CN111581416A - Picture retrieval method, device and storage medium - Google Patents
Picture retrieval method, device and storage medium Download PDFInfo
- Publication number
- CN111581416A CN111581416A CN202010291437.0A CN202010291437A CN111581416A CN 111581416 A CN111581416 A CN 111581416A CN 202010291437 A CN202010291437 A CN 202010291437A CN 111581416 A CN111581416 A CN 111581416A
- Authority
- CN
- China
- Prior art keywords
- picture
- feature vector
- retrieved
- feature
- vector
- 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.)
- Withdrawn
Links
Images
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)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
A picture retrieval method, device and storage medium, including obtaining at least three different feature vectors of the picture to be retrieved; the at least three different feature vectors include: a first feature vector, a second feature vector, a third feature vector; wherein the third feature vector has a higher contrast accuracy than the first and second feature vectors; respectively comparing the acquired picture to be retrieved with each picture in a prestored picture set, and determining whether the picture which is the same as the picture to be retrieved exists in the prestored picture set according to the comparison result with each picture; wherein the pre-stored picture set includes the at least three different feature vectors for each picture. The method and the device not only ensure the accuracy of picture retrieval, but also ensure the speed of picture retrieval.
Description
Technical Field
The present disclosure relates to computer technologies, and in particular, to a method and an apparatus for retrieving pictures and a storage medium.
Background
In the advertisement picture retrieval service, a user needs to judge whether a large number of pictures (million levels or more) from the internet are specified advertisement pictures (hundred levels or more). In a conventional picture retrieval algorithm, after a vector (e.g., a 64-dimensional vector) with a specified length is generated for a picture, the similarity between the vectors corresponding to two pictures is compared to determine whether the two pictures are the same. The picture vector generation algorithm based on the perceptual hash algorithm and HSV color space mapping is introduced, and the traditional local feature algorithm is used for checking, so that the accuracy and the retrieval speed in the advertisement picture retrieval service are ensured.
Disclosure of Invention
The application provides a picture retrieval method, a picture retrieval device and a storage medium, which can achieve the purposes of not only ensuring the accuracy of picture retrieval, but also ensuring the speed of picture retrieval.
The application provides a picture retrieval method, a picture retrieval device and a storage medium, wherein the picture retrieval method comprises the steps of obtaining at least three different feature vectors of a picture to be retrieved; the at least three different feature vectors include: a first feature vector, a second feature vector, a third feature vector; wherein the third feature vector has a higher contrast accuracy than the first and second feature vectors; respectively comparing the acquired picture to be retrieved with each picture in a prestored picture set, and determining whether the picture which is the same as the picture to be retrieved exists in the prestored picture set according to the comparison result with each picture; wherein the pre-stored picture set includes the at least three different feature vectors for each picture.
Compared with the prior art, the embodiment of the application firstly adopts the feature vector with low picture precision to compare when each picture is compared, and determines whether to adopt the feature vector with high precision to compare according to the comparison result of the feature vector with low precision, thereby not only ensuring the accuracy of picture retrieval, but also ensuring the speed of picture retrieval.
Additional features and advantages of the application will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the application. Other advantages of the present application may be realized and attained by the instrumentalities and combinations particularly pointed out in the specification and the drawings.
Drawings
The accompanying drawings are included to provide an understanding of the present disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the examples serve to explain the principles of the disclosure and not to limit the disclosure.
FIG. 1 is a flowchart of a wafer retrieval method according to an embodiment of the present application;
fig. 2 is an exemplary diagram of an advertisement picture retrieval according to an embodiment of the present application.
Detailed Description
The present application describes embodiments, but the description is illustrative rather than limiting and it will be apparent to those of ordinary skill in the art that many more embodiments and implementations are possible within the scope of the embodiments described herein. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are possible. Any feature or element of any embodiment may be used in combination with or instead of any other feature or element in any other embodiment, unless expressly limited otherwise.
The present application includes and contemplates combinations of features and elements known to those of ordinary skill in the art. The embodiments, features and elements disclosed in this application may also be combined with any conventional features or elements to form a unique inventive concept as defined by the claims. Any feature or element of any embodiment may also be combined with features or elements from other inventive aspects to form yet another unique inventive aspect, as defined by the claims. Thus, it should be understood that any of the features shown and/or discussed in this application may be implemented alone or in any suitable combination. Accordingly, the embodiments are not limited except as by the appended claims and their equivalents. Furthermore, various modifications and changes may be made within the scope of the appended claims.
