stored image of specifying images match according to the geological information of the scale invariant feature converting characteristic point of specifying image, obtains the second matching result;
Order module 311, for sorting the pre-stored image in the second matching result from high to low according to the second matching degree;
Acquisition module 312, for obtaining a predetermined number pre-stored image before in sequence, obtains the 3rd matching result;
Weighting block 313, for being weighted according to described first preset weights by the first eigenvector of described appointment image, obtains the third feature vector of described appointment image;
3rd binarization block 314, for specifying the third feature vector of image to carry out binaryzation, obtains the fourth feature vector of specifying image;
Three screening module 315, for according to specifying, the fourth feature of image is vectorial to be filtered out and the pre-stored image of specifying images match from the second feature vector of pre-stored image.
Optionally, the first screening module 304 comprises with lower unit, as shown in Figure 5:
First computing unit 3041, for calculating the first matching degree of the second feature vector of the second feature vector sum pre-stored image of specifying image;
First screening unit 3042, for filtering out the pre-stored image of the first matching degree higher than the first predetermined threshold value from the second feature vector of pre-stored image, obtains the first matching result.
Optionally, the second screening module 310 comprises with lower unit, as shown in Figure 6:
Second computing unit 3101, for calculate specify image the geological information of scale invariant feature converting characteristic point and the first matching result in the second matching degree of geological information;
Second screening unit 3102, for filtering out the pre-stored image of the second matching degree higher than the second predetermined threshold value from the geological information in the first matching result, obtains the second matching result;
Weighted units 3103, for being weighted the geological information of the scale invariant feature converting characteristic point of specifying image according to the second preset weights;
Three screening unit 3104, filters out and the pre-stored image of specifying images match for the geological information after the weighting of the scale invariant feature converting characteristic point according to appointment image, obtains the second matching result from the geological information the first matching result.
Optionally, the first binarization block 303 comprises with lower unit, as shown in Figure 7:
Dimensionality reduction unit 3031, the first eigenvector for tieing up 128 projects in 64 dimensional feature space and carries out dimension-reduction treatment, obtains the first eigenvector of 64 dimensions;
Binarization unit 3032, for carrying out binaryzation to the first eigenvector of 64 dimensions, obtains the second feature vector of 64 dimensions.
Some beneficial effects of the present embodiment can comprise: the feature of the illumination-insensitive utilizing scale invariant feature converting characteristic point to have, yardstick, radiation unchangeability, from pre-stored image, screen matching image by the scale invariant feature converting characteristic point extracting image, thus obtain best matching result.
Those skilled in the art should understand, embodiments of the invention can be provided as method, system or computer program.Therefore, the present invention can adopt the form of complete hardware embodiment, completely software implementation or the embodiment in conjunction with software and hardware aspect.And the present invention can adopt in one or more form wherein including the upper computer program implemented of computer-usable storage medium (including but not limited to magnetic disk memory and optical memory etc.) of computer usable program code.
The present invention describes with reference to according to the process flow diagram of the method for the embodiment of the present invention, equipment (system) and computer program and/or block scheme.Should understand can by the combination of the flow process in each flow process in computer program instructions realization flow figure and/or block scheme and/or square frame and process flow diagram and/or block scheme and/or square frame.These computer program instructions can being provided to the processor of multi-purpose computer, special purpose computer, Embedded Processor or other programmable data processing device to produce a machine, making the instruction performed by the processor of computing machine or other programmable data processing device produce device for realizing the function of specifying in process flow diagram flow process or multiple flow process and/or block scheme square frame or multiple square frame.
These computer program instructions also can be stored in can in the computer-readable memory that works in a specific way of vectoring computer or other programmable data processing device, the instruction making to be stored in this computer-readable memory produces the manufacture comprising command device, and this command device realizes the function of specifying in process flow diagram flow process or multiple flow process and/or block scheme square frame or multiple square frame.
These computer program instructions also can be loaded in computing machine or other programmable data processing device, make on computing machine or other programmable devices, to perform sequence of operations step to produce computer implemented process, thus the instruction performed on computing machine or other programmable devices is provided for the step realizing the function of specifying in process flow diagram flow process or multiple flow process and/or block scheme square frame or multiple square frame.
