CN113568571A - Image de-duplication method based on residual error neural network - Google Patents
Image de-duplication method based on residual error neural network Download PDFInfo
- Publication number
- CN113568571A CN113568571A CN202110722552.3A CN202110722552A CN113568571A CN 113568571 A CN113568571 A CN 113568571A CN 202110722552 A CN202110722552 A CN 202110722552A CN 113568571 A CN113568571 A CN 113568571A
- Authority
- CN
- China
- Prior art keywords
- image
- processed
- client
- characteristic value
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000001514 detection method Methods 0.000 claims abstract description 13
- 230000008447 perception Effects 0.000 claims abstract description 13
- 238000012795 verification Methods 0.000 claims description 43
- 238000007781 pre-processing Methods 0.000 claims description 11
- 238000004891 communication Methods 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 4
- 230000000007 visual effect Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000003993 interaction Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000003997 social interaction Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
- G06F3/0646—Horizontal data movement in storage systems, i.e. moving data in between storage devices or systems
- G06F3/0652—Erasing, e.g. deleting, data cleaning, moving of data to a wastebasket
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an image de-duplication method based on a residual error neural network, which is applied to a server and comprises the following steps: receiving an uploading request; according to the first accurate matching data and the first perception hash characteristic value, similarity detection is carried out on the image to be processed in the database; when a first image matched with the image to be processed is detected in the database, comparing the image quality scores of the image to be processed and the first image; and if the image quality score of the image to be processed is higher than the image quality score of the first image, sending an uploading instruction to the client, and receiving the image to be processed uploaded by the client. According to the invention, the images with the same visual information can be judged as similar images by comparing the first accurate matching data with the first perception hash characteristic value, and the duplicate removal precision is good. In addition, by comparing the image quality of the image to be processed with that of the first image and deleting the redundant image with poor quality, a good deduplication effect can be obtained and the user experience can be improved.
Description
Technical Field
The invention belongs to the field of image processing, and particularly relates to an image de-duplication method based on a residual error neural network.
Background
Images contain rich and intuitive information, and a large number of images are needed to deliver information to users in the fields of social interaction, shopping, tourism and the like. With the increasing number of images, the local storage overhead is increased, and therefore more and more users upload images from the local client to the cloud server for storage.
In order to avoid unnecessary increase of the storage capacity of the cloud server caused by uploading similar or identical images again, how to realize secure deduplication of images from a large amount of data becomes a research hotspot in the field. At present, in the related art, a Client-based Multimedia Data Deduplication method (CSPD) is adopted to achieve image Deduplication, and the method includes three stages of similar image detection, ownership authentication, and quality comparison. However, in the related art, similarity discrimination is performed from two angles of cryptographic hash and discrete cosine transform, so that only an original image and an image with one distortion change can be recognized, and when the original image is subjected to superposition of two or more kinds of distortions, the recognition accuracy of the similar image is reduced, and further the image deduplication precision is reduced.
Disclosure of Invention
In order to solve the above problems in the prior art, the present invention provides an image deduplication method based on a residual error neural network. The technical problem to be solved by the invention is realized by the following technical scheme:
in a first aspect, the present invention provides an image deduplication method based on a residual error neural network, applied to a server, including:
receiving an uploading request from a client; the client side and the server are in communication connection with each other in advance, and the uploading request comprises first accurate matching data and a first perceptual hash characteristic value of the image to be processed;
according to the first accurate matching data and the first perception hash characteristic value, carrying out similarity detection on the image to be processed in a database; the database comprises a plurality of groups of pre-stored images, accurate matching data corresponding to the images and a perception hash characteristic value;
when a first image matching with an image to be processed is detected in the database, comparing the image quality scores of the image to be processed and the first image; and if the image quality score of the image to be processed is higher than the image quality score of the first image, sending an uploading instruction to a client, and receiving the image to be processed uploaded by the client.
