CN109712092A - Archives scan image repair method, device and electronic equipment - Google Patents
Archives scan image repair method, device and electronic equipment Download PDFInfo
- Publication number
- CN109712092A CN109712092A CN201811559084.7A CN201811559084A CN109712092A CN 109712092 A CN109712092 A CN 109712092A CN 201811559084 A CN201811559084 A CN 201811559084A CN 109712092 A CN109712092 A CN 109712092A
- Authority
- CN
- China
- Prior art keywords
- scan image
- image
- archives scan
- archives
- machine learning
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of archives scan image repair method, device and electronic equipments, which comprises obtains archives scan image；Based on the gray value of archives scan image, judge whether archives scan image is blurred picture；If so, repairing according to the gray value of archives scan image to archives scan image based on the machine learning model after training, clearly archival image is obtained.It solves the low technical problem of the reparation precision of archives scan image in the prior art, has reached the technical effect of the reparation precision for the gray scale for improving archives scan image.
Description
Technical field
The present invention relates to field of image processings, in particular to a kind of archives scan image repair method, device and electricity
Sub- equipment.
Background technique
In archival digitalization process often due to artificial, scan input device or archives original part itself
The reason of, cause archives scan image generate ghost image, content Local Damaged, it is out of focus fuzzy situations such as.
In the prior art, the method for repairing blurred picture, which is mainly based upon, is filtered image, by constructing fuzzy graph
The core of picture repairs blurred picture to repair blurred picture, or based on the supervised learning method for fighting network.Archives scan image
Situations such as line thickness in image being usually present is different, image is fuzzy, image texture characteristic is few, present Fuzzy image repair
Method cannot keep all having high-precision repairing effect, thus current blurred picture to the blurred picture of these situations simultaneously
Restorative procedure is low to the reparation precision of archives scan image.
Summary of the invention
The purpose of the present invention is to provide a kind of archives scan image repair method, device and electronic equipments, are intended to mention
The reparation precision of high blurred picture reparation.
In a first aspect, the embodiment of the invention provides a kind of archives scan image repair methods, comprising: obtain archives scan
Image；Based on the gray value of the archives scan image, judge whether the archives scan image is blurred picture；If so, base
Machine learning model after training repairs the archives scan image according to the gray value of the archives scan image
It is multiple, obtain clearly archival image.
Optionally, in the gray scale based on the archives scan image, judge whether the archives scan image is mould
Before pasting image, the method also includes:
The archives scan image is pre-processed.
Optionally, described that the archives scan image is pre-processed specifically: to convert the archives scan image
At gray level image.
Optionally, the gray scale based on the archives scan image judges whether the archives scan image is fuzzy
Image, comprising:
Based on the gray level image, the feature of the archives scan image is extracted, obtains characteristic image；
Based on the characteristic image, judge whether the archives scan image is blurred picture.
It is optionally, described to judge whether the archives scan image is blurred picture based on the characteristic image, comprising:
Obtain the gray value of each pixel of the characteristic image；
Based on the gray value of each pixel, the variance of the gray value of the archives scan image is obtained；
If the variance of the gray value is less than or equal to given threshold, determine that the archives scan image is fuzzy graph
Picture.
Optionally, for repairing the training method of the machine learning model of image, comprising:
Multiple original blurred pictures and multiple original clear images are inputted in machine learning models, the machine learning mould
Type is directed to multiple described original blurred pictures and multiple described original clear images, exports multiple first generation clear images respectively
With multiple the first generation blurred pictures；
It will multiple described first generation clear images and multiple described first described machine learning of generation blurred pictures input
In model, the machine learning model is for multiple described first generation clear images and multiple described first generation fuzzy graphs
Picture exports multiple second generation blurred pictures and multiple second generation clear images respectively；
Based on multiple described first generation clear images and multiple described original clear images, first-loss value is obtained；
Based on multiple described first generation blurred pictures and multiple described original blurred pictures, the second penalty values are obtained；
Based on multiple described second generation clear images, multiple described original clear images, multiple described second generation moulds
Image and multiple described original blurred pictures are pasted, third penalty values are obtained；
It imposes a condition if the first-loss value, second penalty values and the third penalty values meet, deconditioning
The machine learning model, the machine learning model after being trained, the machine learning model after the training will be for that will obscure
Image reverts to clear image.
Optionally, the training method of the machine learning model after the training further include:
If the first-loss value, second penalty values and the third penalty values are unsatisfactory for imposing a condition, institute is adjusted
The training weight of machine learning model is stated, so that the weight phase of the first encoder of the machine learning model and second encoder
Together and the weight of the first generator of the machine learning model and the second generator is identical, and training is adjusted trained power
The machine learning model after weight.
Second aspect, the embodiment of the invention provides a kind of archives scan image fixing apparatus, comprising:
Module is obtained, for obtaining archives scan image；
Processing module, for the gray value based on the archives scan image, judge the archives scan image whether be
Blurred picture；If so, based on the machine learning model after training, according to the gray value of the archives scan image to the archives
Scan image is repaired, and clearly archival image is obtained.