Obviously, those skilled in the art can carry out various change and modification to the present invention and not depart from the spirit and scope of the present invention.Like this, if these amendments of the present invention and modification belong within the scope of the claims in the present invention and equivalent technologies thereof, then the present invention is also intended to comprise these change and modification.
Claims (19)
1. a method for image retrieval, is characterized in that, comprising:
Extract the scale invariant feature converting characteristic point of specifying image;
The scale invariant feature converting characteristic point of described appointment image is described, obtains the first eigenvector of described appointment image;
Binaryzation is carried out to the first eigenvector of described appointment image, obtains the second feature vector of described appointment image;
Second feature vector according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image, obtains the first matching result.
2. the method for claim 1, is characterized in that, described method also comprises:
Extract the scale invariant feature converting characteristic point of described pre-stored image;
The scale invariant feature converting characteristic point of described pre-stored image is described, obtains the first eigenvector of described pre-stored image;
Binaryzation is carried out to the first eigenvector of described pre-stored image, obtains the second feature vector of described pre-stored image;
Store the second feature vector of described pre-stored image.
3. the method for claim 1, is characterized in that, the described vector of the second feature according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image, obtains the first matching result, comprising:
Calculate the first matching degree of the second feature vector of the second feature vector sum pre-stored image of described appointment image;
From the second feature vector of described pre-stored image, filter out the pre-stored image of described first matching degree higher than the first predetermined threshold value, obtain the first matching result.
4. method as claimed in claim 3, it is characterized in that, the described vector of the second feature according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image, and after obtaining the first matching result, described method also comprises:
Extract the geological information of the scale invariant feature converting characteristic point of described appointment image;
Geological information according to the scale invariant feature converting characteristic point of described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result.
5. method as claimed in claim 4, it is characterized in that, described geological information comprises principal direction and the dimensional information of described scale invariant feature converting characteristic point.
6. method as claimed in claim 4, it is characterized in that, the geological information of the described scale invariant feature converting characteristic point according to described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result, comprising:
Calculate the second matching degree of the geological information in the geological information of the scale invariant feature converting characteristic point of described appointment image and described first matching result;
From the geological information described first matching result, filter out the pre-stored image of described second matching degree higher than the second predetermined threshold value, obtain the second matching result.
7. method as claimed in claim 4, it is characterized in that, described method also comprises:
Pre-stored image in described second matching result is sorted from high to low according to described second matching degree;
Obtain a predetermined number pre-stored image before in described sequence, obtain the 3rd matching result.
8. the method for claim 1, is characterized in that, the dimension of described first eigenvector is 128 dimensions;
The described first eigenvector to described appointment image carries out binaryzation, obtains the second feature vector of described appointment image, comprising:
The first eigenvector of described 128 dimensions is projected in 64 dimensional feature space and carries out dimension-reduction treatment, obtain the first eigenvector of 64 dimensions;
Binaryzation is carried out to the first eigenvector of described 64 dimensions, obtains the second feature vector of 64 dimensions.
9. the method for claim 1, is characterized in that,
The second feature vector of described pre-stored image is, extract the scale invariant feature converting characteristic point of described pre-stored image, the scale invariant feature converting characteristic of described pre-stored image o'clock is weighted according to the first preset weights, scale invariant feature converting characteristic point after weighting is described, obtain the first eigenvector of described pre-stored image, the first eigenvector of described pre-stored image is carried out binaryzation, obtains the second feature vector of described pre-stored image;
The described scale invariant feature converting characteristic point to described appointment image is described, and after obtaining the first eigenvector of described appointment image, described method also comprises:
The first eigenvector of described appointment image is weighted according to described first preset weights, obtains the third feature vector of described appointment image;
Binaryzation is carried out to the third feature vector of described appointment image, obtains the fourth feature vector of described appointment image;
Fourth feature vector according to described appointment image filters out the pre-stored image with described appointment images match from the second feature vector of pre-stored image.
10. method as claimed in claim 4, is characterized in that,
Geological information in described first matching result is, obtains the geological information of the scale invariant feature converting characteristic point of described pre-stored image, is weighted by the geological information of the scale invariant feature converting characteristic point of described pre-stored image according to the second preset weights;
The geological information of the described scale invariant feature converting characteristic point according to described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result, comprising:
The geological information of the scale invariant feature converting characteristic point of described appointment image is weighted according to described second preset weights;
From the geological information described first matching result, filter out the pre-stored image with described appointment images match according to the geological information after the weighting of the scale invariant feature converting characteristic point of described appointment image, obtain the second matching result.