In an embodiment of the present invention, the step of performing similarity detection on the to-be-processed image in a database according to the first exact match data and the first perceptual hash feature value includes:
detecting whether second exact match data equal to the first exact match data exist in the exact match data stored in the database;
if the first image exists, determining the image corresponding to the second accurate matching data as a first image; otherwise, calculating a first average absolute error distance between the first perceptual hash characteristic value and each perceptual hash characteristic value stored in the database;
and comparing the first average absolute error distance with a first preset threshold, and determining an image corresponding to the second hash characteristic value as a first image after determining the perceptual hash characteristic value of which the first average absolute error distance is smaller than the preset threshold as the second hash characteristic value.
In an embodiment of the present invention, when a first image matching an image to be processed is detected in the database, the step of comparing the image quality of the image to be processed with that of the first image is preceded by:
and performing ownership verification on the client.
In an embodiment of the present invention, the step of performing ownership verification on the client includes:
after a plurality of image blocks are divided in the first image, selecting a first image block from the first image, and preprocessing the first image block;
the preprocessed first image block and the first image are zoomed to the same size and mixed to obtain a first target image;
inputting the first target image to a first residual error neural network to obtain a perceptual hash characteristic value of the first target image;
generating an ownership verification request, the ownership verification request including network parameters of the first residual neural network;
sending the ownership verification request to a client so that the client can select a second image block from the image to be processed after dividing the image block into a plurality of image blocks, preprocessing the second image block, scaling the preprocessed second image block and the image to be processed to the same size, mixing the preprocessed second image block and the image to be processed to obtain a second target image, setting network parameters of a second residual error neural network according to the network parameters of the first residual error neural network, inputting the second target image to the second residual error neural network, and sending the obtained perceptual hash characteristic value of the second target image to the server;
and judging whether the client passes ownership verification or not according to the perceptual hash characteristic value of the first target image and the perceptual hash characteristic value of the second target image.
In an embodiment of the present invention, the step of determining whether the client passes the ownership verification according to the perceptual hash feature value of the first target image and the perceptual hash feature value of the second target image includes:
calculating a second average absolute error distance between the perceptual hash characteristic value of the first target image and the perceptual hash characteristic value of the second target image;
comparing the second average absolute error distance with a second preset threshold; if the second mean absolute error distance is smaller than a second preset threshold, the client passes ownership verification, otherwise, the client fails ownership verification;
and sending a verification result to the client.
In one embodiment of the invention, the pre-processing comprises a rotation and/or translation operation.
In an embodiment of the present invention, when a first image matching an image to be processed is detected in the database, the step of comparing the image quality scores of the image to be processed and the first image is followed by:
and if the image quality score of the image to be processed is lower than that of the first image, sending the link of the first image to the client.
In an embodiment of the present invention, when there is no first image matching the image to be processed in the database, after the step of performing similarity detection on the image to be processed in the database according to the first exact match data and the first perceptual hash feature value, the method further includes:
and after an uploading instruction is sent to the client, receiving the image to be processed uploaded by the client.
In an embodiment of the present invention, after the step of receiving the image to be processed uploaded by the client, the method further includes:
and storing the image to be processed, the first exact matching data and the first perception hash characteristic value into a database.
In a second aspect, the present invention further provides an image deduplication method based on a residual error neural network, which is applied to a client, and includes:
acquiring an image to be processed;
determining first exact match data of the image to be processed according to a Sha256 algorithm;
determining a first perception hash characteristic value of the image to be processed by utilizing a pre-trained second residual error neural network model;
sending an uploading request of the image to be processed to a server, so that the server performs similarity detection on the image to be processed in a database according to the first accurate matching data and the first perceptual hash characteristic value, and comparing the image quality scores of the image to be processed and a first image when the first image matched with the image to be processed is detected;
if the image quality score of the image to be processed is higher than the image quality score of the first image, receiving an uploading instruction, and uploading the image to be processed to the server; otherwise, a link to the first image is received.