Further, in describing representative embodiments, the specification may have presented the method and/or process as a particular sequence of steps. However, to the extent that the method or process does not rely on the particular order of steps set forth herein, the method or process should not be limited to the particular sequence of steps described. Other orders of steps are possible as will be understood by those of ordinary skill in the art. Therefore, the particular order of the steps set forth in the specification should not be construed as limitations on the claims. Further, the claims directed to the method and/or process should not be limited to the performance of their steps in the order written, and one skilled in the art can readily appreciate that the sequences may be varied and still remain within the spirit and scope of the embodiments of the present application.
As shown in fig. 1, the image retrieval method in the embodiment of the present application includes the following operations:
s1, acquiring at least three different feature vectors of the picture to be retrieved; the at least three different feature vectors include: a first feature vector, a second feature vector, a third feature vector; wherein the third feature vector has a higher contrast accuracy than the first and second feature vectors;
s2, comparing the acquired picture to be retrieved with each picture in a pre-stored picture set respectively, and determining whether the picture which is the same as the picture to be retrieved exists in the pre-stored picture set according to the comparison result with each picture; wherein the pre-stored picture set comprises the at least three different feature vectors of each picture;
wherein the process of comparing with any one of the pictures comprises: and determining whether to use a third feature vector to continue comparison according to the comparison result of the picture to be retrieved and the first feature vector and/or the second feature vector of the picture.
In picture retrieval, a feature vector with high precision is generally adopted for retrieval, the accuracy is high, but the defect is that the processing speed is low; if the retrieval is carried out by using the feature vector with low precision, the processing speed is high, but the precision is low. According to the embodiment of the application, when each picture is compared, the feature vector with low picture precision is firstly adopted for comparison, and whether the feature vector with high precision is adopted for comparison is determined according to the comparison result of the feature vector with low precision, so that the accuracy rate of picture retrieval is ensured, and the speed of picture retrieval is also ensured.
In an exemplary embodiment, the determining whether to use the third feature vector for comparison according to the comparison result between the picture to be retrieved and the first feature vector and/or the second feature vector of the picture in operation S2 includes:
s21, when the similarity obtained by comparing the picture to be retrieved with the first characteristic vector or the second characteristic vector of the picture is larger than a first preset threshold value, comparing the picture to be retrieved with a third characteristic vector of the picture, and taking the comparison result of the third characteristic vector as the comparison result with the picture;
and S22, when the similarity obtained by comparing the first characteristic vector with the second characteristic vector is not larger than a first preset threshold, continuing to compare without using the third characteristic vector, wherein the comparison result is different from that of the picture.
In an exemplary embodiment, the comparing the picture to be retrieved with the third feature vector of the picture in operation S2, and using the comparison result of the third feature vector as the comparison result with the picture includes:
when the similarity obtained by comparing the picture to be retrieved with the third feature vector of the picture is greater than a second preset threshold value, the comparison result of the picture to be retrieved with the picture is the same; and when the similarity obtained by comparing the picture to be retrieved with the third characteristic vector of the picture is not more than a second preset threshold value, the comparison result with the picture is different.
In an exemplary embodiment, the determining whether the picture identical to the picture to be retrieved exists in the pre-stored picture set according to the comparison result with each picture in operation S2 includes:
s23, when the comparison result with at least one picture is the same, determining that the picture which is the same as the picture to be retrieved exists in the prestored picture set;
and S23, when the comparison results of the picture and each picture are different, determining that the picture which is the same as the picture to be retrieved does not exist in the prestored picture set.
In one exemplary embodiment, the first feature vector is a picture compression feature vector with spatial position invariance, and the second feature vector is a spatial color information distribution feature vector; and extracting a feature vector for the local feature point by the third feature vector.
Respectively comparing the acquired picture to be retrieved with each picture in a picture set stored in advance, wherein the process of comparing with any one picture in each picture comprises the following steps: when the similarity of the comparison of the picture compression characteristic vectors or the space color information distribution characteristic vectors with space position invariance is larger than a first preset threshold value, extracting the characteristic vectors by using local characteristic points for comparison; extracting a comparison result of the feature vectors by using the local feature points as a comparison result with the picture; when the similarity of the comparison of the picture compression characteristic vector with space position invariance or the space color information distribution characteristic vector is not more than a first preset threshold, extracting the characteristic vector without using a local characteristic point for comparison; and determining that the comparison result of the picture to be retrieved is different from the comparison result of the picture.
When the similarity obtained by comparing the picture to be retrieved with the local feature point extracted feature vector of the picture is greater than a second preset threshold value, the comparison result of the picture to be retrieved is the same as that of the picture; and when the similarity obtained by comparing the picture to be retrieved with the local feature point extracted feature vector of the picture is not more than a second preset threshold, the comparison result with the picture is different.
When the comparison result between the picture to be retrieved and at least one picture in a pre-stored picture set is the same, determining that the picture which is the same as the picture to be retrieved exists in the pre-stored picture set; and when the comparison result with each picture is different, determining that the picture which is the same as the picture to be retrieved does not exist in the prestored picture set.
The image compression characteristics with spatial position invariance can be extracted by adopting hash algorithms such as p hash, d hash, mean hash and the like; the spatial color information distribution can be extracted by adopting HSV and RGB algorithms; the local feature points may be extracted using an ORB algorithm, a SiFT algorithm, a SURF algorithm, or the like.
In an exemplary embodiment, the picture compression feature vector with spatial position invariance of the picture to be retrieved is obtained as follows: and after the picture to be retrieved is zoomed into a preset size and is converted into a gray scale image, acquiring 64-dimensional feature points of the picture to be retrieved by adopting a perceptual hash algorithm to generate the picture compression feature vector with space position invariance.
In an exemplary embodiment, the spatial color information distribution feature vector of the picture to be retrieved is obtained as follows: and (3) the picture to be retrieved is scaled to a preset size, and a spatial color information distribution algorithm is adopted to obtain 72-dimensional feature points of the picture to be retrieved and generate a spatial color information distribution feature vector for use.
In an exemplary embodiment, the local feature point extraction feature vector of the picture to be retrieved is obtained in the following manner: and extracting up to 200 128-dimensional feature points by adopting a local feature point extraction algorithm to generate a local feature point extraction feature vector of the retrieval picture.
The precision of the picture compression characteristic vector and the space color information distribution characteristic vector with the picture having space position invariance in the embodiment of the application is lower than that of the characteristic vector extracted from the local characteristic point when the pictures are compared. When comparing, firstly, the picture compression characteristic vector with space position invariance and the space color information distribution characteristic vector are used for comparing, and whether the characteristic vector needs to be extracted by local characteristic points is determined according to the comparison result. When the image compression characteristic vector adopting the space position invariance or the space color information distribution characteristic vector is compared identically, the local characteristic point is used for extracting the characteristic vector for further confirmation, and when the image compression characteristic vector adopting the space position invariance and the space color information distribution characteristic vector are compared differently, the judgment is directly different. Therefore, the number of times of comparison for extracting feature vectors by using local feature points can be reduced, the speed of comparison can be improved, and the accuracy of comparison can be ensured.
When the similarity comparison is performed on the feature vectors of the pictures, each feature vector may have respective disadvantages, for example, the compressed feature vectors of the pictures with spatial position invariance have the limitation of matching the directions of the pictures, for example, when the directions of the pictures are different, the pictures are all considered to be different. When the judgment of the image compression characteristic vectors with space position invariance is different, the space color information distribution characteristic vectors are used for comparison so as to overcome the existing conditions.
As shown in fig. 2, in the advertisement service, an advertisement publisher needs to determine whether a video or a picture of an advertisement delivered by the advertisement publisher is actually delivered in a mass of pictures from the internet or other channels. Firstly, an advertisement picture set A is collected, and a fused feature vector va is generated for each advertisement picture a, wherein the fused feature vector va comprises a va 1-picture compression feature vector with space position invariance, a va 2-space color information distribution feature vector and a va 3-local feature point extraction feature vector.
The va 1-compressed feature vector of the picture with the spatial position invariance is obtained by scaling each original image in the A into 8 x 9 size pictures, then converting each original image into a gray scale image, and then using the difference hash algorithm (i.e. d hash) in the perceptual hash algorithm to represent each original image as a 64-dimensional vector va1 represented by 0 or 1.
Wherein the va 2-space color information distribution feature vector is generated by scaling all the original images in A into the same size (e.g., 640 x 480 size) and then converting the RGB color space into HSV color space. After dividing the Hue space into 8 parts, the Saturation space into 3 parts, and the Value space into 3 parts, each pixel after scaling is mapped to an 8 × 3 × 3 space, and the original image can be represented as a 72-dimensional vector va2 composed of positive integers.
The va 3-local feature point extraction feature vector is used for verifying the picture retrieval result by extracting at most 200 128-dimensional feature points by using an ORB algorithm to generate a vector va3 for each picture.
For any picture B to be retrieved in the picture set B, the corresponding feature vector vb is obtained according to the method, and the feature vector vb comprises vb1, vb2 and vb 3.
The feature vector vb1 and the feature vector va1 are compared, and the hamming distance is used to measure the similarity, and if the similarity is greater than 0.9, the similarity is considered to be similar. If the degree of similarity is below 0.9, then matching is performed using vb2 and va 2. At this time, if the similarity is greater than 0.9, it is considered similar.
If the similarity is greater than 0.9, then vb3 and va3 are used for matching, and if the matching similarity is greater than 0.9, finally, the picture a and the picture b are considered to be the same picture.
The application provides a picture retrieval device, which comprises a processor and a memory, and is characterized in that the memory stores a program for picture retrieval; the processor is used for reading the program for the picture retrieval content and executing the method of any one of the above items.
The present application provides a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method of any of the above.
It will be understood by those of ordinary skill in the art that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. In a hardware implementation, the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be performed by several physical components in cooperation. Some or all of the components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media as known to those skilled in the art.
Claims (10)
1. An image retrieval method, comprising:
acquiring at least three different feature vectors of a picture to be retrieved; the at least three different feature vectors include: a first feature vector, a second feature vector, a third feature vector; wherein the third feature vector has a higher contrast accuracy than the first and second feature vectors;
respectively comparing the acquired picture to be retrieved with each picture in a prestored picture set, and determining whether the picture which is the same as the picture to be retrieved exists in the prestored picture set according to the comparison result with each picture; wherein the pre-stored picture set comprises the at least three different feature vectors of each picture;
wherein the process of comparing with any one of the pictures comprises: and determining whether to use a third feature vector to continue comparison according to the comparison result of the picture to be retrieved and the first feature vector and/or the second feature vector of the picture.
2. The retrieving method according to claim 1, wherein the first feature vector is a picture compression feature vector with spatial position invariance, and the second feature vector is a spatial color information distribution feature vector; and extracting a feature vector for the local feature point by the third feature vector.
3. The searching method according to claim 1, wherein the determining whether to use a third feature vector for comparison according to the comparison result between the picture to be searched and the first feature vector and/or the second feature vector of the picture comprises:
when the similarity obtained by comparing the picture to be retrieved with the first characteristic vector or the second characteristic vector of the picture is greater than a first preset threshold value, comparing the picture to be retrieved with a third characteristic vector of the picture, and taking the comparison result of the third characteristic vector as the comparison result with the picture; and when the similarity obtained by comparing the first characteristic vector with the second characteristic vector is not more than a first preset threshold, continuing to compare without using the third characteristic vector, wherein the comparison result is different from that of the picture.
4. The retrieval method according to claim 3, wherein the determining whether the picture identical to the picture to be retrieved exists in the pre-stored picture set according to the comparison result with each picture comprises:
when the comparison result with at least one picture is the same, determining that the picture which is the same as the picture to be retrieved exists in the prestored picture set;
and when the comparison result with each picture is different, determining that the picture which is the same as the picture to be retrieved does not exist in the prestored picture set.
5. The searching method according to claim 3, wherein comparing the picture to be searched with the third feature vector of the picture, and taking the comparison result of the third feature vector as the comparison result with the picture comprises:
when the similarity obtained by comparing the picture to be retrieved with the third feature vector of the picture is greater than a second preset threshold value, the comparison result of the picture to be retrieved with the picture is the same; and when the similarity obtained by comparing the picture to be retrieved with the third characteristic vector of the picture is not more than a second preset threshold value, the comparison result with the picture is different.
6. The retrieval method according to claim 2, wherein the picture compression feature vector with spatial position invariance of the picture to be retrieved is obtained as follows:
and after the picture to be retrieved is zoomed into a preset size and is converted into a gray scale image, acquiring 64-dimensional feature points of the picture to be retrieved by adopting a perceptual hash algorithm to generate the picture compression feature vector with space position invariance.
7. The retrieval method according to claim 2, wherein the spatial color information distribution feature vector of the picture to be retrieved is obtained as follows:
and (3) the picture to be retrieved is scaled to a preset size, and a spatial color information distribution algorithm is adopted to obtain 72-dimensional feature points of the picture to be retrieved and generate a spatial color information distribution feature vector for use.
8. The retrieval method according to claim 2, wherein the local feature point extraction feature vector of the picture to be retrieved is obtained by:
and extracting up to 200 128-dimensional feature points by adopting a local feature point extraction algorithm to generate a local feature point extraction feature vector of the retrieval picture.
9. A picture retrieval device comprises a processor and a memory, and is characterized in that the memory stores a program for picture retrieval; the processor is configured to read the program for picture retrieval, and execute the method of any one of claims 1-8.
10. A computer storage medium on which a computer program is stored, which computer program, when being executed by a processor, carries out the method according to any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010291437.0A CN111581416A (en) | 2020-04-14 | 2020-04-14 | Picture retrieval method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010291437.0A CN111581416A (en) | 2020-04-14 | 2020-04-14 | Picture retrieval method, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111581416A true CN111581416A (en) | 2020-08-25 |
Family
ID=72118840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010291437.0A Withdrawn CN111581416A (en) | 2020-04-14 | 2020-04-14 | Picture retrieval method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111581416A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112749668A (en) * | 2021-01-18 | 2021-05-04 | 上海明略人工智能(集团)有限公司 | Target image clustering method and device, electronic equipment and computer readable medium |
-
2020
- 2020-04-14 CN CN202010291437.0A patent/CN111581416A/en not_active Withdrawn
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112749668A (en) * | 2021-01-18 | 2021-05-04 | 上海明略人工智能(集团)有限公司 | Target image clustering method and device, electronic equipment and computer readable medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9036905B2 (en) | Training classifiers for deblurring images | |
| CN107633526B (en) | Image tracking point acquisition method and device and storage medium | |
| JP5180156B2 (en) | System and method for finding picture images in an image collection using localized two-dimensional visual fingerprints | |
| CN110853033B (en) | Video detection method and device based on inter-frame similarity | |
| JP4907938B2 (en) | Method of representing at least one image and group of images, representation of image or group of images, method of comparing images and / or groups of images, method of encoding images or group of images, method of decoding images or sequence of images, code Use of structured data, apparatus for representing an image or group of images, apparatus for comparing images and / or group of images, computer program, system, and computer-readable storage medium | |
| JP5097280B2 (en) | Method and apparatus for representing, comparing and retrieving images and image groups, program, and computer-readable storage medium | |
| US7277584B2 (en) | Form recognition system, form recognition method, program and storage medium | |
| JP5117670B2 (en) | Image and method for representing image group, method for comparing image or image group, method for retrieving image or image group, apparatus and system for executing each method, program, and computer-readable storage medium | |
| US20080267452A1 (en) | Apparatus and method of determining similar image | |
| KR20140102038A (en) | Video matching device and video matching method | |
| US20100158375A1 (en) | Signal processing apparatus, signal processing method, computer-readable medium and computer data signal | |
| US8254690B2 (en) | Information processing apparatus, information processing method, and program | |
| US10699156B2 (en) | Method and a device for image matching | |
| CN110049309B (en) | Method and device for detecting stability of image frame in video stream | |
| CN111783770B (en) | Image correction method, device and computer readable storage medium | |
| CN111914921A (en) | Similarity image retrieval method and system based on multi-feature fusion | |
| CN111581416A (en) | Picture retrieval method, device and storage medium | |
| JP2007199749A (en) | Image retrieval method and image retrieval device | |
| CN112287905A (en) | Vehicle damage identification method, device, equipment and storage medium | |
| US20220414393A1 (en) | Methods and Systems for Generating Composite Image Descriptors | |
| US10360471B2 (en) | Image retrieving device, image retrieving method, and recording medium | |
| KR20130080743A (en) | Method of searching object field by matching the descritor set | |
| JP6468642B2 (en) | Information terminal equipment | |
| Essa et al. | High Order Volumetric Directional Pattern for Video‐Based Face Recognition | |
| CN112396551A (en) | Watermark embedding method and device |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200825 |