The device of 11. 1 kinds of image retrievals, is characterized in that, comprising:
First extraction module, for extracting the scale invariant feature converting characteristic point of specifying image;
First describing module, for being described the scale invariant feature converting characteristic point of described appointment image, obtains the first eigenvector of described appointment image;
First binarization block, for carrying out binaryzation to the first eigenvector of described appointment image, obtains the second feature vector of described appointment image;
First screening module, filters out the pre-stored image with described appointment images match for the second feature vector according to described appointment image, obtains the first matching result from the second feature vector of pre-stored image.
12. devices as claimed in claim 11, it is characterized in that, described device also comprises:
Second extraction module, for extracting the scale invariant feature converting characteristic point of described pre-stored image;
Second describing module, for being described the scale invariant feature converting characteristic point of described pre-stored image, obtains the first eigenvector of described pre-stored image;
Second binarization block, for carrying out binaryzation to the first eigenvector of described pre-stored image, obtains the second feature vector of described pre-stored image;
Memory module, for storing the second feature vector of described pre-stored image.
13. devices as claimed in claim 11, is characterized in that, described first screening module comprises:
First computing unit, for calculating the first matching degree of the second feature vector of the second feature vector sum pre-stored image of described appointment image;
First screening unit, for filtering out the pre-stored image of described first matching degree higher than the first predetermined threshold value from the second feature vector of described pre-stored image, obtains the first matching result.
14. devices as claimed in claim 13, it is characterized in that, described device also comprises:
3rd extraction module, for extracting the geological information of the scale invariant feature converting characteristic point of described appointment image;
Second screening module, the geological information for the scale invariant feature converting characteristic point according to described appointment image filters out the pre-stored image with described appointment images match from the geological information described first matching result, obtains the second matching result.
15. devices as claimed in claim 14, is characterized in that, described second screening module comprises:
Second computing unit, for calculating the second matching degree of the geological information in the geological information of the scale invariant feature converting characteristic point of described appointment image and described first matching result;
Second screening unit, for filtering out the pre-stored image of described second matching degree higher than the second predetermined threshold value from the geological information in described first matching result, obtains the second matching result.
16. devices as claimed in claim 14, it is characterized in that, described device also comprises:
Order module, for sorting the pre-stored image in described second matching result from high to low according to described second matching degree;
Acquisition module, for obtaining a predetermined number pre-stored image before in described sequence, obtains the 3rd matching result.
17. devices as claimed in claim 11, it is characterized in that, described first binarization block comprises:
Dimensionality reduction unit, carrying out dimension-reduction treatment for being projected in 64 dimensional feature space by the first eigenvector of described 128 dimensions, obtaining the first eigenvector of 64 dimensions;
Binarization unit, for carrying out binaryzation to the first eigenvector of described 64 dimensions, obtains the second feature vector of 64 dimensions.
18. devices as claimed in claim 11, it is characterized in that, described device also comprises:
Weighting block, for being weighted according to described first preset weights by the first eigenvector of described appointment image, obtains the third feature vector of described appointment image;
3rd binarization block, for carrying out binaryzation to the third feature vector of described appointment image, obtains the fourth feature vector of described appointment image;
Three screening module, filters out the pre-stored image with described appointment images match for the fourth feature vector according to described appointment image from the second feature vector of pre-stored image.
19. devices as claimed in claim 14, is characterized in that, described second screening module comprises:
Weighted units, the geological information for the scale invariant feature converting characteristic point by described appointment image is weighted according to described second preset weights;
Three screening unit, from the geological information described first matching result, filter out the pre-stored image with described appointment images match for the geological information after the weighting of the scale invariant feature converting characteristic point according to described appointment image, obtain the second matching result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410470435.2A CN104217006A (en) | 2014-09-15 | 2014-09-15 | Method and device for searching image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410470435.2A CN104217006A (en) | 2014-09-15 | 2014-09-15 | Method and device for searching image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104217006A true CN104217006A (en) | 2014-12-17 |
Family
ID=52098496
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410470435.2A Pending CN104217006A (en) | 2014-09-15 | 2014-09-15 | Method and device for searching image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104217006A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106980834A (en) * | 2017-03-24 | 2017-07-25 | 中山大学 | A kind of automatic classification method of facial image |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004786A (en) * | 2010-12-02 | 2011-04-06 | 上海交通大学 | Acceleration method in image retrieval system |
| CN102722731A (en) * | 2012-05-28 | 2012-10-10 | 南京航空航天大学 | Efficient image matching method based on improved scale invariant feature transform (SIFT) algorithm |
| US8352494B1 (en) * | 2009-12-07 | 2013-01-08 | Google Inc. | Distributed image search |
| CN103236068A (en) * | 2013-05-14 | 2013-08-07 | 中国科学院自动化研究所 | Method for matching local images |
| CN103617431A (en) * | 2013-11-05 | 2014-03-05 | 北京工业大学 | Maximum average entropy-based scale-invariant feature transform (SIFT) descriptor binaryzation and similarity matching method |
-
2014
- 2014-09-15 CN CN201410470435.2A patent/CN104217006A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8352494B1 (en) * | 2009-12-07 | 2013-01-08 | Google Inc. | Distributed image search |
| CN102004786A (en) * | 2010-12-02 | 2011-04-06 | 上海交通大学 | Acceleration method in image retrieval system |
| CN102722731A (en) * | 2012-05-28 | 2012-10-10 | 南京航空航天大学 | Efficient image matching method based on improved scale invariant feature transform (SIFT) algorithm |
| CN103236068A (en) * | 2013-05-14 | 2013-08-07 | 中国科学院自动化研究所 | Method for matching local images |
| CN103617431A (en) * | 2013-11-05 | 2014-03-05 | 北京工业大学 | Maximum average entropy-based scale-invariant feature transform (SIFT) descriptor binaryzation and similarity matching method |
Non-Patent Citations (4)
| Title |
|---|
| 国务院学位委员会办公室 等: "《第六届博士生学术年会论文集 下》", 28 February 2009 * |
| 杨化超: "《图像局部不变性特征及其匹配问题研究与应用》", 31 December 2013 * |
| 陈宗海: "《系统仿真技术及其应用 第11卷》", 31 July 2009 * |
| 颜色和SIFT特征融合的图像检索算法: "《第十六届全国图象图形学学术会议暨第六届立体图象技术学术研讨会论文集》", 31 July 2012 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106980834A (en) * | 2017-03-24 | 2017-07-25 | 中山大学 | A kind of automatic classification method of facial image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mohedano et al. | Bags of local convolutional features for scalable instance search | |
| Strecha et al. | LDAHash: Improved matching with smaller descriptors | |
| US9349072B2 (en) | Local feature based image compression | |
| Berton et al. | Deep visual geo-localization benchmark | |
| Aly et al. | Indexing in large scale image collections: Scaling properties and benchmark | |
| KR101650994B1 (en) | Systems and methods to improve feature generation in object recognition | |
| CN108875487B (en) | Training of pedestrian re-recognition network and pedestrian re-recognition based on training | |
| Lahdenoja et al. | Towards understanding the formation of uniform local binary patterns | |
| Iakovidou et al. | Localizing global descriptors for content-based image retrieval | |
| CN104462199A (en) | Near-duplicate image search method in network environment | |
| CN112784009B (en) | Method and device for mining subject term, electronic equipment and storage medium | |
| CN111291695B (en) | Training method and recognition method for recognition model of personnel illegal behaviors and computer equipment | |
| Bi et al. | Multiple instance dense connected convolution neural network for aerial image scene classification | |
| Lin et al. | Robust fisher codes for large scale image retrieval | |
| CN105095468A (en) | Novel image retrieval method and system | |
| Fraundorfer et al. | Visual word based location recognition in 3d models using distance augmented weighting | |
| KR20120051952A (en) | Method and apparatus for image searching using feature point | |
| Amato et al. | Aggregating binary local descriptors for image retrieval | |
| WO2016037848A1 (en) | Image recognition using descriptor pruning | |
| CN104217006A (en) | Method and device for searching image | |
| Pratikakis et al. | Predictive digitisation of cultural heritage objects | |
| Law et al. | Hybrid pooling fusion in the bow pipeline | |
| CN113688261A (en) | Image data cleaning method and device, electronic equipment and readable storage medium | |
| CN113191401A (en) | Method and device for three-dimensional model recognition based on visual saliency sharing | |
| CN110765291A (en) | Retrieval method and device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20141217 |