Compared with the prior art, the invention has the beneficial effects that:
in the image deduplication method based on the residual error neural network, the server receives an uploading request from the client, and the uploading request comprises first accurate matching data and a first perceptual hash characteristic value of an image to be processed, so that the same image can be judged by comparing the first accurate matching data with the accurate matching data of the image stored in the database in advance, and the similar image can be judged by comparing the first perceptual hash characteristic value with the perceptual hash characteristic value of the image stored in the database in advance, so that the images with the same visual information can be judged to be similar, and the deduplication precision is good. In addition, when a first image which is the same as or similar to the image to be uploaded exists in the database, the image quality of the first image and the image quality of the second image are compared, and a redundant image with poor quality is deleted, so that good de-duplication effect is obtained, and user experience is improved.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Drawings
Fig. 1 is a schematic flowchart of an image deduplication method based on a residual neural network according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an image deduplication method based on a residual error neural network according to an embodiment of the present invention;
fig. 3 is a schematic diagram of performing ownership verification on a server according to an embodiment of the present invention;
fig. 4 is a schematic diagram of interaction between a client and a server in an ownership verification process provided by an embodiment of the present invention;
fig. 5 is a schematic flowchart of another image deduplication method based on a residual neural network according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to specific examples, but the embodiments of the present invention are not limited thereto.
As shown in fig. 1, the image deduplication method based on the residual neural network provided in the embodiment of the present invention is applied to a server, and includes:
s101, receiving an uploading request from a client; the method comprises the steps that communication connection is established between a client and a server in advance, and an uploading request comprises first accurate matching data and a first perception hash characteristic value of an image to be processed;
s102, carrying out similarity detection on the image to be processed in a database according to the first accurate matching data and the first perception hash characteristic value; the database comprises a plurality of groups of pre-stored images, accurate matching data corresponding to the images and perceptual hash characteristic values;
s103, when a first image matched with the image to be processed is detected in the database, comparing the image quality scores of the image to be processed and the first image; and if the image quality score of the image to be processed is higher than the image quality score of the first image, sending an uploading instruction to the client, and receiving the image to be processed uploaded by the client.
In this embodiment, the server and the client establish a communication connection in advance, and the database of the server includes a plurality of sets of images stored in advance, exact matching data corresponding to the images, and perceptual hash feature values, where the images stored in advance may be uploaded to the server by the client. Optionally, the upload request sent by the client to the server includes a feature sequence of the image to be processed, where the feature sequence includes first exact match data and a first perceptual hash feature value. And after receiving the uploading request, the server searches in the database according to the first accurate matching data and the first perceptual hash characteristics, and detects whether an image similar to the image to be processed exists or not. It should be noted that the image similar to the image to be processed in this embodiment includes both an image having the same encoding method as the image to be processed and a similar image having the same semantic information after the image to be processed is subjected to operations such as rotation, filtering, and compression.
Optionally, the client determines first exact match data of the image to be processed by using a Sha256 algorithm, and determines a first perceptual hash feature value of the image to be processed by using a pre-trained second residual neural network model. In the embodiment, the residual neural network and the Hash technology are introduced into the similarity discrimination of the images, and the images with the same visual information can be judged to be similar, so that the robustness and the distinguishability of the image deduplication are improved.
Of course, in some other embodiments of the present invention, other algorithms may be used to calculate the first exact match data and the first perceptual hash feature value of the image to be processed, which is not limited by the present invention.
In the above step S103, when there is a first image similar to the image to be processed in the database, the image quality scores of the two images are further compared, wherein a higher image quality score indicates better image quality. When the image quality of the image to be processed is better than that of the first image, the image to be processed which is requested to be uploaded by the client side is better in quality, therefore, the server sends an uploading instruction to the client side and receives the image to be processed which is uploaded by the client side, the design method can delete the redundant image with poor quality, and therefore good de-duplication effect is achieved, and user experience is improved.