The third aspect, the embodiment of the invention provides a kind of computer readable storage mediums, are stored thereon with computer journey
Sequence, when which is executed by processor the step of realization any of the above-described the method.
Fourth aspect, the embodiment of the invention provides a kind of electronic equipment, which is characterized in that including memory, processor
And the computer program that can be run on a memory and on a processor is stored, the processor is realized when executing described program
The step of stating any one the method.
Compared with the prior art, the invention has the following advantages:
The embodiment of the invention provides a kind of archives scan image repair method, device and electronic equipment, the method packets
It includes: obtaining archives scan image；Based on the gray value of archives scan image, judge whether archives scan image is blurred picture；
If so, archives scan image is repaired according to the gray value of archives scan image based on the machine learning model after training,
Obtain clearly archival image.Judge whether archives scan image is blurred picture by the gray scale based on archives scan image,
For that may have, line thickness is different, image is fuzzy, whether the archives scan image of image texture characteristic is blurred picture
Judgement it is accurate；The gray scale of image can characterize the characteristic of image, according to the gray value of archives scan image, to archives scan figure
As being repaired, the repairing effect of the gray scale of archives scan image can be improved；Archives scan is based on by machine learning model
The gray scale of image repairs archives scan image, improves the reparation precision of the gray scale of archives scan image.It solves existing
The technical problem for having the reparation precision of archives scan image in technology low, has reached repairing for the gray scale for improving archives scan image
The technical effect of multiple precision.
Other feature and advantage of the embodiment of the present invention will illustrate in subsequent specification, also, partly from specification
In become apparent, or by implement understanding of the embodiment of the present invention.The objectives and other advantages of the invention can be by institute
Specifically noted structure is achieved and obtained in specification, claims and the attached drawing write.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows a kind of flow chart of archives scan image repair method provided in an embodiment of the present invention.
Fig. 2 shows the flow charts of another archives scan image repair method provided in an embodiment of the present invention.
Fig. 3 shows a kind of frame structure signal of archives scan image fixing apparatus 200 provided in an embodiment of the present invention
Figure.
Fig. 4 shows the frame structure schematic diagram of a kind of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
Exemplary embodiments of the present disclosure are described in more detail below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
It is limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
It is fully disclosed to those skilled in the art.
The embodiment of the invention provides the embodiment of the invention provides a kind of archives scan image repair method, device and electricity
Sub- equipment, the low technical problem of reparation precision to solve archives scan image in the prior art.
Embodiment
A kind of archives scan image repair method provided in an embodiment of the present invention, including S100~S400 as shown in Figure 1,
S100~S400 is illustrated below in conjunction with Fig. 1.
S100: archives scan image is obtained.
S200: the gray value based on archives scan image judges whether archives scan image is blurred picture.
S300: if archives scan image is blurred picture, based on the machine learning model after training, according to archives scan figure
The gray value of picture repairs archives scan image, obtains clearly archival image.
In embodiments of the present invention, archives scan image refers to the image that acquisition is scanned to archives of paper quality.It is scanning
During archives are to obtain archives scan image, often due to manual operation when paper place inclination, scan input device
Scanning accuracy is low or the reasons such as the degeneration of archives original part itself, cause archives scan image occur ghost image, content part by
Situations such as damaging, is out of focus fuzzy causes archives scan image fuzzy, reduces the quality of archives scan image.However, for passing through
It the methods of improves the precision of manual operation, improve the scanning accuracy of scan input device or prevents archives original part itself to degenerate
Improve archives scan image quality, cost be it is expensive, implement also extremely difficult.It therefore, can be by using base
In the image repair method of electronic equipment, the archives scan image for occurring fuzzy is repaired, obtains clearly archival image.
In the prior art, mainly fuzzy image is repaired by the deep learning method based on supervision, this method mainly according to
Lai Yu fuzzy image has feature apparent, that type is single, for example, the method pixel based on deep learning is to pixel
It (pix2pix) is reparation algorithm under supervised learning scene, which relies on a large amount of contents and match one by one, however, for multiple
For archives scan image, not the feature of all files scan image is all clear, thus in the method based on deep learning from
The image repair mode of pixel to pixel (pix2pix) cannot be to the reparation precision of the unsharp archives scan image of feature
It is low.
In embodiments of the present invention before being repaired to archives scan image, first judge archives scan image whether be
Blurred picture.Specifically, judging whether archives scan image is blurred picture by S200.
First archives scan image is pre-processed before executing S200 as a kind of optional real-time mode.Specifically
, it pre-processes the specific can be that archives scan image is converted into gray level image, is filtered archives scan image
Deng.
As an alternative embodiment, S200 includes S200-1 and S200-2 shown in Fig. 2, S200-1: it is based on
Gray level image extracts the feature of archives scan image, obtains characteristic image.S200-2: it is based on characteristic image, judges archives scan
Whether image is blurred picture.
Wherein, for S200-1, specifically: shelves are obtained in gray level image by Laplace operator (Laplacian)
The feature of case scan image is filtered gray level image specifically by Laplace operator, obtains archives scan figure
The characteristic image of picture.In this way, the feature spy obtained based on gray level image is into the gray feature including archives scan image.Pass through
Laplace operator is filtered gray level image, and detection speed is fast, accuracy is high, controllability is good, improves archives
The accuracy of the characteristic image of scan image.It is to be understood that extracting the feature of archives scan image and unlimited in S200-1
Due to the feature for obtaining archives scan image in gray level image by Laplace operator (Laplacian), can also pass through
Luo Baici operator (Roberts), Canny operator obtain the feature of archives scan image.
As an alternative embodiment, S200-2 specifically: the gray value of each pixel of characteristic image is obtained,
Based on the gray value of each pixel, obtain the variance of the gray value of archives scan image, if the variance of gray value be less than or
Equal to given threshold, then determine archives scan image for blurred picture.Based on the gray value of each pixel, archives scan is obtained
The variance of the gray scale of image specifically: the variance for calculating the gray value of the pixel in characteristic image, using the variance as archives
The variance of the gray scale of scan image.For example, characteristic image includes multiple pixels, for example, given threshold is 1000, archives scan
The characteristic image of image includes 4 pixels, and the gray value of 4 pixels is 150,130,0 and 255 respectively, mean value k=
[(150+130+0+255)/4], k=134, wherein [(150+130+0+255)/4] indicate to (150+130+0+255)/4 into
Row rounds up, then the variance of the gray value of the pixel in characteristic imageS is greater than 1000, then sweeps archives
Tracing is as being determined as clear image.If the variance S=100 of the gray value of the pixel in the characteristic image of archives scan image,
100 less than 1000, then are blurred picture by archives scan spectral discrimination.If the pixel in the characteristic image of archives scan image
Gray value variance S=1000,1000 are equal to 1000, then are blurred picture by archives scan spectral discrimination.
By using above scheme, independent of spies such as lines, color pixel values, image textures in archives scan image
Sign, the gray value of the pixel based on image judge whether archives scan image is blurred picture, high reliablity, strong applicability.
If archives scan image is blurred picture, need to restore archives scan image, to obtain clearly archives
Image.In embodiments of the present invention, fuzzy archives scan image is restored by S300.
For S300, by the machine after the gray value input training of fuzzy archives scan image and archives scan image
In learning model, machine learning model repairs archives scan image, exports clearly archival image.Implement in the present invention
Example in, will obscure archives scan image input training after machine learning model in front of, need training machine learn mould
Type.
As an alternative embodiment, the training method of the machine learning model for repairing image specifically: will
In multiple original blurred pictures and multiple original clear image input machine learning models, machine learning model is original for multiple
Blurred picture and multiple original clear images export multiple first generation clear images and multiple first generation fuzzy graphs respectively
Picture；Multiple first generation clear images and multiple first generation blurred pictures are inputted in described machine learning models, the machine
Device learning model exports multiple second generations for multiple first generation clear images and multiple first generation blurred pictures respectively
Blurred picture and multiple second generation clear images；Based on multiple the first generation clear images and multiple original clear images, obtain
Obtain first-loss value；Based on multiple the first generation blurred pictures and multiple original blurred pictures, the second penalty values are obtained；Based on more
It opens second and generates clear image, multiple original clear images, multiple the second generation blurred pictures and multiple original blurred pictures, obtain
Obtain third penalty values；It imposes a condition if first-loss value, the second penalty values and third penalty values meet, machine described in deconditioning
Learning model, the machine learning model after being trained, the machine learning model after training are used for blur ed image restoration Cheng Qing
Clear image.
As an alternative embodiment, machine learning model is that circulation generates confrontation network model
(cycleGenerative Adversarial Network,cycleGAN)。
As an alternative embodiment, shown in the specific calculation such as formula (1) of first-loss value:
LGY(GY,DY, X, Y) and=EY~Y[logDY(y)+EX~X[log(1-DY(GY(x)))] (1)
Wherein, LGY(GY,DY, X, Y) and indicate first-loss value, DY(y) indicate that circulation generates the differentiation in confrontation network model
Device D is to the differentiation of certain original clear image y as a result, EY~Y[logDY(y) indicate original to every in clear image set Y
The differentiation result D of clear image yY(y) mean value of logarithm.GY(x) indicate that circulation generates the generator G in confrontation network model
Certain original blurred picture x is converted into corresponding with original blurred picture x first and generates clear image, DY(GY(x)) it indicates to sentence
Other device D generates clear image G to firstY(x) differentiation result.
As an alternative embodiment, shown in the specific calculation such as formula (2) of the second penalty values:
LGX(GY,DY, X, Y) and=EX~X[logDX(x)+EY~Y[log(1-DX(GX(y)))] (2)
Wherein, LGX(GY,DY, X, Y) and indicate the second penalty values, DX(x) indicate arbiter D to certain original blurred picture x's
Differentiate as a result, GX(y) the first blurred picture that generator is converted to certain original clear image y, D are indicatedX(GX(y)) table
Show arbiter D to the first blurred picture GX(y) differentiation result.
As an alternative embodiment, shown in the specific calculation such as formula (3) of third penalty values:
Lcyc(GY,GX, X, Y) and=EY~Y[||(GY(GX(y)))-y||1]+EX~X[||(GX(GY(x)))-x||1 (3)
Wherein, Lcyc(GY,GX, X, Y) and indicate third penalty values, GY(GX(y)) it indicates to convert certain the first blurred picture
The second obtained clear image.GX(GY(x)) the second blurred picture for being converted into certain the first clear image is indicated.By obtaining
Obtain the second clear image GY(GX(y)) mould of the difference between original clear image y | | (GY(GX(y)))-y||1, then based on institute
It states every that mould gets a distinct image in set Y and clearly schemes corresponding mould and get a distinct image the mean value E of setY~Y[||(GY
(GX(y)))-y||1], likewise, GYIt (x) is to be based on certain original blurred picture x being converted into the first clear image, (GX(GY
It (x)) is) by the first clear image GY(x) the second blurred picture being converted into.
By using above scheme, by the difference and the second clear image that are based on the second blurred picture and original blurred picture
And the difference of original clear image, obtain for determines circulation generate confrontation network model to original blurred picture and it is original clearly
The third penalty values of the precision size of the conversion of image, third penalty values can completely portray circulation and generate confrontation network model
Performance.
As an alternative embodiment, if first-loss value, the second penalty values and third penalty values meet setting item
Part, machine learning model described in deconditioning, the machine learning model after being trained, specifically: when first-loss value, second
When penalty values and third penalty values converge to steady, indicate that machine learning model is trained to stable, i.e. machine learning model
Original blurred picture is converted into the first clear image and reaches the precision that original clear image is converted into the first blurred picture
To requirement, therefore, deconditioning machine learning model, at this time train after machine learning model can by blur ed image restoration at
Clear image.
As an alternative embodiment, if first-loss value, the second penalty values and third penalty values are unsatisfactory for setting
Condition adjusts the training weight of machine learning model, so that the first encoder of machine learning model and the power of second encoder
Heavy phase is same and the first generator of machine learning model and the weight of the second generator it is identical, training is adjusted trained power
Machine learning model after weight imposes a condition until first-loss value, the second penalty values and third penalty values meet.
The embodiment of the invention provides a kind of archives scan image repair method the described method includes: obtaining archives scan figure
Picture；Based on the gray value of archives scan image, judge whether archives scan image is blurred picture；If so, after based on training
Machine learning model repairs archives scan image according to the gray value of archives scan image, obtains clearly archives figure
Picture.Judge whether archives scan image is blurred picture by the gray scale based on archives scan image, for that there may be lines
Width is different, image is fuzzy, the archives scan image of the characteristic of image texture whether be blurred picture judgement it is accurate；Image
Gray scale can characterize the characteristic of image, according to the gray scale of archives scan image, repair, can be improved to archives scan image
The repairing effect of the gray scale of archives scan image；Gray scale by machine learning model based on archives scan image is to archives scan
Image is repaired, and the reparation precision of the gray scale of archives scan image is improved.Solves archives scan image in the prior art
The low technical problem of reparation precision, reached the technical effect of the reparation precision for the gray scale for improving archives scan image.
A kind of archives scan image repair method is provided for above-described embodiment, the embodiment of the present application also correspondence provides one kind
For executing the executing subject of above-mentioned step, which can be archives scan image prosthetic device 200 in Fig. 3.Please
With reference to Fig. 3, which includes:
Module 210 is obtained, for obtaining archives scan image；
Whether processing module 220 judges the archives scan image for the gray value based on the archives scan image
It is blurred picture；If so, based on the machine learning model after training, according to the gray value of the archives scan image to the shelves
Case scan image is repaired, and clearly archival image is obtained.
As an alternative embodiment, the processing module 220 is also used to: being carried out to the archives scan image pre-
Processing.
As an alternative embodiment, the processing module 220 is specifically also used to: the archives scan image is turned
It is melted into gray level image.
As an alternative embodiment, the processing module 220 is specifically also used to: being based on the gray level image, mention
The feature of the archives scan image is taken, characteristic image is obtained；Based on the characteristic image, judge that the archives scan image is
No is blurred picture.
As an alternative embodiment, the processing module 220 is specifically also used to: obtaining the every of the characteristic image
The gray value of a pixel；Based on the gray value of each pixel, the side of the gray value of the archives scan image is obtained
Difference；If the variance of the gray value is less than or equal to given threshold, determine that the archives scan image is blurred picture.
As an alternative embodiment, the processing module 220 is specifically also used to: training the machine for repairing image
Device learning model, specifically: multiple original blurred pictures and multiple original clear images are inputted in machine learning models, it is described
Machine learning model is directed to multiple described original blurred pictures and multiple described original clear images, exports multiple first lifes respectively
At clear image and multiple first generation blurred pictures；It will multiple described first generation clear images and multiple described first generations
Blurred picture inputs in the machine learning model, the machine learning model for it is described multiple first generate clear images and
Multiple described first generation blurred pictures, export multiple second generation blurred pictures and multiple second generation clear images respectively；
Based on multiple described first generation clear images and multiple described original clear images, first-loss value is obtained；Based on described more
It opens first and generates blurred picture and multiple described original blurred pictures, obtain the second penalty values；Based on multiple described second generations
Clear image, multiple described original clear images, multiple described second generation blurred pictures and multiple described original blurred pictures,
Obtain third penalty values；It imposes a condition, stops if the first-loss value, second penalty values and the third penalty values meet
The machine learning model, the machine learning model after being trained only are trained, the machine learning model after the training is used for
By blur ed image restoration at clear image.
As an alternative embodiment, the processing module 220 is specifically also used to: if the first-loss value, institute
It states the second penalty values and the third penalty values is unsatisfactory for imposing a condition, adjust the training weight of the machine learning model, with
Keep the first encoder of the machine learning model identical with the weight of second encoder and the machine learning model
The weight of one generator and the second generator is identical, and training is adjusted the machine learning model after trained weight.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
The embodiment of the invention also provides a kind of electronic equipment, as shown in figure 4, include memory 504, processor 502 and
It is stored in the computer program that can be run on memory 504 and on processor 502, the processor 502 executes described program
The step of either Shi Shixian archives scan image repair method described previously method.
Wherein, in Fig. 4, bus architecture (is represented) with bus 500, and bus 500 may include any number of interconnection
Bus and bridge, bus 500 will include the one or more processors represented by processor 502 and what memory 504 represented deposits
The various circuits of reservoir link together.Bus 500 can also will peripheral equipment, voltage-stablizer and management circuit etc. it
Various other circuits of class link together, and these are all it is known in the art, therefore, no longer further retouch to it herein
It states.Bus interface 505 provides interface between bus 500 and receiver 501 and transmitter 503.Receiver 501 and transmitter
503 can be the same element, i.e. transceiver, provide the unit for communicating over a transmission medium with various other devices.Place
It manages device 502 and is responsible for management bus 500 and common processing, and memory 504 can be used for storage processor 502 and execute behaviour
Used data when making.
The embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer program, the journey
The step of either archives scan image repair method described previously method is realized when sequence is executed by processor.
Algorithm and display are not inherently related to any particular computer, virtual system, or other device provided herein.
Various general-purpose systems can also be used together with teachings based herein.As described above, it constructs required by this kind of system
Structure be obvious.In addition, the present invention is also not directed to any particular programming language.It should be understood that can use various
Programming language realizes summary of the invention described herein, and the description done above to language-specific is to disclose this hair
Bright preferred forms.
In the instructions provided here, numerous specific details are set forth.It is to be appreciated, however, that implementation of the invention
Example can be practiced without these specific details.In some instances, well known method, structure is not been shown in detail
And technology, so as not to obscure the understanding of this specification.
Similarly, it should be understood that in order to simplify the disclosure and help to understand one or more of the various inventive aspects,
Above in the description of exemplary embodiment of the present invention, each feature of the invention is grouped together into single implementation sometimes
In example, figure or descriptions thereof.However, the disclosed method should not be interpreted as reflecting the following intention: i.e. required to protect
Shield the present invention claims features more more than feature expressly recited in each claim.More precisely, as following
Claims reflect as, inventive aspect is all features less than single embodiment disclosed above.Therefore,
Thus the claims for following specific embodiment are expressly incorporated in the specific embodiment, wherein each claim itself
All as a separate embodiment of the present invention.
Those skilled in the art will understand that can be carried out adaptively to the module in the equipment in embodiment
Change and they are arranged in one or more devices different from this embodiment.It can be the module or list in embodiment
Member or component are combined into a module or unit or component, and furthermore they can be divided into multiple submodule or subelement or
Sub-component.Other than such feature and/or at least some of process or unit exclude each other, it can use any
Combination is to all features disclosed in this specification (including adjoint claim, abstract and attached drawing) and so disclosed
All process or units of what method or apparatus are combined.Unless expressly stated otherwise, this specification is (including adjoint power
Benefit require, abstract and attached drawing) disclosed in each feature can carry out generation with an alternative feature that provides the same, equivalent, or similar purpose
It replaces.
In addition, it will be appreciated by those of skill in the art that although some embodiments in this include institute in other embodiments
Including certain features rather than other feature, but the combination of the feature of different embodiment means in the scope of the present invention
Within and form different embodiments.For example, in the following claims, embodiment claimed it is any it
One can in any combination mode come using.
Various component embodiments of the invention can be implemented in hardware, or to run on one or more processors
Software module realize, or be implemented in a combination thereof.It will be understood by those of skill in the art that can be used in practice
Microprocessor or digital signal processor (DSP) realize some or all portions in device according to an embodiment of the present invention
The some or all functions of part.The present invention is also implemented as a part or complete for executing method as described herein
The device or device program (for example, computer program and computer program product) in portion.It is such to realize program of the invention
It can store on a computer-readable medium, or may be in the form of one or more signals.Such signal can be with
It downloads from internet website, is perhaps provided on the carrier signal or is provided in any other form.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and ability
Field technique personnel can be designed alternative embodiment without departing from the scope of the appended claims.In the claims,
Any reference symbol between parentheses should not be configured to limitations on claims.Word "comprising" does not exclude the presence of not
Element or step listed in the claims.Word "a" or "an" located in front of the element does not exclude the presence of multiple such
Element.The present invention can be by means of including the hardware of several different elements and being come by means of properly programmed computer real
It is existing.In the unit claims listing several devices, several in these devices can be through the same hardware branch
To embody.The use of word first, second, and third does not indicate any sequence.These words can be explained and be run after fame
Claim.
Claims (10)
1. a kind of archives scan image repair method characterized by comprising
Obtain archives scan image；
Based on the gray value of the archives scan image, judge whether the archives scan image is blurred picture；
If so, based on the machine learning model after training, according to the gray value of the archives scan image to the archives scan
Image is repaired, and clearly archival image is obtained.
2. the method according to claim 1, wherein in the gray value based on the archives scan image,
Before judging whether the archives scan image is blurred picture, the method also includes:
The archives scan image is pre-processed.
3. according to the method described in claim 2, it is characterized in that, described pre-process specifically to the archives scan image
Are as follows: the archives scan image is converted to gray level image.
4. according to the method described in claim 3, it is characterized in that, the gray scale based on the archives scan image, judgement
Whether the archives scan image is blurred picture, comprising:
Based on the gray level image, the feature of the archives scan image is extracted, obtains characteristic image；
Based on the characteristic image, judge whether the archives scan image is blurred picture.
5. according to the method described in claim 4, it is characterized in that, described judge the archives scan based on the characteristic image
Whether image is blurred picture, comprising:
Obtain the gray value of each pixel of the characteristic image；
Based on the gray value of each pixel, the variance of the gray value of the archives scan image is obtained；
If the variance of the gray value is less than or equal to given threshold, determine that the archives scan image is blurred picture.
6. method according to claim 1-5, which is characterized in that for repairing the machine learning model of image
Training method, comprising:
Multiple original blurred pictures and multiple original clear images are inputted in machine learning models, the machine learning model needle
To multiple described original blurred pictures and multiple described original clear images, multiple are exported respectively and first generates clear images and more
It opens first and generates blurred picture；
It will multiple described first generation clear images and multiple described first described machine learning models of generation blurred pictures input
In, the machine learning model is divided for multiple described first generation clear images and multiple described first generation blurred pictures
Multiple second generation blurred pictures and multiple second generation clear images are not exported；
Based on multiple described first generation clear images and multiple described original clear images, first-loss value is obtained；
Based on multiple described first generation blurred pictures and multiple described original blurred pictures, the second penalty values are obtained；
Based on multiple described second generation clear images, multiple described original clear images, multiple described second generation fuzzy graphs
Picture and multiple described original blurred pictures obtain third penalty values；
It imposes a condition if the first-loss value, second penalty values and the third penalty values meet, described in deconditioning
Machine learning model, the machine learning model after being trained, the machine learning model after the training are used for blurred picture
Revert to clear image.
7. according to the method described in claim 6, it is characterized in that, the training method of the machine learning model after the training also
Include:
If the first-loss value, second penalty values and the third penalty values are unsatisfactory for imposing a condition, the machine is adjusted
The training weight of device learning model, so that the first encoder of the machine learning model is identical with the weight of second encoder,
And the first generator of the machine learning model and the weight of the second generator it is identical, training be adjusted trained weight after
The machine learning model.
8. a kind of archives scan image fixing apparatus characterized by comprising
Module is obtained, for obtaining archives scan image；
Processing module judges whether the archives scan image is fuzzy for the gray value based on the archives scan image
Image；If so, based on the machine learning model after training, according to the gray value of the archives scan image to the archives scan
Image is repaired, and clearly archival image is obtained.
9. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is held by processor
The step of any one of claim 1-7 the method is realized when row.
10. a kind of electronic equipment, which is characterized in that on a memory and can be in processor including memory, processor and storage
The computer program of upper operation, the processor realize the step of any one of claim 1-7 the method when executing described program
Suddenly.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811559084.7A CN109712092B (en) | 2018-12-18 | 2018-12-18 | File scanning image restoration method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811559084.7A CN109712092B (en) | 2018-12-18 | 2018-12-18 | File scanning image restoration method and device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109712092A true CN109712092A (en) | 2019-05-03 |
| CN109712092B CN109712092B (en) | 2021-01-05 |
Family
ID=66255979
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811559084.7A Active CN109712092B (en) | 2018-12-18 | 2018-12-18 | File scanning image restoration method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109712092B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570375A (en) * | 2019-09-06 | 2019-12-13 | 腾讯科技（深圳）有限公司 | image processing method, image processing device, electronic device and storage medium |
| CN112416864A (en) * | 2020-11-18 | 2021-02-26 | 广东电网有限责任公司佛山供电局 | Automatic quality inspection method for digital files |
| CN113222843A (en) * | 2021-05-10 | 2021-08-06 | 北京有竹居网络技术有限公司 | Image restoration method and related equipment thereof |
| CN113792169A (en) * | 2021-09-16 | 2021-12-14 | 烟台市蓬莱区档案馆 | Digital archive management method and system based on big data application |
| CN117876270A (en) * | 2024-01-11 | 2024-04-12 | 安徽博凌信息科技有限公司 | Archive scanning image restoration method, archive scanning image restoration device and storage medium |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101609452A (en) * | 2009-07-10 | 2009-12-23 | 南方医科大学 | The fuzzy SVM feedback that is used for the medical image target identification is estimated method |
| US20120154581A1 (en) * | 2010-12-16 | 2012-06-21 | Industrial Technology Research Institute | Cascadable camera tampering detection transceiver module |
| WO2013009651A1 (en) * | 2011-07-12 | 2013-01-17 | Dolby Laboratories Licensing Corporation | Method of adapting a source image content to a target display |
| CN103632342A (en) * | 2013-11-12 | 2014-03-12 | 华南理工大学 | Fuzzy enhancement method for X-ray image in integrated circuit packaging |
| US20150109489A1 (en) * | 2011-12-19 | 2015-04-23 | Ziva Corporation | Computational imaging using variable optical transfer function |
| CN106952239A (en) * | 2017-03-28 | 2017-07-14 | 厦门幻世网络科技有限公司 | image generating method and device |
| CN107103590A (en) * | 2017-03-22 | 2017-08-29 | 华南理工大学 | A kind of image for resisting generation network based on depth convolution reflects minimizing technology |
| CN107133934A (en) * | 2017-05-18 | 2017-09-05 | 北京小米移动软件有限公司 | Image completion method and device |
| CN107945140A (en) * | 2017-12-20 | 2018-04-20 | 中国科学院深圳先进技术研究院 | A kind of image repair method, device and equipment |
| KR20180062819A (en) * | 2016-12-01 | 2018-06-11 | 한화에어로스페이스 주식회사 | Apparatus and method for processing image |
| CN108269245A (en) * | 2018-01-26 | 2018-07-10 | 深圳市唯特视科技有限公司 | A kind of eyes image restorative procedure based on novel generation confrontation network |
| CN108520504A (en) * | 2018-04-16 | 2018-09-11 | 湘潭大学 | A kind of blurred picture blind restoration method based on generation confrontation network end-to-end |
| CN108550118A (en) * | 2018-03-22 | 2018-09-18 | 深圳大学 | Fuzzy processing method, device, equipment and the storage medium of motion blur image |
| CN108711141A (en) * | 2018-05-17 | 2018-10-26 | 重庆大学 | The motion blur image blind restoration method of network is fought using improved production |
-
2018
- 2018-12-18 CN CN201811559084.7A patent/CN109712092B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101609452A (en) * | 2009-07-10 | 2009-12-23 | 南方医科大学 | The fuzzy SVM feedback that is used for the medical image target identification is estimated method |
| US20120154581A1 (en) * | 2010-12-16 | 2012-06-21 | Industrial Technology Research Institute | Cascadable camera tampering detection transceiver module |
| WO2013009651A1 (en) * | 2011-07-12 | 2013-01-17 | Dolby Laboratories Licensing Corporation | Method of adapting a source image content to a target display |
| US20150109489A1 (en) * | 2011-12-19 | 2015-04-23 | Ziva Corporation | Computational imaging using variable optical transfer function |
| CN103632342A (en) * | 2013-11-12 | 2014-03-12 | 华南理工大学 | Fuzzy enhancement method for X-ray image in integrated circuit packaging |
| KR20180062819A (en) * | 2016-12-01 | 2018-06-11 | 한화에어로스페이스 주식회사 | Apparatus and method for processing image |
| CN107103590A (en) * | 2017-03-22 | 2017-08-29 | 华南理工大学 | A kind of image for resisting generation network based on depth convolution reflects minimizing technology |
| CN106952239A (en) * | 2017-03-28 | 2017-07-14 | 厦门幻世网络科技有限公司 | image generating method and device |
| CN107133934A (en) * | 2017-05-18 | 2017-09-05 | 北京小米移动软件有限公司 | Image completion method and device |
| CN107945140A (en) * | 2017-12-20 | 2018-04-20 | 中国科学院深圳先进技术研究院 | A kind of image repair method, device and equipment |
| CN108269245A (en) * | 2018-01-26 | 2018-07-10 | 深圳市唯特视科技有限公司 | A kind of eyes image restorative procedure based on novel generation confrontation network |
| CN108550118A (en) * | 2018-03-22 | 2018-09-18 | 深圳大学 | Fuzzy processing method, device, equipment and the storage medium of motion blur image |
| CN108520504A (en) * | 2018-04-16 | 2018-09-11 | 湘潭大学 | A kind of blurred picture blind restoration method based on generation confrontation network end-to-end |
| CN108711141A (en) * | 2018-05-17 | 2018-10-26 | 重庆大学 | The motion blur image blind restoration method of network is fought using improved production |
Non-Patent Citations (3)
| Title |
|---|
| SCHULER C J等: "Learning to deblur", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| 吴梦婷等: "双框架卷积神经网络用于运动模糊图像盲复原", 《计算机辅助设计与图形学学报》 * |
| 胡成燕: "图像模糊篡改检测技术的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570375A (en) * | 2019-09-06 | 2019-12-13 | 腾讯科技（深圳）有限公司 | image processing method, image processing device, electronic device and storage medium |
| CN110570375B (en) * | 2019-09-06 | 2022-12-09 | 腾讯科技（深圳）有限公司 | Image processing method, device, electronic device and storage medium |
| CN112416864A (en) * | 2020-11-18 | 2021-02-26 | 广东电网有限责任公司佛山供电局 | Automatic quality inspection method for digital files |
| CN113222843A (en) * | 2021-05-10 | 2021-08-06 | 北京有竹居网络技术有限公司 | Image restoration method and related equipment thereof |
| CN113222843B (en) * | 2021-05-10 | 2023-11-10 | 北京有竹居网络技术有限公司 | Image restoration method and related equipment thereof |
| CN113792169A (en) * | 2021-09-16 | 2021-12-14 | 烟台市蓬莱区档案馆 | Digital archive management method and system based on big data application |
| CN113792169B (en) * | 2021-09-16 | 2022-05-10 | 烟台市蓬莱区档案馆 | Digital archive management method and system based on big data application |
| CN117876270A (en) * | 2024-01-11 | 2024-04-12 | 安徽博凌信息科技有限公司 | Archive scanning image restoration method, archive scanning image restoration device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109712092B (en) | 2021-01-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109712092A (en) | Archives scan image repair method, device and electronic equipment | |
| Lv et al. | Attention guided low-light image enhancement with a large scale low-light simulation dataset | |
| Wei et al. | Deep retinex decomposition for low-light enhancement | |
| CN111292264B (en) | Image high dynamic range reconstruction method based on deep learning | |
| CN110675336A (en) | Low-illumination image enhancement method and device | |
| WO2021218119A1 (en) | Image toning enhancement method and method for training image toning enhancement neural network | |
| CN111144491B (en) | Image processing method, device and electronic system | |
| CN111669514A (en) | High dynamic range imaging method and apparatus | |
| CN112102185B (en) | Image deblurring method and device based on deep learning and electronic equipment | |
| CN111179196B (en) | Multi-resolution depth network image highlight removing method based on divide-and-conquer | |
| CN110689495A (en) | Image restoration method for deep learning | |
| Leavline et al. | On teaching digital image processing with MATLAB | |
| CN112528782A (en) | Underwater fish target detection method and device | |
| CN107729885B (en) | Face enhancement method based on multiple residual error learning | |
| CN112801911B (en) | Method and device for removing text noise in natural image and storage medium | |
| Soni et al. | Removal of high density salt and pepper noise removal by modified median filter | |
| CN114299573A (en) | Video processing method and device, electronic equipment and storage medium | |
| CN113222856A (en) | Inverse halftone image processing method, terminal equipment and readable storage medium | |
| CN111968039B (en) | Day and night general image processing method, device and equipment based on silicon sensor camera | |
| Wang et al. | Mixed distortion image enhancement method based on joint of deep residuals learning and reinforcement learning | |
| CN116188720A (en) | Digital person generation method, device, electronic equipment and storage medium | |
| CN110766153A (en) | Neural network model training method and device and terminal equipment | |
| CN104639844B (en) | A kind of method and device of the long exposure image of calibrated analog | |
| CN113496468A (en) | Method and device for restoring depth image and storage medium | |
| CN114418863B (en) | Cell image restoration method, cell image restoration device, computer storage medium and electronic equipment |
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 | ||
| CB02 | Change of applicant information |
Address after: 200040, room 710, 302 Changping Road, Shanghai, Jingan District Applicant after: Shanghai Xinlian Information Development Co., Ltd Address before: 200040, room 710, 302 Changping Road, Shanghai, Jingan District Applicant before: SHANGHAI ZHONGXIN INFORMATION DEVELOPMENT Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |