CN108665509A - A kind of ultra-resolution ratio reconstructing method, device, equipment and readable storage medium storing program for executing - Google Patents
A kind of ultra-resolution ratio reconstructing method, device, equipment and readable storage medium storing program for executing Download PDFInfo
- Publication number
- CN108665509A CN108665509A CN201810442314.5A CN201810442314A CN108665509A CN 108665509 A CN108665509 A CN 108665509A CN 201810442314 A CN201810442314 A CN 201810442314A CN 108665509 A CN108665509 A CN 108665509A
- Authority
- CN
- China
- Prior art keywords
- image
- resolution
- super
- terahertz image
- terahertz
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000012545 processing Methods 0.000 claims abstract description 103
- 239000011159 matrix material Substances 0.000 claims description 28
- 230000011218 segmentation Effects 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 11
- 238000003709 image segmentation Methods 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 230000000306 recurrent effect Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 5
- 239000002699 waste material Substances 0.000 abstract description 5
- 230000001737 promoting effect Effects 0.000 abstract 1
- 230000008569 process Effects 0.000 description 9
- 238000001514 detection method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 239000013598 vector Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 239000000463 material Substances 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000001328 terahertz time-domain spectroscopy Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/001—Texturing; Colouring; Generation of texture or colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of ultra-resolution ratio reconstructing methods, and this approach includes the following steps：Obtain target Terahertz image；Processing is split to target Terahertz image, obtains multiple subgraphs；According to preset rules, multiple subgraphs are separately input to handle in corresponding super-resolution processing model；By treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.After target Terahertz image is split, corresponding super-resolution processing is carried out for different subgraphs.The waste of computer resource can be also reduced while promoting image resolution ratio, promote image processing efficiency.The invention also discloses a kind of super-resolution reconstruction device, equipment and readable storage medium storing program for executing, have corresponding technique effect.
Description
Technical field
The present invention relates to safety inspection technical fields, more particularly to a kind of ultra-resolution ratio reconstructing method, device, equipment
And readable storage medium storing program for executing.
Background technology
Terahertz time-domain spectroscopy detection technique can realize penetrability, lossless detection to nonmetallic substance.Contained by analysis
There are physical property of the Earth's materials and the terahertz light spectrum information of chemical property, may recognize that whether keep drugs, inflammable thing in detectable substance
Or the prohibited items such as bomb.
But the Terahertz image that generates of terahertz time-domain spectroscopy detection technique there is a problem that one it is obvious,
That is the resolution ratio of image is relatively low.Lower resolution ratio is unfavorable for obtaining and detecting the dangerous material in Terahertz image.It is existing
Super-resolution processing technology needs to occupy a large amount of computer resource and longer processing time to be that Terahertz image promotes resolution
Rate, this can cause safety check speed excessively slow, it is low there are safety check efficiency the problems such as, it is clear that unfavorable real-time safety check.
In conclusion the problems such as how effectively improving Terahertz image resolution ratio, is that current those skilled in the art are anxious
The technical issues of need to solving.
Invention content
The object of the present invention is to provide a kind of ultra-resolution ratio reconstructing method, device, equipment and readable storage medium storing program for executing, to carry
High Terahertz image resolution ratio.
In order to solve the above technical problems, the present invention provides the following technical solutions：
A kind of ultra-resolution ratio reconstructing method, including：
Obtain target Terahertz image；
Processing is split to the target Terahertz image, obtains multiple subgraphs；
According to preset rules, multiple subgraphs are separately input to carry out in corresponding super-resolution processing model
Processing；
By treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.
Preferably, processing is split to the target Terahertz image, obtains multiple subgraphs, including：
By the target Terahertz image according to 1:2:1 ratio, be divided into top subgraph, middle part subgraph and under
Portion's subgraph.
Preferably, processing is split to the target Terahertz image, obtains multiple subgraphs, including：
By the target Terahertz image according to 1:1:1 ratio is divided into left part subgraph, middle part subgraph and the right side
Portion's subgraph.
Preferably, described according to preset rules, multiple subgraphs are separately input to corresponding super-resolution processing
It is handled in model, including：
The subgraph for belonging to default key component is inputted into the first super-resolution image and handles model, will not belong to described
The subgraph of default key component inputs the second super-resolution image and handles model, to improve resolution ratio；Wherein, described first
It is that recurrent neural network handles model that super-resolution image, which handles model, and the second super-resolution image processing module is volume
Product Processing with Neural Network model.
Preferably, described by treated, multiple subgraphs are reconstructed, and obtain the Terahertz figure of super-resolution
Picture, including：
By treated, multiple subgraphs are reconstructed according to the corresponding position of segmentation, obtain super-resolution too
Hertz image.
Preferably, the acquisition target Terahertz image, including：
Obtain original Terahertz image；
Background difference processing is done to the original Terahertz image, obtains target Terahertz image.
Preferably, described that background difference processing is done to the original Terahertz image, obtain target Terahertz image, packet
It includes：
Calculate the difference matrix of the original Terahertz image corresponding matrix and default background matrix；
All elements in the difference matrix are traversed, and element value is more than to the corresponding figure of element set of predetermined threshold value
As determining target Terahertz image.
A kind of super-resolution reconstruction device, including：
Target Terahertz image collection module, for obtaining target Terahertz image；
Image segmentation module obtains multiple subgraphs for being split processing to the target Terahertz image；
Super-resolution processing module, for according to preset rules, multiple subgraphs being separately input to super accordingly
It is handled in resolution processes model；
Image reconstruction module, for will treated that multiple subgraphs are reconstructed, obtain the terahertz of super-resolution
Hereby image.
A kind of super-resolution reconstruction equipment, including：
Memory, for storing computer program；
Processor, the step of above-mentioned ultra-resolution ratio reconstructing method is realized when for executing the computer program.
A kind of readable storage medium storing program for executing is stored with computer program, the computer program quilt on the readable storage medium storing program for executing
The step of processor realizes above-mentioned ultra-resolution ratio reconstructing method when executing.
The method provided using the embodiment of the present invention obtains target Terahertz image；Target Terahertz image is carried out
Dividing processing obtains multiple subgraphs；According to preset rules, multiple subgraphs are separately input at corresponding super-resolution
It is handled in reason model；By treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.Consider
Into practical application, in a Terahertz image, the object to be identified only accounts for the part in entire Terahertz image.
Therefore, before doing the resolution processes for improving target Terahertz image, can first segmentation object Terahertz image, then will segmentation
The subgraph obtained afterwards is inputted in corresponding super-resolution processing model and is handled according to preset rules respectively.That is, can incite somebody to action
The subgraph for belonging to different location part uses different super-resolution processing models.Specifically, critical subgraph picture can be inputted
High-precision super-resolution processing model is handled, and more careful image is obtained, and for the subgraph of non-key part
It is handled as then carrying out simple super-resolution processing model.Then, by the subgraph after progress super-resolution processing respectively
As being reconstructed, the final Terahertz image for obtaining super-resolution.Emphasis improves the resolution ratio of parts of images, can reduce calculating
The waste of machine resource promotes image processing efficiency.When doing further object identification based on the Terahertz image, also it can be improved
The accuracy rate of identification.
Correspondingly, the embodiment of the present invention additionally provides Super-resolution reconstruction corresponding with above-mentioned ultra-resolution ratio reconstructing method
Structure device, equipment and readable storage medium storing program for executing, have above-mentioned technique effect, and details are not described herein.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will to embodiment or
Attached drawing needed to be used in the description of the prior art is briefly described, it should be apparent that, the accompanying drawings in the following description is only
Some embodiments of the present invention, for those of ordinary skill in the art, without creative efforts, also
It can be obtain other attached drawings according to these attached drawings.
Fig. 1 is a kind of implementing procedure figure of ultra-resolution ratio reconstructing method in the embodiment of the present invention；
Fig. 2 is image segmentation schematic diagram in a kind of ultra-resolution ratio reconstructing method in the embodiment of the present invention；
Fig. 3 is the implementing procedure figure of another ultra-resolution ratio reconstructing method in the embodiment of the present invention；
Fig. 4 is image reconstruction schematic diagram in a kind of ultra-resolution ratio reconstructing method in the embodiment of the present invention；
Fig. 5 is the implementing procedure figure of another ultra-resolution ratio reconstructing method in the embodiment of the present invention；
Fig. 6 is a kind of recurrent neural network processing model schematic in the embodiment of the present invention；
Fig. 7 is a kind of convolutional neural networks processing model schematic in the embodiment of the present invention；
Fig. 8 is a kind of structural schematic diagram of super-resolution reconstruction device in the embodiment of the present invention；
Fig. 9 is a kind of structural schematic diagram of super-resolution reconstruction equipment in the embodiment of the present invention.
Specific implementation mode
Core of the invention is to provide a kind of ultra-resolution ratio reconstructing method, and target Terahertz image is split, is divided into
Then multiple portions image inputs the multiple portions image that segmentation obtains in different super-resolution processing models, finally will
The high-definition picture of output is reconstructed, and obtains final high-definition picture.In this way, can quickly and effectively be promoted too
The resolution ratio of hertz image, is more advantageous to object identification.
Another core of the present invention is to provide a kind of super-resolution reconstruction device, equipment and readable storage medium storing program for executing, has upper
Technique effect is stated, details are not described herein.
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction with the accompanying drawings and specific embodiment party
The present invention is described in further detail for formula.Obviously, described embodiments are only a part of the embodiments of the present invention, and
The embodiment being not all of.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work
Under the premise of the every other embodiment that is obtained, shall fall within the protection scope of the present invention.
Embodiment one：
Referring to FIG. 1, Fig. 1 is a kind of flow chart of ultra-resolution ratio reconstructing method in the embodiment of the present invention, this method includes
Following steps：
S101, target Terahertz image is obtained.
Wherein, Terahertz (terahertz, THz) is one of vibration frequency unit, also known as terahertz or terahertz.
Signified Terahertz image is the image obtained by THz imaging technology in the embodiment of the present invention.It should be noted that
Terahertz energy very little will not generate destruction to substance.The mode for obtaining target Terahertz image can be to utilize terahertz
The Terahertz image that hereby imaging technique acquires in real time can also read from preset image library and acquire and store too in advance
Hertz image.Certainly, target Terahertz image can be the initial pictures directly collected, can also be by a series of
Terahertz image after processing.
S102, processing is split to target Terahertz image, obtains multiple subgraphs.
In practical applications, it can be found that important information in Terahertz image, i.e. object image are all distributed in too substantially
Middle section in hertz image.Therefore, when carrying out Image Super Resolution Processing, can emphasis be directed to object image
Middle section.And there is the smaller marginal portion of object image possibility and are then not necessarily to carry out the image procossing of degree of precision in other.
In this way, can emphasis augmented image resolution ratio, edge image is carried out without wasting excessive computer resource
Processing.In order to reach this purpose, processing can be split to target Terahertz image, then obtain multiple subgraphs.It needs
It is noted that when being split, for ease of operation, can simply it be divided, such as horizontal partition, longitudinally split, segmentation
The size of the subgraph gone out can be the same or different.Specifically, reference can be made to partitioning scheme shown in Fig. 2.It needs to illustrate
, the multiple subgraphs divided are reconfigurable into complete target Terahertz image.
S103, according to preset rules, multiple subgraphs are separately input to carry out in corresponding super-resolution processing model
Processing.
Wherein, super-resolution (Super-Resolution) is to improve original image by the method for hardware or software
Resolution ratio.The method of the super-resolution technique combination deep learning applied herein, with the deep learning of two different structures
Image processing model carries out the super-resolution processing of image.
It should be noted that the processing mode that can be directed to subgraph in embodiments of the present invention lays down a regulation.For example,
It is provided in preset rules：Which super-resolution processing model belongs to the subgraph of which part should be input in and handled.
Specifically, object image being there is a possibility that, higher subgraph inputs increasingly complex, the oversubscription that treatment effect becomes apparent
It is handled in resolution processing model, and object image be there is a possibility that into lower subgraph the input phase to treatment effect phase
To poor, but handled in the relatively simple super-resolution processing model of structure.
Specifically, after obtaining multiple subgraphs, multiple subgraphs can be separately input to phase according to preset rules
It is handled in the super-resolution processing model answered.It should be noted which super-resolution processing mould arrived for specifically used
The concrete structure of type and super-resolution processing model does not influence the embodiment of the present invention and realizes raising Terahertz image resolution ratio
Purpose.Therefore, existing super-resolution processing model can refer to for the concrete processing procedure of super-resolution processing model,
Details are not described herein.
S104, by treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.
After subgraph is handled by corresponding super-resolution processing model, resolution ratio has been obtained for phase
The promotion answered.Then subgraph that these are improved to resolution ratio is reconstructed, and can get the Terahertz image of super-resolution.
That is, the resolution ratio of the Terahertz image after reconstruct is compared to target Terahertz image higher, and the Terahertz image
The resolution ratio of the various pieces of resolution ratio is different.Specifically, the resolution of the higher part of possibility existing for object image
Rate there is a possibility that lower part higher compared to object image.In this way, being more advantageous to article identification.
The method provided using the embodiment of the present invention obtains target Terahertz image；Target Terahertz image is carried out
Dividing processing obtains multiple subgraphs；According to preset rules, multiple subgraphs are separately input at corresponding super-resolution
It is handled in reason model；By treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.Consider
Into practical application, in a Terahertz image, the object to be identified only accounts for the part in entire Terahertz image.
Therefore, before doing the resolution processes for improving target Terahertz image, can first segmentation object Terahertz image, then will segmentation
The subgraph obtained afterwards is inputted in corresponding super-resolution processing model and is handled according to preset rules respectively.That is, can incite somebody to action
The subgraph for belonging to different location part uses different super-resolution processing models.Specifically, critical subgraph picture can be inputted
High-precision super-resolution processing model is handled, and more careful image is obtained, and for the subgraph of non-key part
It is handled as then carrying out simple super-resolution processing model.Then, by the subgraph after progress super-resolution processing respectively
As being reconstructed, the final Terahertz image for obtaining super-resolution.Emphasis improves the resolution ratio of parts of images, can reduce calculating
The waste of machine resource promotes image processing efficiency.When doing further object identification based on the Terahertz image, also it can be improved
The accuracy rate of identification.
It should be noted that being based on above-described embodiment one, the embodiment of the present invention additionally provides the scheme of being correspondingly improved.
Involved in subsequent embodiment with can mutually be referred between same steps or corresponding steps in above-described embodiment one, it is corresponding beneficial
Effect also can be cross-referenced, is no longer repeated one by one in improvement embodiment below.
Embodiment two：
Referring to FIG. 3, Fig. 3 is the flow chart of another ultra-resolution ratio reconstructing method in the embodiment of the present invention, this method packet
Include following steps：
S201, original Terahertz image is obtained.
Original Terahertz image in the embodiment of the present invention can be Terahertz safety check image.When because of safety check, scene is detected
It is relatively fixed, therefore acquired original Terahertz image also has relatively-stationary background.It should be noted that institute here
The original Terahertz image said can also be the image after simple processing operation.For example, original Terahertz image can be with
For the image after denoising.
S202, background difference processing is done to original Terahertz image, obtain target Terahertz image.
Because the original Terahertz image got in step s 201 has relatively-stationary background.And in practical application
In, it is contemplated that the dangerous material that common people carry are concentrated mainly on the middle part of human body, i.e., in the image of on-fixed background parts.Cause
And when the object in Terahertz image is identified, background is only used as redundancy section to exist.Therefore, it is carried on the back to reduce image
Scape generates interference to image processing process, and improves the resolution ratio of article in image, further to detect in images of items
Article whether be dangerous material, first the background in accessed Terahertz image can be detached, obtain target Terahertz
Image.
S203, by target Terahertz image according to 1:2:1 ratio, be divided into top subgraph, middle part subgraph and under
Portion's subgraph.
According to the general experience of life, it can be found that the important information in Terahertz image, i.e. object image all divide substantially
Cloth is in the middle section of Terahertz image.Therefore, middle section and marginal portion can be split, so that emphasis is to centre
Part carries out super-resolution processing.For example, when carrying out image segmentation, it can be according to 1： 2：1 ratio is by target Terahertz image
Divide top subgraph, middle part subgraph and lower part subgraph.Specifically, the matrix of the image by the target object of interception
Line number s and columns l, be in proportion 1/4,1/2,1/4 this ratio is split：
Wherein, aijFor pixel, X'1, X'2, X'3Correspondence image is split respectively top subgraph, middle part subgraph
Picture and lower part subgraph.
Certainly, in other embodiments of the invention, the ration of division of image can also be other ratios, specific to compare
Example can predefine, and can also be determined and adjust according to actual conditions, not limit herein.In addition, in the present embodiment
In, top subgraph, middle part subgraph and lower part subgraph are divided the image into, may be used also in other embodiments of the invention
It is divided into 2 or 2 or more subgraphs.
S204, according to preset rules, multiple subgraphs are separately input to carry out in corresponding super-resolution processing model
Processing.
All it is distributed in the middle section of Terahertz image substantially based on object image, it in the present embodiment, can be by middle part
Subgraph is input in the better super-resolution processing model of process performance and is handled, by top subgraph and lower part subgraph
It is handled as being input in the relatively common super-resolution processing model of process performance.In this way, can for object image appearance
The higher middle part subgraph of energy property carries out high-precision processing, and the relatively low top of possibility occurs for object image
It is handled in subgraph and lower part subgraph.While improving object image resolution ratio, computer resource can be also reduced
Waste.
S205, by treated, multiple subgraphs are reconstructed according to the corresponding position of segmentation, obtain super-resolution too
Hertz image.
Referring to FIG. 4, by treated, multiple subgraphs are reconstructed according to the corresponding position of segmentation, obtain super-resolution
The Terahertz image of rate.
Preferably, step S202 includes in other embodiments of the invention：
Step 1: calculating the difference matrix of original Terahertz image corresponding matrix and default background matrix；
Step 2: all elements in traversal difference matrix, and element value is corresponding more than the element set of predetermined threshold value
Image determines target Terahertz image.
For ease of description.Above-mentioned 2 steps are combined below and are illustrated.
Because Terahertz Image Acquisition is usually applied in such as relatively-stationary scene of safety detection background environment.In reality
In the application of border, the background of accessed original Terahertz image hardly generates variation.Based on this, implement in the present invention
In example, the corresponding matrix T of image of the background residing for an object can be obtained in advance.Then, in Terahertz image to be carried out
When super-resolution processing, then the pending corresponding matrix W of Terahertz image can be made the difference two matrixes：Δ=W-T,
Y=Δs={ aij0≤i≤N, 0≤j≤M, wherein i and j is respectively the line number and row number of matrix, can determine that picture by i and j
Vegetarian refreshments position, Y are the matrix of target Terahertz image.Work as aij≤ n (n is predetermined threshold value, as n can be 10) when, illustrate the picture
Vegetarian refreshments is not belonging to background, but a part for object.All pixels point less than predetermined threshold value is extracted, can extract too
The corresponding pixel of object image in the corresponding matrix W of hertz image, to obtain the corresponding matrix Y of target Terahertz image.
That is, the corresponding matrix Y of the target Terahertz image is after Terahertz image carries out background separation, the target obtained is too
Hertz image.
After background is detached, interference of the image background when carrying out super-resolution image processing can be reduced.And
The Terahertz image of the super-resolution obtained after separating background, when being applied in object identification, detection, because of no background interference,
It also may make recognition detection result more accurate.
Embodiment three：
Referring to FIG. 5, Fig. 5 is the flow chart of another ultra-resolution ratio reconstructing method in the embodiment of the present invention, this method packet
Include following steps：
S301, target Terahertz image is obtained.
S302, by target Terahertz image according to 1:1:1 ratio is divided into left part subgraph, middle part subgraph and the right side
Portion's subgraph.
According to the general experience of life, it can be found that the important information in Terahertz image, i.e. object image all divide substantially
Cloth is in the middle section of Terahertz image.Therefore, before carrying out super-resolution processing to image, image can be divided
It cuts, so that emphasis is to there is the middle section of important information to carry out image procossing.It, can be according to 1 when to image segmentation：1：1 ratio
Example (i.e. trisection), by target Terahertz image segmentation top subgraph, middle part subgraph and lower part subgraph.
S303, the subgraph for belonging to default key component is inputted to the first super-resolution image processing model, will not belong to
The subgraph of default key component inputs the second super-resolution image and handles model, to improve resolution ratio.
Wherein, the first super-resolution image processing model is that recurrent neural network handles model, the second super-resolution image
Processing module is that convolutional neural networks handle model.
In the present embodiment, the key component of image can be pre-set.For example, can be by the middle part of target Terahertz image
It is set to off key section, correspondingly, the middle part subgraph obtained after segmentation is to pre-set the corresponding subgraph of key component
Picture.
Middle part subgraph is inputted the first super-resolution image processing model to handle, by left part subgraph and right part
Subgraph inputs the second super-resolution image processing model and is handled.Wherein, the first super-resolution image processing model can be with
Model is handled for recurrent neural network, it is that convolutional neural networks handle model that the second super-resolution image, which handles model,.
Wherein, the first super-resolution image handles model, the super-resolution image of such as Fig. 6, i.e. recurrent neural network composition
It includes three networks to handle model：
First layer network is convolutional network, and the extraction of the feature for image, first layer network representation is operation F1：F1
(Y)=max (0, W1*Y+B1), wherein Y is input picture, W1、B1Indicate that filter and deviation, " * " indicate convolution fortune respectively
It calculates.Filter W1It is specifically as follows and supports n × f3×f3N1Filter.Intuitively, filter W1To image application n1Filter
Convolution is carried out, and the kernel size of each convolution is c × f1×f1, wherein c is the port number in input picture, f1It is filtering
The space size of device.Output is by n1Filter characteristic figure forms.B1It is n1Dimensional vector, each of which element are associated with filter.
ReLU (max (0, x)) functional form is used in filter response.
Second layer network is a Recursive Networks, be used for image Nonlinear Mapping, such Recursive Networks solve for
Realization is preferably non-linear to be caused to avoid the problem that excessive network parameter using more convolutional layers increase network receptive field.This
The Recursive Networks of a part input reconstructed network, ensure Image Information Processing by acquiring the output of each go-between
Integrality.
The input matrix H of Recursive Networks0And calculating matrix export HN.All operations can use identical weight W and partially
Lean on matrix b.G is enabled to indicate by the function of single recusive modeling of recurrence layer：
G (H)=max (0, W*H+b), Hn=g (Hn-1)=max (0, W*Hn-1+b)；For n=1 ..., N, (n expressions are passed
Return the feedback convolution number in network).The output Y of Recursive Networks is equivalent to the combination of these basic functions g：Y=(zero g zero ... of g
Zero) g (H)=gn(H), wherein operator "○" representative function is constituted, gnIndicate the n times of product of g.
Third layer network is reconstructed network, and image reconstruction is combined into height according to the weight of different outputs above
Image in different resolution.
Reconstruction network representation is Y1, input hidden state Hn, export corresponding target image Y'(high-resolution).Specifically
Ground rebuilds the inverse operation that network is Recursive Networks.The formula of reconstructed network is as follows：
Hn+1=max (0, Wn+1*Hn+bn+1), Y 'n=max (0, Wn+2*Hn+1+bn+2).The result Y ' of final output is all
The weighted average of medium range forecast：Wherein n=1 ..., N.Wherein WnIt indicates in a recursive process from every
The prediction weight that a intermediate hidden state is rebuild.These weights can learn to arrive in training.
Specifically, the second super-resolution image handles model, the super-resolution figure of such as Fig. 7, i.e. convolutional neural networks composition
As processing model, to improve resolution ratio.The model is made of three convolutional layers：
First layer is extraction of the convolutional network for the feature of image, former with the convolution layer network of above-mentioned network model
Reason is the same, and details are not described herein.
Nonlinear Mapping of the second layer convolutional network for image is handled, and realizes the raising of image block resolution ratio.First layer
Extract the n of each image block1Dimensional feature.In second layer network operation, by these n1Dimension DUAL PROBLEMS OF VECTOR MAPPING is at n2Dimension.Second
The operation of layer is expressed as F2：F2(Y)=max (0, W2*F1(Y)+B2), wherein W2Including size is n × f1×f1N2Filter,
B2It is n2The deviation of dimension.Each n2The vector table for the high-definition picture block that the output vector of dimension conceptually can be used for rebuilding
Show.It in practical applications, can be non-linear to increase by adding more convolutional layers.
Third layer convolutional network is used for the reconstruct of image block, exports final high-definition picture.High-definition picture block
It is usually averaged to generate final complete image.The predefined filter that average value can be considered as on one group of characteristic pattern.Root
According to this principle, convolution layer operation F is defined to generate final high-definition picture：
F (Y)=W3*F2(Y)+B3, wherein W3It is n × f corresponding to size3×f3C filters, B3It is c dimensional vectors.
S304, by treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.
After obtaining Terahertz image, it can also output it.
Corresponding to above method embodiment, the embodiment of the present invention additionally provides a kind of super-resolution reconstruction device, hereafter
The super-resolution reconstruction device of description can correspond reference with above-described ultra-resolution ratio reconstructing method.
Shown in Figure 8, which comprises the following modules：
Target Terahertz image collection module 101, for obtaining target Terahertz image；
Image segmentation module 102 obtains multiple subgraphs for being split processing to target Terahertz image；
Super-resolution processing module 103, for according to preset rules, multiple subgraphs being separately input to super accordingly
It is handled in resolution processes model；
Image reconstruction module 104, for will treated that multiple subgraphs are reconstructed, obtain the terahertz of super-resolution
Hereby image.
The device provided using the embodiment of the present invention obtains target Terahertz image；Target Terahertz image is carried out
Dividing processing obtains multiple subgraphs；According to preset rules, multiple subgraphs are separately input at corresponding super-resolution
It is handled in reason model；By treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.Consider
Into practical application, in a Terahertz image, the object to be identified only accounts for the part in entire Terahertz image.
Therefore, before doing the resolution processes for improving target Terahertz image, can first segmentation object Terahertz image, then will segmentation
The subgraph obtained afterwards is inputted in corresponding super-resolution processing model and is handled according to preset rules respectively.That is, can incite somebody to action
The subgraph for belonging to different location part uses different super-resolution processing models.Specifically, critical subgraph picture can be inputted
High-precision super-resolution processing model is handled, and more careful image is obtained, and for the subgraph of non-key part
It is handled as then carrying out simple super-resolution processing model.Then, by the subgraph after progress super-resolution processing respectively
As being reconstructed, the final Terahertz image for obtaining super-resolution.Emphasis improves the resolution ratio of parts of images, can reduce calculating
The waste of machine resource promotes image processing efficiency.When doing further object identification based on the Terahertz image, also it can be improved
The accuracy rate of identification.
In a kind of specific implementation mode of the present invention, image segmentation module 102 is specifically used for target Terahertz figure
As according to 1:2:1 ratio is divided into top subgraph, middle part subgraph and lower part subgraph.
In a kind of specific implementation mode of the present invention, image segmentation module 102 is specifically used for target Terahertz figure
As according to 1:1:1 ratio is divided into left part subgraph, middle part subgraph and right part subgraph.
In a kind of specific implementation mode of the present invention, super-resolution processing module 103 is default specifically for that will belong to
The subgraph of key component inputs the first super-resolution image and handles model, and the subgraph that will not belong to default key component is defeated
Enter the second super-resolution image processing model, to improve resolution ratio；Wherein, the first super-resolution image processing model is recurrence
Processing with Neural Network model, the second super-resolution image processing module are that convolutional neural networks handle model.
In a kind of specific implementation mode of the present invention, image reconstruction module 104, being specifically used for will that treated be multiple
Subgraph is reconstructed according to the corresponding position of segmentation, obtains the Terahertz image of super-resolution.
In a kind of specific implementation mode of the present invention, target Terahertz image collection module 101, including：
Original Terahertz image acquisition unit, for obtaining original Terahertz image；
Difference processing unit obtains target Terahertz image for doing background difference processing to original Terahertz image.
In a kind of specific implementation mode of the present invention, difference processing unit, including：
Difference matrix computation subunit, for calculating the corresponding matrix of original Terahertz image and default background matrix
Difference matrix；
Target Terahertz image determines subunit, is more than in advance for traversing all elements in difference matrix, and by element value
If the corresponding image of the element set of threshold value determines target Terahertz image.
Corresponding to above method embodiment, the embodiment of the present invention additionally provides a kind of super-resolution reconstruction equipment, hereafter
A kind of super-resolution reconstruction equipment of description can correspond reference with a kind of above-described ultra-resolution ratio reconstructing method.
Shown in Figure 9, which includes：
Memory D1, for storing computer program；
Processor D2 realizes the ultra-resolution ratio reconstructing method of above method embodiment when for executing computer program
Step.
Corresponding to above method embodiment, the embodiment of the present invention additionally provides a kind of readable storage medium storing program for executing, is described below
A kind of readable storage medium storing program for executing can correspond reference with a kind of above-described ultra-resolution ratio reconstructing method.
A kind of readable storage medium storing program for executing is stored with computer program on readable storage medium storing program for executing, and computer program is held by processor
The step of ultra-resolution ratio reconstructing method of above method embodiment is realized when row.
Each embodiment is described by the way of progressive in this specification, the highlights of each of the examples are with its
The difference of its embodiment, just to refer each other for same or similar part between each embodiment.Disclosed in embodiment
For device, since it is corresponded to the methods disclosed in the examples, so description is fairly simple, related place is referring to method
Part illustrates.
Professional further appreciates that, list described in conjunction with the examples disclosed in the embodiments of the present disclosure
Member and algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, hard in order to clearly demonstrate
The interchangeability of part and software generally describes each exemplary composition and step according to function in the above description.
These functions are implemented in hardware or software actually, depend on the specific application and design constraint of technical solution.
Professional technician can use different methods to achieve the described function each specific application, but this reality
Now it should not be considered as beyond the scope of the present invention.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly use hardware, processor
The combination of the software module or the two of execution is implemented.Software module can be placed in random access memory (RAM), memory, only
Read memory (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or
In any other form of storage medium well known in technical field.
Principle and implementation of the present invention are described for specific case used herein, above example
Illustrate to be merely used to help understand technical scheme of the present invention and its core concept.It should be pointed out that for the general of the art
For logical technical staff, without departing from the principle of the present invention, can with several improvements and modifications are made to the present invention,
These improvement and modification are also fallen within the protection scope of the claims of the present invention.
Claims (10)
1. a kind of ultra-resolution ratio reconstructing method, which is characterized in that including：
Obtain target Terahertz image；
Processing is split to the target Terahertz image, obtains multiple subgraphs；
According to preset rules, multiple subgraphs are separately input to handle in corresponding super-resolution processing model；
By treated, multiple subgraphs are reconstructed, and obtain the Terahertz image of super-resolution.
2. ultra-resolution ratio reconstructing method according to claim 1, which is characterized in that carried out to the target Terahertz image
Dividing processing obtains multiple subgraphs, including：
By the target Terahertz image according to 1:2:1 ratio is divided into top subgraph, middle part subgraph and lower part subgraph
Picture.
3. ultra-resolution ratio reconstructing method according to claim 1, which is characterized in that carried out to the target Terahertz image
Dividing processing obtains multiple subgraphs, including：
By the target Terahertz image according to 1:1:1 ratio is divided into left part subgraph, middle part subgraph and right part subgraph
Picture.
4. ultra-resolution ratio reconstructing method according to claim 1, which is characterized in that it is described according to preset rules, it will be multiple
The subgraph is separately input to be handled in corresponding super-resolution processing model, including：
The subgraph for belonging to default key component is inputted into the first super-resolution image and handles model, will not belong to the default pass
The subgraph of key section inputs the second super-resolution image and handles model, to improve resolution ratio；Wherein, first super-resolution
Image processing model is that recurrent neural network handles model, and the second super-resolution image processing module is convolutional neural networks
Handle model.
5. ultra-resolution ratio reconstructing method according to claim 1, which is characterized in that it is described will treated multiple sons
Image is reconstructed, and obtains the Terahertz image of super-resolution, including：
By treated, multiple subgraphs are reconstructed according to the corresponding position of segmentation, obtain the Terahertz figure of super-resolution
Picture.
6. ultra-resolution ratio reconstructing method according to any one of claims 1 to 5, which is characterized in that the acquisition target is too
Hertz image, including：
Obtain original Terahertz image；
Background difference processing is done to the original Terahertz image, obtains target Terahertz image.
7. ultra-resolution ratio reconstructing method according to claim 6, which is characterized in that described to the original Terahertz image
Background difference processing is done, target Terahertz image is obtained, including：
Calculate the difference matrix of the original Terahertz image corresponding matrix and default background matrix；
All elements in the difference matrix are traversed, and element value is more than to the corresponding image determination of element set of predetermined threshold value
Target Terahertz image.
8. a kind of super-resolution reconstruction device, which is characterized in that including：
Target Terahertz image collection module, for obtaining target Terahertz image；
Image segmentation module obtains multiple subgraphs for being split processing to the target Terahertz image；
Super-resolution processing module, for according to preset rules, multiple subgraphs to be separately input to corresponding super-resolution
It is handled in rate processing model；
Image reconstruction module, for will treated that multiple subgraphs are reconstructed, obtain the Terahertz figure of super-resolution
Picture.
9. a kind of super-resolution reconstruction equipment, which is characterized in that including：
Memory, for storing computer program；
Processor realizes the super-resolution reconstruction side as described in any one of claim 1 to 7 when for executing the computer program
The step of method.
10. a kind of readable storage medium storing program for executing, which is characterized in that be stored with computer program, the meter on the readable storage medium storing program for executing
It is realized when calculation machine program is executed by processor as described in any one of claim 1 to 7 the step of ultra-resolution ratio reconstructing method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810442314.5A CN108665509A (en) | 2018-05-10 | 2018-05-10 | A kind of ultra-resolution ratio reconstructing method, device, equipment and readable storage medium storing program for executing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810442314.5A CN108665509A (en) | 2018-05-10 | 2018-05-10 | A kind of ultra-resolution ratio reconstructing method, device, equipment and readable storage medium storing program for executing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108665509A true CN108665509A (en) | 2018-10-16 |
Family
ID=63778237
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810442314.5A Pending CN108665509A (en) | 2018-05-10 | 2018-05-10 | A kind of ultra-resolution ratio reconstructing method, device, equipment and readable storage medium storing program for executing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108665509A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109903289A (en) * | 2019-04-17 | 2019-06-18 | 广东工业大学 | A kind of method, apparatus and equipment of terahertz image non-destructive testing |
| CN109978827A (en) * | 2019-02-25 | 2019-07-05 | 平安科技（深圳）有限公司 | Violated object recognition methods, device, equipment and storage medium based on artificial intelligence |
| CN110288530A (en) * | 2019-06-28 | 2019-09-27 | 北京金山云网络技术有限公司 | A kind of pair of image carries out the processing method and processing device of super-resolution rebuilding |
| CN110298790A (en) * | 2019-06-28 | 2019-10-01 | 北京金山云网络技术有限公司 | A kind of pair of image carries out the processing method and processing device of super-resolution rebuilding |
| CN110473145A (en) * | 2019-08-15 | 2019-11-19 | 广东工业大学 | Passive type terahertz image merges the detection method and device that ultrahigh resolution is rebuild |
| WO2020134848A1 (en) * | 2018-12-28 | 2020-07-02 | 深圳市华讯方舟太赫兹科技有限公司 | Intelligent detection method and device applied to millimeter wave security check instrument, and storage device |
| CN114005514A (en) * | 2021-11-26 | 2022-02-01 | 杭州涿溪脑与智能研究所 | Medical image diagnosis method, system and device |
| CN118134765A (en) * | 2024-04-30 | 2024-06-04 | 国家超级计算天津中心 | Image processing method, apparatus and storage medium |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101290682A (en) * | 2008-06-25 | 2008-10-22 | 北京中星微电子有限公司 | Movement target checking method and apparatus |
| CN105678728A (en) * | 2016-01-19 | 2016-06-15 | 西安电子科技大学 | High-efficiency super-resolution imaging device and method with regional management |
| CN105976318A (en) * | 2016-04-28 | 2016-09-28 | 北京工业大学 | Image super-resolution reconstruction method |
| CN106228512A (en) * | 2016-07-19 | 2016-12-14 | 北京工业大学 | Based on learning rate adaptive convolutional neural networks image super-resolution rebuilding method |
| CN106709872A (en) * | 2016-12-08 | 2017-05-24 | 中国人民解放军国防科学技术大学 | Quick image super-resolution reconstruction method |
| US20170193680A1 (en) * | 2016-01-04 | 2017-07-06 | Kla-Tencor Corporation | Generating high resolution images from low resolution images for semiconductor applications |
| CN107194872A (en) * | 2017-05-02 | 2017-09-22 | 武汉大学 | Remote sensed image super-resolution reconstruction method based on perception of content deep learning network |
| CN107481188A (en) * | 2017-06-23 | 2017-12-15 | 珠海经济特区远宏科技有限公司 | A kind of image super-resolution reconstructing method |
| CN107729948A (en) * | 2017-10-31 | 2018-02-23 | 京东方科技集团股份有限公司 | Image processing method and device, computer product and storage medium |
| CN107944379A (en) * | 2017-11-20 | 2018-04-20 | 中国科学院自动化研究所 | White of the eye image super-resolution rebuilding and image enchancing method based on deep learning |
| CN107967669A (en) * | 2017-11-24 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of picture processing |
| US20180124415A1 (en) * | 2016-05-06 | 2018-05-03 | Magic Pony Technology Limited | Encoder pre-analyser |
-
2018
- 2018-05-10 CN CN201810442314.5A patent/CN108665509A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101290682A (en) * | 2008-06-25 | 2008-10-22 | 北京中星微电子有限公司 | Movement target checking method and apparatus |
| US20170193680A1 (en) * | 2016-01-04 | 2017-07-06 | Kla-Tencor Corporation | Generating high resolution images from low resolution images for semiconductor applications |
| CN105678728A (en) * | 2016-01-19 | 2016-06-15 | 西安电子科技大学 | High-efficiency super-resolution imaging device and method with regional management |
| CN105976318A (en) * | 2016-04-28 | 2016-09-28 | 北京工业大学 | Image super-resolution reconstruction method |
| US20180124415A1 (en) * | 2016-05-06 | 2018-05-03 | Magic Pony Technology Limited | Encoder pre-analyser |
| CN106228512A (en) * | 2016-07-19 | 2016-12-14 | 北京工业大学 | Based on learning rate adaptive convolutional neural networks image super-resolution rebuilding method |
| CN106709872A (en) * | 2016-12-08 | 2017-05-24 | 中国人民解放军国防科学技术大学 | Quick image super-resolution reconstruction method |
| CN107194872A (en) * | 2017-05-02 | 2017-09-22 | 武汉大学 | Remote sensed image super-resolution reconstruction method based on perception of content deep learning network |
| CN107481188A (en) * | 2017-06-23 | 2017-12-15 | 珠海经济特区远宏科技有限公司 | A kind of image super-resolution reconstructing method |
| CN107729948A (en) * | 2017-10-31 | 2018-02-23 | 京东方科技集团股份有限公司 | Image processing method and device, computer product and storage medium |
| CN107944379A (en) * | 2017-11-20 | 2018-04-20 | 中国科学院自动化研究所 | White of the eye image super-resolution rebuilding and image enchancing method based on deep learning |
| CN107967669A (en) * | 2017-11-24 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of picture processing |
Non-Patent Citations (3)
| Title |
|---|
| 周亮等: ""基于分割的图像超分辨率重构算法"", 《数据采集与处理》 * |
| 胡彦婷等: ""超分辨率重建技术研究进展"", 《信息技术》 * |
| 谢云宇等: ""超分辨率重建技术及其在太赫兹图像中的应用"", 《***仿真技术》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020134848A1 (en) * | 2018-12-28 | 2020-07-02 | 深圳市华讯方舟太赫兹科技有限公司 | Intelligent detection method and device applied to millimeter wave security check instrument, and storage device |
| CN109978827A (en) * | 2019-02-25 | 2019-07-05 | 平安科技（深圳）有限公司 | Violated object recognition methods, device, equipment and storage medium based on artificial intelligence |
| CN109903289A (en) * | 2019-04-17 | 2019-06-18 | 广东工业大学 | A kind of method, apparatus and equipment of terahertz image non-destructive testing |
| CN110288530A (en) * | 2019-06-28 | 2019-09-27 | 北京金山云网络技术有限公司 | A kind of pair of image carries out the processing method and processing device of super-resolution rebuilding |
| CN110298790A (en) * | 2019-06-28 | 2019-10-01 | 北京金山云网络技术有限公司 | A kind of pair of image carries out the processing method and processing device of super-resolution rebuilding |
| CN110473145A (en) * | 2019-08-15 | 2019-11-19 | 广东工业大学 | Passive type terahertz image merges the detection method and device that ultrahigh resolution is rebuild |
| CN114005514A (en) * | 2021-11-26 | 2022-02-01 | 杭州涿溪脑与智能研究所 | Medical image diagnosis method, system and device |
| CN114005514B (en) * | 2021-11-26 | 2022-07-29 | 杭州涿溪脑与智能研究所 | Medical image diagnosis method, system and device |
| CN118134765A (en) * | 2024-04-30 | 2024-06-04 | 国家超级计算天津中心 | Image processing method, apparatus and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108665509A (en) | A kind of ultra-resolution ratio reconstructing method, device, equipment and readable storage medium storing program for executing | |
| Zhu et al. | A benchmark for edge-preserving image smoothing | |
| CN109858461B (en) | Method, device, equipment and storage medium for counting dense population | |
| CN112308200B (en) | Searching method and device for neural network | |
| CN111369440B (en) | Model training and image super-resolution processing method, device, terminal and storage medium | |
| CN109271933A (en) | The method for carrying out 3 D human body Attitude estimation based on video flowing | |
| CN109117944B (en) | Super-resolution reconstruction method and system for ship target remote sensing image | |
| CN110222598B (en) | Video behavior identification method and device, storage medium and server | |
| CN108475415A (en) | Method and system for image procossing | |
| CN105046651B (en) | A kind of ultra-resolution ratio reconstructing method and device of image | |
| CN109671020A (en) | Image processing method, device, electronic equipment and computer storage medium | |
| CN113159143B (en) | Infrared and visible light image fusion method and device based on jump connection convolution layer | |
| CN109214366A (en) | Localized target recognition methods, apparatus and system again | |
| Yang et al. | Multi-focus image fusion and super-resolution with convolutional neural network | |
| CN105046672A (en) | Method for image super-resolution reconstruction | |
| CN113469074B (en) | Remote sensing image change detection method and system based on twin attention fusion network | |
| CN111160288A (en) | Gesture key point detection method and device, computer equipment and storage medium | |
| CN113469072B (en) | Remote sensing image change detection method and system based on GSoP and twin fusion network | |
| CN109829353A (en) | A kind of facial image stylizing method based on space constraint | |
| CN110210278A (en) | A kind of video object detection method, device and storage medium | |
| CN107977930A (en) | A kind of image super-resolution method and its system | |
| CN110084181A (en) | A kind of remote sensing images Ship Target Detection method based on sparse MobileNetV2 network | |
| CN109064402A (en) | Based on the single image super resolution ratio reconstruction method for enhancing non local total variation model priori | |
| CN116485646A (en) | Micro-attention-based light-weight image super-resolution reconstruction method and device | |
| CN110335196A (en) | A kind of super-resolution image reconstruction method and system based on fractal decoding |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181016 |