In addition, in the above step S103, if the image quality score of the image to be processed is lower than the image quality score of the first image, a link of the first image is transmitted to the client so that the user uses the first image according to the link download.
Of course, in this embodiment, if the server does not retrieve the first image, it indicates that the image that is the same as or similar to the image to be processed is not uploaded, so the server can send an upload instruction to the client and receive the image to be processed uploaded by the client.
Optionally, in the step S102, the step of performing similarity detection on the image to be processed in the database according to the first exact match data and the first perceptual hash feature value includes:
detecting whether second accurate matching data equal to the first accurate matching data exists in the accurate matching data stored in the database;
if the first image exists, determining the image corresponding to the second accurate matching data as the first image; otherwise, calculating a first average absolute error distance between the first perceptual hash characteristic value and each perceptual hash characteristic value stored in the database;
and comparing the first average absolute error distance with a first preset threshold, and determining an image corresponding to the second hash characteristic value as a first image after determining the perceptual hash characteristic value of which the first average absolute error distance is smaller than the preset threshold as the second hash characteristic value.
In this embodiment, after receiving the upload request, the server first detects whether there is second exact match data equal to the first exact match data in the exact match data stored in the database; if the first image exists, determining the image corresponding to the second accurate matching data as a first image, wherein the first image and the image to be processed are the same image; on the contrary, it means that the same image as the image to be processed does not exist in the server, at this time, a first average absolute error distance between the first perceptual hash feature value and each perceptual hash feature value stored in the database is further calculated, after the perceptual hash feature value of which the first average absolute error distance is smaller than a preset threshold value is determined as a second perceptual hash feature value, the image corresponding to the second perceptual hash feature value is determined as a first image, and the first image and the image to be processed are similar images.
Optionally, when the first image matching the image to be processed is detected in the database, the step of comparing the image quality of the image to be processed with that of the first image is preceded by the following step:
and carrying out ownership verification on the client.
In this embodiment, when the first image is stored in the server, the ownership of the client is verified. It should be understood that when deduplication is performed according to the feature sequence, if an attacker guesses or illegally acquires the feature sequence of the image to be processed, the attacker may obtain the use right by using the feature sequence and further acquire the first image, which results in data leakage, and therefore it is necessary to perform security authentication on the client.
Fig. 3 is a schematic diagram of performing ownership verification at a server according to an embodiment of the present invention, and fig. 4 is a schematic diagram of interaction between a client and a server in an ownership verification process according to an embodiment of the present invention.
Referring to fig. 3-4, the step of performing ownership verification on the client includes:
after a plurality of image blocks are divided in a first image, selecting the first image block from the first image, and preprocessing the first image block;
the preprocessed first image block and the first image are zoomed to the same size and mixed to obtain a first target image;
inputting a first target image to a first residual error neural network to obtain a perceptual hash characteristic value of the first target image;
generating an ownership verification request, the ownership verification request including network parameters of the first residual neural network;
sending an ownership verification request to a client so that the client can select a second image block from the image to be processed after dividing the image block into a plurality of image blocks, preprocessing the second image block, scaling the preprocessed second image block and the image to be processed to the same size, mixing the preprocessed second image block and the image to be processed to obtain a second target image, setting network parameters of a second residual error neural network according to the network parameters of a first residual error neural network, inputting the second target image to the second residual error neural network, and sending the obtained perceptual hash characteristic value of the second target image to a server;
and judging whether the client passes ownership verification according to the perceptual hash characteristic value of the first target image and the perceptual hash characteristic value of the second target image.
Specifically, as shown in fig. 3, in the process of performing ownership verification on the client, the server first divides the first image into a plurality of image blocks, and randomly selects a first image block from the image blocks; then, preprocessing the first image block, wherein the preprocessing comprises rotation and/or translation operation, and then scaling the preprocessed first image block; and finally, after the unprocessed first image is scaled to the size same as that of the first image block, the unprocessed first image and the first image block are mixed to generate a first target image. Further, as shown in fig. 4, after the server generates the first target image, the first target image is input to the first residual neural network to obtain the perceptual hash feature value of the first target image.
It can be understood that in this embodiment, the ownership verification of the client is performed by comparing the first target image generated by the server and the second target image generated by the client, and the generation process of the second target image is the same as that of the first target image, that is, the client divides the image to be processed into a plurality of image blocks, randomly selects the second image block from the image blocks, performs the same preprocessing operation as that of the first image block on the second image block, scales and mixes the image to be processed and the processed second image block, and then inputs the mixed second target image into the second residual neural network, so as to obtain the perceptual hash feature value of the second target image.
It should be noted that, after the server side completes the calculation of the perceptual hash characteristic value of the first target image, an ownership verification request is generated and sent, the ownership verification request contains the network parameters of the first residual error neural network, and the client side sets the network parameters of the second residual error neural network according to the network parameters of the first residual error neural network.
Further, the client sends the perceptual hash feature value of the second target image to the server, and the server can judge whether the client passes ownership verification according to the perceptual hash feature value of the first target image and the perceptual hash feature value of the second target image.
Optionally, with reference to fig. 4, the step of determining whether the client passes the ownership verification according to the perceptual hash feature value of the first target image and the perceptual hash feature value of the second target image includes:
calculating a second average absolute error distance between the perceptual hash characteristic value of the first target image and the perceptual hash characteristic value of the second target image;
comparing the second average absolute error distance with a second preset threshold; if the second average absolute error distance is smaller than a second preset threshold, the client passes ownership verification, otherwise, the client does not pass ownership verification;
and sending the verification result to the client.
Specifically, the server calculates a second average absolute error distance between a perceptual hash characteristic value of the first target image and a perceptual hash characteristic value of the second target image, and when the second average absolute error distance is smaller than a second preset threshold, the client passes ownership verification; and when the second average absolute error distance is larger than or equal to a second preset threshold, the client side does not pass ownership verification, and the server rejects the uploading request.
Therefore, the invention does not need to introduce a server specially used for ownership verification and set an additional reference image library, and the perceptual hash characteristic value used for ownership verification is not reused, thereby ensuring that the image deduplication method has excellent safety and reliability.
It should be noted that, no matter whether the image to be processed is finally uploaded to the server, the image to be processed, the first exact matching data and the first perceptual hash feature value may be stored in the database, and the updated database is favorable for being used as a basis for similarity determination in the subsequent deduplication process of other images.
As shown in fig. 5, an embodiment of the present invention further provides an image deduplication method based on a residual neural network, which is applied to a client, and includes:
s501, acquiring an image to be processed;
s502, determining first accurate matching data of the image to be processed according to a Sha256 algorithm;
s503, determining a first perception hash characteristic value of the image to be processed by utilizing a pre-trained second residual error neural network model;
s504, sending an uploading request of the image to be processed to a server, so that the server performs similarity detection on the image to be processed in a database according to the first accurate matching data and the first perceptual hash characteristic value, and comparing the image quality scores of the image to be processed and the first image when the first image matched with the image to be processed is detected;
s505, if the image quality score of the image to be processed is higher than that of the first image, receiving an uploading instruction, and uploading the image to be processed to a server;
s506, receiving the link of the first image if the image quality score of the image to be processed is lower than the image quality score of the first image.
In the image deduplication method based on the residual error neural network, the server receives an uploading request from the client, and the uploading request comprises first accurate matching data and a first perceptual hash characteristic value of an image to be processed, so that the same image can be judged by comparing the first accurate matching data with the accurate matching data of the image stored in the database in advance, and the similar image can be judged by comparing the first perceptual hash characteristic value with the perceptual hash characteristic value of the image stored in the database in advance, so that the images with the same visual information can be judged to be similar, and the deduplication precision is good. In addition, when a first image which is the same as or similar to the image to be uploaded exists in the database, the image quality of the first image and the image quality of the second image are compared, and a redundant image with poor quality is deleted, so that good de-duplication effect is obtained, and user experience is improved.
In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples described in this specification can be combined and combined by those skilled in the art.
While the present application has been described in connection with various embodiments, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed application, from a review of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the word "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.
The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.
Claims (10)
1. An image deduplication method based on a residual error neural network is applied to a server and comprises the following steps:
receiving an uploading request from a client; the client side and the server are in communication connection with each other in advance, and the uploading request comprises first accurate matching data and a first perceptual hash characteristic value of the image to be processed;
according to the first accurate matching data and the first perception hash characteristic value, carrying out similarity detection on the image to be processed in a database; the database comprises a plurality of groups of pre-stored images, accurate matching data corresponding to the images and a perception hash characteristic value;
when a first image matching with an image to be processed is detected in the database, comparing the image quality scores of the image to be processed and the first image; and if the image quality score of the image to be processed is higher than the image quality score of the first image, sending an uploading instruction to a client, and receiving the image to be processed uploaded by the client.
2. The residual neural network-based image deduplication method of claim 1, wherein the step of performing similarity detection on the to-be-processed image in a database according to the first exact match data and the first perceptual hash feature value comprises:
detecting whether second exact match data equal to the first exact match data exist in the exact match data stored in the database;
if the first image exists, determining the image corresponding to the second accurate matching data as a first image; otherwise, calculating a first average absolute error distance between the first perceptual hash characteristic value and each perceptual hash characteristic value stored in the database;
and comparing the first average absolute error distance with a first preset threshold, and determining an image corresponding to the second hash characteristic value as a first image after determining the perceptual hash characteristic value of which the first average absolute error distance is smaller than the preset threshold as the second hash characteristic value.
3. The residual neural network-based image deduplication method of claim 2, wherein when a first image matching a to-be-processed image is detected in the database, the step of comparing the image quality of the to-be-processed image with that of the first image is preceded by the step of:
and performing ownership verification on the client.
4. The residual neural network-based image deduplication method of claim 3, wherein the step of performing ownership verification on the client comprises:
after a plurality of image blocks are divided in the first image, selecting a first image block from the first image, and preprocessing the first image block;
the preprocessed first image block and the first image are zoomed to the same size and mixed to obtain a first target image;
inputting the first target image to a first residual error neural network to obtain a perceptual hash characteristic value of the first target image;
generating an ownership verification request, the ownership verification request including network parameters of the first residual neural network;
sending the ownership verification request to a client so that the client can select a second image block from the image to be processed after dividing the image block into a plurality of image blocks, preprocessing the second image block, scaling the preprocessed second image block and the image to be processed to the same size, mixing the preprocessed second image block and the image to be processed to obtain a second target image, setting network parameters of a second residual error neural network according to the network parameters of the first residual error neural network, inputting the second target image to the second residual error neural network, and sending the obtained perceptual hash characteristic value of the second target image to the server;
and judging whether the client passes ownership verification or not according to the perceptual hash characteristic value of the first target image and the perceptual hash characteristic value of the second target image.
5. The residual neural network-based image deduplication method of claim 4, wherein the step of determining whether the client passes ownership verification according to the perceptual hash feature value of the first target image and the perceptual hash feature value of the second target image comprises:
calculating a second average absolute error distance between the perceptual hash characteristic value of the first target image and the perceptual hash characteristic value of the second target image;
comparing the second average absolute error distance with a second preset threshold; if the second mean absolute error distance is smaller than a second preset threshold, the client passes ownership verification, otherwise, the client fails ownership verification;
and sending a verification result to the client.
6. The residual neural network-based image deduplication method of claim 4, wherein the preprocessing comprises rotation and/or translation operations.
7. The residual neural network-based image deduplication method of claim 1, wherein when a first image matching a to-be-processed image is detected in the database, the step of comparing the to-be-processed image with an image quality score of the first image is followed by further comprising:
and if the image quality score of the image to be processed is lower than that of the first image, sending the link of the first image to the client.
8. The residual neural network-based image deduplication method of claim 1, wherein when there is no first image matching the image to be processed in the database, the step of performing similarity detection on the image to be processed in the database according to the first exact match data and the first perceptual hash feature value further comprises:
and after an uploading instruction is sent to the client, receiving the image to be processed uploaded by the client.
9. The residual neural network-based image deduplication method according to claim 1 or 8, wherein after the step of receiving the to-be-processed image uploaded by the client, the method further comprises:
and storing the image to be processed, the first exact matching data and the first perception hash characteristic value into a database.
10. An image deduplication method based on a residual error neural network is applied to a client side, and comprises the following steps:
acquiring an image to be processed;
determining first exact match data of the image to be processed according to a Sha256 algorithm;
determining a first perception hash characteristic value of the image to be processed by utilizing a pre-trained second residual error neural network model;
sending an uploading request of the image to be processed to a server, so that the server performs similarity detection on the image to be processed in a database according to the first accurate matching data and the first perceptual hash characteristic value, and comparing the image quality scores of the image to be processed and a first image when the first image matched with the image to be processed is detected;
if the image quality score of the image to be processed is higher than the image quality score of the first image, receiving an uploading instruction, and uploading the image to be processed to the server; otherwise, a link to the first image is received.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110722552.3A CN113568571B (en) | 2021-06-28 | 2021-06-28 | Image de-duplication method based on residual neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110722552.3A CN113568571B (en) | 2021-06-28 | 2021-06-28 | Image de-duplication method based on residual neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113568571A true CN113568571A (en) | 2021-10-29 |
| CN113568571B CN113568571B (en) | 2024-06-04 |
Family
ID=78162915
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110722552.3A Active CN113568571B (en) | 2021-06-28 | 2021-06-28 | Image de-duplication method based on residual neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113568571B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115225278A (en) * | 2022-06-07 | 2022-10-21 | 西安电子科技大学 | Data-perception-hash-based data possession proving method |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016177259A1 (en) * | 2015-05-07 | 2016-11-10 | 阿里巴巴集团控股有限公司 | Similar image recognition method and device |
| CN108021457A (en) * | 2017-10-31 | 2018-05-11 | 阿里巴巴集团控股有限公司 | Data migration method and device |
| CN108055121A (en) * | 2017-10-23 | 2018-05-18 | 北京邮电大学 | The encryption method and decryption method of image |
| CN108182367A (en) * | 2017-12-15 | 2018-06-19 | 西安电子科技大学 | A kind of encrypted data chunk client De-weight method for supporting data update |
| CN109472690A (en) * | 2018-10-25 | 2019-03-15 | 深圳壹账通智能科技有限公司 | The loan measures and procedures for the examination and approval, device, storage medium and electronic equipment based on block chain |
| CN110321447A (en) * | 2019-07-08 | 2019-10-11 | 北京字节跳动网络技术有限公司 | Determination method, apparatus, electronic equipment and the storage medium of multiimage |
| CN111310129A (en) * | 2020-02-20 | 2020-06-19 | 北京海益同展信息科技有限公司 | Method, device and storage medium for maintaining right of image |
| CN111553259A (en) * | 2020-04-26 | 2020-08-18 | 北京宙心科技有限公司 | Image duplicate removal method and system |
| US20210075788A1 (en) * | 2019-09-11 | 2021-03-11 | Jumio Corporation | Identification document database |
| CN112507159A (en) * | 2020-11-20 | 2021-03-16 | 有米科技股份有限公司 | Hash network training method, advertisement image material retrieval method and related device |
-
2021
- 2021-06-28 CN CN202110722552.3A patent/CN113568571B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016177259A1 (en) * | 2015-05-07 | 2016-11-10 | 阿里巴巴集团控股有限公司 | Similar image recognition method and device |
| CN108055121A (en) * | 2017-10-23 | 2018-05-18 | 北京邮电大学 | The encryption method and decryption method of image |
| CN108021457A (en) * | 2017-10-31 | 2018-05-11 | 阿里巴巴集团控股有限公司 | Data migration method and device |
| CN108182367A (en) * | 2017-12-15 | 2018-06-19 | 西安电子科技大学 | A kind of encrypted data chunk client De-weight method for supporting data update |
| CN109472690A (en) * | 2018-10-25 | 2019-03-15 | 深圳壹账通智能科技有限公司 | The loan measures and procedures for the examination and approval, device, storage medium and electronic equipment based on block chain |
| CN110321447A (en) * | 2019-07-08 | 2019-10-11 | 北京字节跳动网络技术有限公司 | Determination method, apparatus, electronic equipment and the storage medium of multiimage |
| US20210075788A1 (en) * | 2019-09-11 | 2021-03-11 | Jumio Corporation | Identification document database |
| CN111310129A (en) * | 2020-02-20 | 2020-06-19 | 北京海益同展信息科技有限公司 | Method, device and storage medium for maintaining right of image |
| CN111553259A (en) * | 2020-04-26 | 2020-08-18 | 北京宙心科技有限公司 | Image duplicate removal method and system |
| CN112507159A (en) * | 2020-11-20 | 2021-03-16 | 有米科技股份有限公司 | Hash network training method, advertisement image material retrieval method and related device |
Non-Patent Citations (1)
| Title |
|---|
| 王宏达;: "一种基于混沌系统的新型图像加密算法", 光学技术, no. 03, 15 May 2017 (2017-05-15), pages 260 - 266 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115225278A (en) * | 2022-06-07 | 2022-10-21 | 西安电子科技大学 | Data-perception-hash-based data possession proving method |
| CN115225278B (en) * | 2022-06-07 | 2024-06-04 | 西安电子科技大学 | Data-holding proving method based on data-aware hash |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113568571B (en) | 2024-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8837769B2 (en) | Video signature based on image hashing and shot detection | |
| US11727053B2 (en) | Entity recognition from an image | |
| JP6190041B2 (en) | Safety confirmation system and concealed data similarity search method | |
| US11409789B2 (en) | Determining identity in an image that has multiple people | |
| CN109766707B (en) | Data processing method, device, equipment and medium based on block chain | |
| CN110019640B (en) | Secret-related file checking method and device | |
| WO2021142072A1 (en) | Peceptual video fingerprinting | |
| CN103415856A (en) | Device and method for online storage, transmission device and method, and receiving device and method | |
| US12081827B2 (en) | Determining video provenance utilizing deep learning | |
| CN102891751A (en) | Method and equipment for generating business code from fingerprint image | |
| CN113568571B (en) | Image de-duplication method based on residual neural network | |
| US8913851B1 (en) | Fingerprinting image using points of interest for robust image identification | |
| CN111680181A (en) | Abnormal object identification method and terminal equipment | |
| CN112860933B (en) | Ciphertext image retrieval method, device, terminal equipment and storage medium | |
| CN106682130B (en) | Similar picture detection method and device | |
| CN110737633A (en) | resource management method and system based on cloud management platform | |
| CN113688762B (en) | Face recognition method, device, equipment and medium based on deep learning | |
| WO2019002556A1 (en) | Online user verification without prior knowledge of the user | |
| Gao et al. | Similarity-based Secure Deduplication for IIoT Cloud Management System | |
| CN113609513B (en) | File signature verification method based on signature code | |
| CN115225278B (en) | Data-holding proving method based on data-aware hash | |
| CN111382286B (en) | Data processing method and related product | |
| CN117688265B (en) | Cross-platform information integration and collection system | |
| CN116884042A (en) | Fingerprint identification method and device | |
| CN118364461A (en) | File detection method, device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |