CN110335330A - Image simulation generation method and its system, deep learning algorithm training method and electronic equipment - Google Patents
Image simulation generation method and its system, deep learning algorithm training method and electronic equipment Download PDFInfo
- Publication number
- CN110335330A CN110335330A CN201910634805.4A CN201910634805A CN110335330A CN 110335330 A CN110335330 A CN 110335330A CN 201910634805 A CN201910634805 A CN 201910634805A CN 110335330 A CN110335330 A CN 110335330A
- Authority
- CN
- China
- Prior art keywords
- image
- initial pictures
- simulation
- processing
- threshold value
- 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
- 238000004088 simulation Methods 0.000 title claims abstract description 113
- 238000000034 method Methods 0.000 title claims abstract description 102
- 238000012549 training Methods 0.000 title claims abstract description 33
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 20
- 238000013135 deep learning Methods 0.000 title claims abstract description 20
- 238000013519 translation Methods 0.000 claims abstract description 39
- 230000003321 amplification Effects 0.000 claims abstract description 20
- 238000003199 nucleic acid amplification method Methods 0.000 claims abstract description 20
- 238000004458 analytical method Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims description 80
- 238000006073 displacement reaction Methods 0.000 claims description 34
- 238000004590 computer program Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000005855 radiation Effects 0.000 claims description 7
- 230000008901 benefit Effects 0.000 claims description 4
- 230000004075 alteration Effects 0.000 abstract description 8
- 230000000694 effects Effects 0.000 abstract description 4
- 230000003287 optical effect Effects 0.000 description 20
- 238000010586 diagram Methods 0.000 description 19
- 201000009310 astigmatism Diseases 0.000 description 11
- 238000013041 optical simulation Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 6
- 230000006854 communication Effects 0.000 description 4
- 238000007796 conventional method Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000005457 Black-body radiation Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 230000005291 magnetic effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000750 progressive effect Effects 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 210000003127 knee Anatomy 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000009466 transformation Effects 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
- G06T11/40—Filling a planar surface by adding surface attributes, e.g. colour or texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/60—Editing figures and text; Combining figures or text
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
A kind of image simulation generation method provided by the present invention, it is based on the preliminary analysis to initial pictures, determine simulation degree threshold value, the translation predetermined pixel value that the initial pictures are carried out at least two directions is determined based on simulation degree threshold value, and the image after translation is weighted and averaged superposition, to export required analog image.It can solve the technical issues of conventional images analogue technique can not effectively simulate brings image aberration such as camera lens, optics, photosensitive element etc. using the above method.Described image, which simulates generation system and its electronic equipment, has technical effect identical with method.It can also solve the problems, such as that available data amplification technique can not will generate image and be modeled to true samples figure well based on deep learning algorithm training method provided by the present invention, to obtain more preferably training effect.
Description
[technical field]
The present invention relates to image data simulation field, in particular to a kind of image simulation generation method and its system, depth
Spend learning algorithm training method and electronic equipment.
[background technique]
The development of depth learning technology is so that computer picture vision technique has obtained strong support, but depth
The network structure of habit usually contains a large amount of parameter, in order to allow network that there is better performance (accuracy, Generalization Capability) etc.,
Need to provide training of a large amount of data for network.And in actual application, the acquisition of data and mark etc. are to be difficult
Meet magnitude required for training network, it would therefore be highly desirable to provide the technology of a kind of new image data quick obtaining or generation
Scheme.
[summary of the invention]
To solve conventional images data quick obtaining, the present invention provides a kind of image simulation generation side
Method and its system, deep learning algorithm training method and electronic equipment.
The present invention is in order to solve the above technical problems, offer the following technical solution: a kind of image simulation generation method, packet
Include: step S1 obtains initial pictures, and analyzes the size characteristic and color space structure feature of the initial pictures；Specifically,
It can be based on the scale and color space structure of the determination initial pictures, to generate original image or input image to be processed；Step
Rapid S2, size characteristic and color space structure feature based on the initial pictures determine simulation degree threshold value；And step S3, it will
The predetermined pixel value that the initial pictures are translated along preset direction, by initial pictures with translation after image be weighted and averaged it is folded
Add, to export required analog image；Wherein, the predetermined pixel value is determined based on simulation degree threshold value.
Preferably, above-mentioned steps S3 specifically comprises the following steps: step S31, sets a first direction and a second direction,
Predetermined pixel value successively is translated along the first direction and the second direction to the initial pictures, it is flat to obtain first respectively
Move image and the second displacement images；Step S32 handles first displacement images and second displacement images, with
Obtain the numerical value to match with position coordinates；And step S33, will in first displacement images and second displacement images
The numerical value that same position coordinate matches is weighted and averaged superposition, to export required analog image.
Preferably, between above-mentioned steps S2 and step S3 further include: step S201 amplifies the initial pictures
Or diminution processing；Wherein, the multiple that the initial pictures zoom in or out processing can be set based on the simulation degree threshold value
It is fixed.
Preferably, after above-mentioned steps S3, further includes: step S4 uses reflection model simulation or image interpolation method
Absent region is filled up.
It preferably, further include following steps after above-mentioned steps S2: step S3A, analog light source colour temperature, based on simulation
Degree threshold value is matched in conjunction with the ratio between black matrix light radiation colour temperature and rgb value, carries out colour temperature to the initial pictures
Adjustment；And/or the color space structure of the initial pictures is converted to YUV color space by step S3B, analogue exposure, packet
The channel Y is included, the channel Y is luminance channel, is exposed amendment to the channel Y, wherein exposure amendment can be based on the simulation journey
Degree threshold value is handled；And/or step S3C, data amplification processing is carried out to the initial pictures；Wherein, above-mentioned steps S3A,
Step S3B and step S3C can the optionally combination of one or several and step S3 progress random order.
Preferably, in above-mentioned steps S3C, the data amplification processing further comprises: overturning to initial pictures
Processing, rotation processing, scaling processing, cutting processing, translation processing, noise processed, color channel are handled, at image Shear Transform
Any one or several combination in reason.
The present invention is in order to solve the above technical problems, to provide another technical solution as follows: a kind of image simulation generates system,
Include: image preliminary analysis module, be configurable for obtain initial pictures, and analyze the initial pictures size characteristic and
Color space structure feature；Simulation degree threshold value obtains module, is configurable for the size characteristic based on the initial pictures
And color space structure feature determines simulation degree threshold value；And analog image generation module, being configurable for will be described initial
Initial pictures are weighted and averaged with the image after translation and are superimposed, with defeated by the predetermined pixel value that image is translated along preset direction
Required analog image out；Wherein, the predetermined pixel value is determined based on simulation degree threshold value.
Preferably, described image simulates generation system further include: picture size adjustment module is used for the initial pictures
Zoom in or out processing；Module is filled up in absent region, for using reflection model or image values average value to missing area
Domain is filled up；Analog light source colour temperature module, for being based on simulation degree threshold value, in conjunction with black matrix light radiation colour temperature and rgb value it
Between ratio matched, to the initial pictures carry out colour temperature adjustment；Analogue exposure module is used for the initial pictures
Color space structure be converted to YUV color space；And/or data expand module, for carrying out data to the initial pictures
Amplification processing.
The present invention is in order to solve the above technical problems, to provide another technical solution as follows: a kind of deep learning algorithm training side
Method, wherein the deep learning algorithm uses image simulation generation method as described above analog image combination initial graph obtained
As being trained and obtaining as training set.
The present invention is in order to solve the above technical problems, to provide another technical solution as follows: a kind of electronic equipment comprising storage
Unit and processing unit, the storage unit are used for single by the storage for storing computer program, the processing unit
The computer program of member storage executes image simulation generation method as described above.
Compared with prior art, a kind of image simulation generation method and its system, deep learning provided by the present invention given
Algorithm training method and electronic equipment have it is following the utility model has the advantages that
Object of the present invention is to which optical parallax is introduced image simulation by optical analog to generate field, so as to based on a small amount of
Basic image data, can be realized the enhancing and extension to image data, and can will generate data set enhance with quasi-
Close truthful data.The present invention is passed by can optically be supplemented in the insufficient situation of data volume based on optical simulation
The problem of system method data amplification cannot cover truth whole possibility.Processing for generating image data has good
Good results of Physical.Training set can be allowed closer to truth using this method, be conducive to the internetworking finally trained
It can be promoted.Specifically, in the present invention, based on the preliminary analysis to initial pictures, simulation degree threshold value, the initial graph are determined
As the predetermined pixel value of translation is with the image after translate, wherein the predetermined pixel value based on the determination of simulation degree threshold value,
And initial pictures are weighted and averaged with the image after translation and are superimposed, to export required analog image.
Image simulation generation method provided by the present invention can solve conventional images analogue technique can not effectively to camera lens,
The technical issues of brings image aberration such as optics, photosensitive element etc. is simulated.Meanwhile being based on image provided by the present invention
Simulation-generation method, which can also solve available data amplification technique for the data set of generation, can not will generate the fine simulation of image
The problem of at true sample graph.As it can be seen that image simulation generation method provided by the present invention has wider applicability.
In image simulation generation method provided by the present invention, based on the preliminary analysis to initial pictures, simulation journey is determined
Threshold value is spent, the predetermined pixel value that the initial pictures are carried out with the translation at least two directions is determined based on simulation degree threshold value,
And the image after translation is weighted and averaged superposition and specifically comprises the following steps: to set a first direction and a second direction,
Predetermined pixel value successively is translated along the first direction and the second direction to the initial pictures, it is flat to obtain first respectively
Move image and the second displacement images；First displacement images and second displacement images are handled, with acquisition and position
Set the numerical value that coordinate matches；And it will be with same position coordinate phase in first displacement images and second displacement images
The numerical value matched is weighted and averaged superposition, to export required analog image.Based on the numerical value to match to same position coordinate into
Row weighted average superposition, can get simulation astigmatism or image out of focus.
Further right, in order to which the initial pictures to different sizes are effectively handled, described image simulation is generated
Method further includes that the initial pictures are zoomed in or out processing；Wherein, the initial pictures zoom in or out processing
Multiple can be set based on the simulation degree threshold value.Based on this, can make to zoom in or out processing to the initial pictures
Multiple precision it is higher.
In order to avoid leading to analog image due to the scaling processing or translation processing generation absent region to initial pictures
Unavailable, described image simulation-generation method further includes being filled out using reflection model or image values average value to absent region
It mends.
It in the present invention, can also include to the analog light source colour temperature of initial pictures is handled, analogue exposure handles and data
Amplification processing, these processing can be carried out based on simulation degree threshold value, so that the accuracy of simulation process can be improved.Further
Ground, in image simulation generation method provided by the present invention, corresponding simulation astigmatism is simulated at out of focus, analog light source colour temperature
The step processing sequence of reason, analogue exposure processing and data amplification processing can be replaced arbitrarily, to meet the essence of different initial pictures
Quasi- processing.
In the present invention, data amplification processing further comprises: carrying out overturning processing, at rotation to initial pictures
Reason, scaling processing, cutting processing, translation processing, noise processed, color channel handle, are any in the processing of image Shear Transform
Or several combinations, it is based on this, a variety of different image simulation generation methods can be met, has described image simulation-generation method
There is wider array of applicability.
The present invention also provides a kind of image simulations to generate system and its electronic equipment, has and above-mentioned image simulation generation side
The identical beneficial effect of method, the present invention can be in the insufficient situations of data volume optically by being based on optical simulation
The problem of supplement conventional method data amplification cannot cover truth whole possibility.For generating the place of image data
Reason has good results of Physical.Training set can be allowed closer to truth using this method, be conducive to finally to train
Network performance is promoted.Specifically, in the present invention, it based on the preliminary analysis to initial pictures, determines simulation degree threshold value, is based on
Simulation degree threshold value determines the predetermined pixel value that the initial pictures are carried out with the translation at least two directions, and will be after translation
Image is weighted and averaged superposition, to export required analog image.
The present invention also provides a kind of deep learning algorithm training methods, use image simulation generation method institute as described above
The analog image combination initial pictures of acquisition are trained as training set and are obtained.The deep learning algorithm training method
Conventional method data amplification can be optically supplemented in the insufficient situation of data volume, and cannot to cover truth complete
The problem of portion's possibility.To provide more fully training data for the training of the deep learning algorithm, so that institute can be improved
State the effect of deep learning algorithm training.
[Detailed description of the invention]
Fig. 1 is the step flow diagram of image simulation generation method provided in first embodiment of the invention.
Fig. 2 is the specific steps flow diagram of content shown in step S3 in Fig. 1.
Fig. 3 is the step flow diagram of one of variant embodiment of image simulation generation method shown in Fig. 1.
Fig. 4 is two step flow diagram of the variant embodiment of image simulation generation method shown in Fig. 1.
Fig. 5 be using simulate it is out of focus as imagery optical simulate method image translation before schematic diagram.
Fig. 6 be using simulate it is out of focus as imagery optical simulate method image translation after schematic diagram.
Fig. 7 is to simulate the schematic diagram before the image translation for the method that astigmatism is simulated as imagery optical.
Fig. 8 is to simulate the schematic diagram after the image translation for the method that astigmatism is simulated as imagery optical.
Fig. 9 is the step flow diagram of one of variant embodiment of image simulation generation method shown in Fig. 1.
Figure 10 is the functional block diagram that image simulation provided in second embodiment of the invention generates system.
Figure 11 is the functional block diagram of analog image generation module shown in Figure 10.
Figure 12 is the functional block diagram of one of embodiment that image simulation shown in Figure 10 generates system.
Figure 13 is two the functional block diagram of the embodiment that image simulation shown in Figure 10 generates system.
Figure 14 is the functional block diagram of electronic equipment provided in fourth embodiment of the invention.
Description of drawing identification:
20, image simulation generates system；21, image preliminary analysis module；22, simulation degree threshold value obtains module；23, mould
Quasi- image generation module；231, translation unit；232, position coordinates matching unit；233, it is weighted and averaged superpositing unit；24, image
Size adjustment module；25, module is filled up in absent region；26, analog light source colour temperature module；27, analogue exposure module；28, data
Expand module
30, electronic equipment；31, storage unit；32, processing unit.
[specific embodiment]
In order to make the purpose of the present invention, technical solution and advantage are more clearly understood, below in conjunction with attached drawing and embodiment,
The present invention will be described in further detail.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention,
It is not intended to limit the present invention.
Referring to Fig. 1, providing a kind of image simulation generation method S10 in the first embodiment of the invention comprising as follows
Step:
Step S1 obtains initial pictures, and analyzes the size characteristic and color space structure feature of the initial pictures；Tool
Body, it can be based on the scale and color space structure of the determination initial pictures, to generate original image or input figure to be processed
Picture；
Step S2, size characteristic and color space structure feature based on the initial pictures determine simulation degree threshold value；
And
Step S3, the predetermined pixel value that the initial pictures are translated along preset direction, after initial pictures and translation
Image is weighted and averaged superposition, to export required analog image；Wherein, it is true to be based on simulation degree threshold value for the predetermined pixel value
It is fixed.
In above-mentioned steps S1, the size characteristic refers to the actual size size of the initial pictures, based on specific
Size characteristic, it may be determined whether need the processing that zooms in or out to image, in order to it is subsequent to the initial pictures into
Row optical aberration simulation process.Wherein, the optical aberrations may include spherical aberration, astigmatism, the curvature of field, distortion, chromatism of position and multiplying power
Color difference etc..
The color space structure feature may include that primary colours color space and color are showed the score from color space.The former typical case is
RGB (Red, Green, Blue), further include CMY (Cyan, Magenta, Yellow), CMYK (Cyan, Magenta, Yellow,
Black), CIE XYZ (Color System) etc.；The latter includes YCC/YUV, Lab and a collection of form and aspect class color space.
Further, in above-mentioned steps S2, size characteristic and color space structure feature based on the initial pictures
The degree threshold value of imagery optical simulation is determined, that is, can set based on the specific feature of initial pictures and obtain more accurate mould
Quasi- degree threshold value.
In the present invention, simulation degree threshold value may include carrying out image simulation based on imagery optical simulation, specifically, described
The method of imagery optical simulation may include but be not only restricted to simulate any one or several in out of focus, simulation astigmatism, analog color difference etc.
Combination.
Wherein, simulation degree threshold value can be directly reflected as analog image and initial pictures obtained after imagery optical simulation
Between difference value, wherein required difference value is bigger, then corresponding degree threshold value is also bigger, and required difference value is got over
Small, then corresponding degree threshold value is also smaller.
As shown in Figure 2, in above-mentioned steps S3, the predetermined pixel value that the initial pictures are translated along preset direction,
Initial pictures are weighted and averaged with the image after translation and are superimposed, to export required analog image, specific includes following
Step:
Step S31 sets a first direction and a second direction, to the initial pictures successively along the first direction and
The second direction translates predetermined pixel value, to obtain the first displacement images and the second displacement images respectively；
Step S32 handles first displacement images and second displacement images, with acquisition and position coordinates
The numerical value to match；And
Step S33, by what is matched with same position coordinate in first displacement images and second displacement images
Numerical value is weighted and averaged superposition, to export required analog image.
Specifically, in other some embodiments, the preset direction specifically can actual features and use based on picture
The demand at family does corresponding adjustment, specifically may further be two, four, six or eight, in some particular embodiments
In, it is also possible to three, five, seven etc., herein only as an example, not as the limitation of the invention.
Referring to Fig. 3, in order to obtain more diversified analog image, in some variant embodiments of the present embodiment,
It may also include the steps of: between above-mentioned steps S2 and step S3
The initial pictures are zoomed in or out processing by step S201.
Specifically, the multiple that the initial pictures zoom in or out processing can be set based on the simulation degree threshold value
It is fixed.
Further, referring to Fig. 4, in order to avoid due to occurring in the image of weighted average superposition after multi-direction translation
Absent region further after detecting that the superimposed image of weighted average has the region that numerical value is zero, then further comprises
Following steps:
Step S4 carries out absent region using reflection model (Reflection Model) simulation or image difference method
It fills up.
Wherein, the reflection model includes Cook-Torrance reflection model, Blinn-Phong reflection model and Phong
It is any in reflection model.
Described image number average value refers to that calculating obtains the initial pictures and corresponds to the numerical value of each pixel, and acquires
The average value of its numerical value.
In order to be preferably illustrated to above-described embodiment, with the generation of DM (Data Matrix) two dimensional code analog image
For, namely obtain DM two dimensional code initial pictures.
Specifically, in the present embodiment, as shown in Fig. 5 and Fig. 6, to simulate the side out of focus simulated as imagery optical
Method specifically includes:
For the DM two dimensional code initial pictures, the out of focus of camera is very common image error source.Based on above-mentioned
Step S2 obtains the simulation degree threshold value of corresponding DM two dimensional code initial pictures, and the degree of image focus is obtained with this, specifically, with
N/m is expressed as the astigmatism degree of rational, after m*m times of initial pictures is amplified to image, translates 1 pixel, 2 along multiple directions
Pixel is weighted and averaged superposition to obtain (n+1) and open the corresponding numerical value of new images until n-pixel.Such as some specific
In embodiment, after initial pictures A as shown in Figure 5 is amplified m*m times, it is based on X positive direction, X negative direction, Y positive direction and Y
Negative direction translates 1 pixel, after (i.e. n=1), obtains four new displacement images A ', is as shown in Figure 6 to move in X direction
Displacement images A ' obtained after dynamic n-pixel.
Initial pictures numerical value corresponding with four new displacement images A ' is further weighted and averaged superposition；Further
Ground, due to occurring absent region on the image after translation, the statistical average for further using image is filled up.
Wherein, it in order to obtain richer simulation image out of focus, can further can also be based on specifically in its direction translated
Content adjusts, and above-mentioned example is only as an example, not as the limitation of the invention.
In the present embodiment, as shown in Fig. 7 and Fig. 8, the method simulated using simulating astigmatism as imagery optical is specific
Include:
The astigmatism degree that rational is indicated using n/m, after image to be amplified to m*m times of initial pictures, along preset direction
1 pixel, 2 pixels are translated until n-pixel, obtained (n+1) picture is weighted and averaged and is superimposed.Such as it is specific real at other
It applies in mode, after initial pictures B as shown in Figure 7 is amplified m*m times, 2 pixels (i.e. n=2) is translated based on Y positive direction,
Two new displacement images B ' are obtained, are as shown in Figure 8 the displacement images B ' obtained after Y positive direction translation n-pixel.
Initial pictures are weighted and averaged with two new displacement images B ' corresponding datas further and are superimposed.
In the above example, the degree out of focus, the astigmatism degree are the simulation degree threshold value.
In the present embodiment, the method simulated using analog color difference as imagery optical, specifically includes:
One optical axis X is set to the initial pictures, it is assumed that length direction (the i.e. first party of DM two dimensional code initial pictures
To) it include n pixel, width direction (i.e. second direction) includes m pixel, is carried out along the direction opposite with the optical axis X flat
It moves, analog color difference result after superposition.In addition it is also possible to generate color difference analog parameter and carry out color difference simulation.
Since the light of different wave length has nuance by the focal length after lens group, different journeys are carried out to tri- channels RGB
After the processing out of focus of degree, image out of focus is simulated to obtain.
Referring to Fig. 9, in the other some embodiments of the present embodiment, in the present invention in order to meet various initial graphs
The optical aberration simulation demand of picture may also include following step further after above-mentioned steps S2:
Step S3A, analog light source colour temperature are based on simulation degree threshold value, in conjunction between black matrix light radiation colour temperature and rgb value
Ratio is matched, and the initial pictures are carried out with the adjustment of colour temperature；And/or
The color space structure of the initial pictures is converted to YUV (YCrCb, color coding by step S3B, analogue exposure
Method) color space, wherein the channel Y is luminance channel, is exposed amendment to the channel Y, wherein exposure amendment can be based on described
Simulation degree threshold value is handled；And/or
Step S3C carries out data amplification processing to the initial pictures, carries out at overturning including to initial pictures
Reason, rotation processing, scaling processing, cutting processing, translation processing, noise processed, color channel processing, the processing of image Shear Transform
Any one or several combination in.
As shown in Figure 9, the sequence between above three step be step S3A, step S3B and step S3C optionally its
One.This sequence of steps is only as an example, not as the limitation of the invention.Further, above-mentioned steps S3A, step S3B and step
S3C can the optionally combination of one or several and step S3 progress random order.
In above-mentioned steps S3A, in the black body radiation, with temperature difference, the color of light is different, described black
Body, which is presented, can indicate black by the progressive formation namely the black matrix light radiation colour temperature of red-orange red-Huang-yellowish-white-white-Lan Bai
In the progressive formation that body varies with temperature.The rgb value (Red, Green, Blue) is that the initial pictures correspond to each picture
R value, G value and the B value of vegetarian refreshments.The tune of corresponding colour temperature is carried out with the R value of corresponding pixel points, G value and B value based on the black body radiation
It is whole.
In above-mentioned steps S3B, the color space structure of the initial pictures is converted into YUV color space, wherein institute
Stating the channel Y is luminance channel, amendment is exposed to the channel Y, to can get new diagram.
Specifically, in above-mentioned steps S3C, Image Reversal processing, which refers to, carries out turning over for horizontal or vertical direction for image
Turn；Image rotation processing refers to the rotation that image is carried out to 90 degree, 180 degree, the different angles such as 270 degree, portion of techniques scheme pair
Postrotational picture will do it cutting to guarantee that image scaled is constant.Scaling processing refers to the contracting that different scale is carried out to image
It puts, different technologies scheme may use different interpolation methods, will do it cutting to the image after scaling or completion is (possible
Method has reflection, translation etc.).Random cropping processing refers to image into after row stochastic cutting, and the part of cutting is scaled
At full size；Translation processing refers to the translation that random direction is carried out to image, and portion of techniques scheme is to the absent region after translation
It is filled up；The mode of random noise processing has salt-pepper noise, the various ways such as Gaussian noise, by the random picture in image
Element is modified, to enhance the generalization ability of deep learning network；Color channel processing refers to logical by carrying out color to image
The modification in road generates the offset the result is that overall color style；The processing of image Shear Transform, refers to the matrixing by image
Deformed image is generated, and absent region is carried out to image and is filled up.
For example, as above-mentioned cited DM two dimensional code, in conjunction with above-mentioned specific processing step, due to final application field
Luminous environment colour temperature is about 6600K in scape, and color temperature parameters are adjusted to 6600K.Further set exposure status and overexposure with
Under-exposed degree.Can also decide whether out of focus or astigmatism, and generate random parameter in the appropriate range.To generate result with
It generates original image and carries out appropriate stochastic transformation composition data pair.
Referring to Fig. 10, the second embodiment of the present invention, which provides an image simulation, generates system 20 comprising:
Image preliminary analysis module 21 is configurable for obtaining initial pictures, and analyzes the size of the initial pictures
Feature and color space structure feature；
Simulation degree threshold value obtains module 22, is configurable for size characteristic and color sky based on the initial pictures
Between structure feature determine simulation degree threshold value；And
Analog image generation module 23 is configured as the predetermined pixel value for translating the initial pictures along preset direction,
Initial pictures are weighted and averaged with the image after translation and are superimposed, to export required analog image；Wherein, the intended pixel
Value is determined based on simulation degree threshold value.
As shown in Figure 11, in order to improve accuracy that the analog image generates to obtain satisfactory simulation drawing
Picture, the analog image generation module 23 further comprises:
Translation unit 231, for setting a first direction and a second direction, to the initial pictures successively along described the
One direction and the second direction translate predetermined pixel value, to obtain the first displacement images and the second displacement images respectively；
Position coordinates matching unit 232, for handling first displacement images and second displacement images,
To obtain the numerical value to match with position coordinates；And
Be weighted and averaged superpositing unit 233, for will with it is same in first displacement images and second displacement images
The numerical value that position coordinates match is weighted and averaged superposition, to export required analog image.
As shown in Figure 12, in some embodiments of the present embodiment, it is further that described image simulates generation system 20
Include:
Picture size adjustment module 24, for the initial pictures to be zoomed in or out processing；And
Module 25 is filled up in absent region, for using reflection model (Reflection Model) simulation or image interpolation side
Method fills up absent region.
Wherein, the reflection model includes Cook-Torrance reflection model, Blinn-Phong reflection model and Phong
It is any in reflection model.
Described image interpolation method refers to the gray scale that unknown pixel point is obtained using the gray value of known vicinity points
Value, to bear the image with higher resolution by original image, described image interpolation method may include arest neighbors value method,
Bilinear interpolation method etc..
As shown in Figure 13, various first in order to meet in the present embodiment in other embodiments of the present embodiment
The optical aberration of beginning image simulates demand, and described image simulation generation system 20 further comprises:
Analog light source colour temperature module 26, for being based on simulation degree threshold value, in conjunction between black matrix light radiation colour temperature and rgb value
Ratio matched, to the initial pictures carry out colour temperature adjustment；And/or
Analogue exposure module 27, for the color space structure of the initial pictures to be converted to YUV, (YCrCb, color are compiled
Code method) color space, wherein the channel Y is luminance channel, is exposed amendment to the channel Y, wherein exposure amendment can be based on institute
Simulation degree threshold value is stated to be handled；And/or
Data expand module 28, for carrying out data amplification processing to the initial pictures.
Specifically, wherein the data amplification module 28 includes carrying out overturning processing, rotation processing, contracting to initial pictures
It puts any or several in processing, cutting processing, translation processing, noise processed, color channel processing, the processing of image Shear Transform etc.
The combination of kind.
In the present embodiment, the related content generated in relation to specific image simulation related in above-mentioned first embodiment is retouched
It states unanimously, details are not described herein.
The third embodiment of the present invention provides a deep learning algorithm training method, uses as in above-mentioned first embodiment
Described image simulation-generation method analog image combination initial pictures obtained are trained as training set and are obtained.
Wherein, the particular content of described image simulation-generation method is identical as described in above-mentioned first embodiment, herein not
It repeats again.
Figure 14 is please referred to, the fourth embodiment of the present invention provides an electronic equipment 30, and the electronic equipment 30 includes storage
Unit 31 and processing unit 32, the storage unit 31 is for storing computer program, and the processing unit 32 is for passing through institute
The computer program for stating the storage of storage unit 31 executes the data checking method that exceptional value is examined described in above-mentioned first embodiment
Specific steps.
In some specific embodiments of the present invention, the electronic equipment 30 can be hardware, be also possible to software.Work as electricity
When sub- equipment 30 is hardware, the various electronic equipments of video playing are can be with display screen and supported, including but not limited to
Smart phone, tablet computer, E-book reader, MP3 player (Moving Picture Experts Group Audio
Layer III, dynamic image expert's compression standard audio level 3), MP4 (Moving Picture Experts Group
Audio Layer IV, dynamic image expert's compression standard audio level 4) player, pocket computer on knee and desk-top meter
Calculation machine etc..When electronic equipment 30 is software, may be mounted in above-mentioned cited electronic equipment.It may be implemented into more
A software or software module (such as providing multiple softwares of Distributed Services or software module), also may be implemented into single
Software or software module.It is not specifically limited herein.
The storage unit 31 includes the storage unit of read-only memory (ROM), random access storage device (RAM) and hard disk etc.
Point etc., the processing unit 32 according to the program being stored in the read-only memory (ROM) or can be loaded into random visit
It asks the program in memory (RAM) and executes various movements appropriate and processing.In random access storage device (RAM), also deposit
It contains the electronic equipment 30 and operates required various programs and data.
The electronic equipment 30 may also include the importation (not shown) of keyboard, mouse etc.；The electronic equipment 30 is also
Can further comprise cathode-ray tube (CRT), liquid crystal display (LCD) etc. and loudspeaker etc. output par, c (figure not
Show)；And the electronic equipment 30 can further comprise the communication unit of the network interface card of LAN card, modem etc.
Divide (not shown).The communications portion executes communication process via the network of such as internet.
Particularly, disclosed embodiment, the process described above with reference to flow chart may be implemented as counting according to the present invention
Calculation machine software program.For example, disclosed embodiment of this invention may include a kind of computer program product comprising be carried on meter
Computer program on calculation machine readable medium, the computer program include the program generation for method shown in execution flow chart
Code.In such embodiments, which can be downloaded and installed from network by communications portion.
When the computer program is executed by the processing unit 32, executes the described of the application and have anti-fraud functional mind
The above-mentioned function of being limited in training method through network model.It should be noted that computer-readable medium described herein
It can be computer-readable signal media or computer readable storage medium either the two any combination.Computer
Readable storage medium storing program for executing for example may be-but not limited to-the system of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, dress
It sets or device, or any above combination.The more specific example of computer readable storage medium can include but is not limited to:
Electrical connection, portable computer diskette, hard disk, random access storage device (RAM), read-only storage with one or more conducting wires
Device (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
In this application, computer readable storage medium can also be any tangible medium for including or store program, should
Program can be commanded execution system, device or device use or in connection.And in this application, computer can
The signal media of reading may include in a base band or as the data-signal that carrier wave a part is propagated, wherein carrying computer
Readable program code.The data-signal of this propagation can take various forms, including but not limited to electromagnetic signal, optical signal
Or above-mentioned any appropriate combination.Computer-readable signal media can also be appointing other than computer readable storage medium
What computer-readable medium, the computer-readable medium can be sent, propagated or transmitted for by instruction execution system, device
Either device use or program in connection.The program code for including on computer-readable medium can be fitted with any
When medium transmission, including but not limited to: wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
One or more programming languages or combinations thereof can be used to write the calculating for executing operation of the invention
Machine program code, described program design language include object oriented program language -- such as Java, Smalltalk, C+
+, it further include conventional procedural programming language-such as " C " language or similar programming language.Program code can
Fully to execute, partly execute on the user computer on the user computer, be executed as an independent software package,
Part executes on the remote computer or executes on a remote computer or server completely on the user computer for part.
In situations involving remote computers, remote computer can pass through the network of any kind --- including local area network (LAN)
Or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as utilize Internet service
Provider is connected by internet).
Flow chart and block diagram in attached drawing of the invention illustrate the system according to the various embodiments of the application, method
With the architecture, function and operation in the cards of computer program product.In this regard, each of flowchart or block diagram
Box can represent a part of a module, program segment or code, and a part of the module, program segment or code includes one
A or multiple executable instructions for implementing the specified logical function.It should also be noted that in some realization sides as replacement
In case, function marked in the box may also be distinct from that the sequence marked in attached drawing occurs.For example, two succeedingly indicate
Box can actually be basically executed in parallel, they can also execute in the opposite order sometimes, herein based on being related to
Function and determine.It is significant to note that in each box and block diagram and or flow chart in block diagram and or flow chart
Box combination, can the dedicated hardware based systems of the functions or operations as defined in executing realize, or can be with
It realizes using a combination of dedicated hardware and computer instructions.
Involved unit can be realized by way of software in an embodiment of the present invention, can also pass through hardware
Mode realize.Described unit also can be set in the processor.
As on the other hand, the fifth embodiment of the present invention additionally provides a kind of computer-readable medium, which can
Reading medium can be included in device described in above-described embodiment；It is also possible to individualism, and without the supplying dress
In setting.Above-mentioned computer-readable medium carries one or more program, and described program specifically includes, and obtains initial pictures,
And analyze the size characteristic and color space structure feature of the initial pictures；Size characteristic and color based on the initial pictures
Color space structure feature determines simulation degree threshold value；And the predetermined pixel value for translating the initial pictures along preset direction, it will
Initial pictures are weighted and averaged with the image after translation and are superimposed, to export required analog image；Wherein, the predetermined pixel value
It is determined based on simulation degree threshold value.
Compared with prior art, a kind of image simulation generation method and its system, deep learning provided by the present invention given
Algorithm training method and electronic equipment have it is following the utility model has the advantages that
Object of the present invention is to which optical parallax is introduced image simulation by optical analog to generate field, so as to based on a small amount of
Basic image data, can be realized the enhancing and extension to image data, and can will generate data set enhance with quasi-
Close truthful data.The present invention is passed by can optically be supplemented in the insufficient situation of data volume based on optical simulation
The problem of system method data amplification cannot cover truth whole possibility.Processing for generating image data has good
Good results of Physical.Training set can be allowed closer to truth using this method, be conducive to the internetworking finally trained
It can be promoted.Specifically, in the present invention, based on the preliminary analysis to initial pictures, simulation degree threshold value is determined, based on simulation journey
Spend threshold value determine to the initial pictures carry out at least two directions translation predetermined pixel value, and by the image after translation into
Row weighted average superposition, to export required analog image.
Image simulation generation method provided by the present invention can solve conventional images analogue technique can not effectively to camera lens,
The technical issues of aberration etc. of the brings image such as optics, photosensitive element is simulated.Meanwhile being based on figure provided by the present invention
Image mould well can not will be generated as data set of the simulation-generation method for generation can also solve available data amplification technique
The problem of intending into true sample graph.As it can be seen that image simulation generation method provided by the present invention has wider applicability.
The present invention also provides a kind of image simulations to generate system and its electronic equipment, has and above-mentioned image simulation generation side
The identical beneficial effect of method, the present invention can be in the insufficient situations of data volume optically by being based on optical simulation
The problem of supplement conventional method data amplification cannot cover truth whole possibility.For generating the place of image data
Reason has good results of Physical.Training set can be allowed closer to truth using this method, be conducive to finally to train
Network performance is promoted.Specifically, in the present invention, it based on the preliminary analysis to initial pictures, determines simulation degree threshold value, is based on
Simulation degree threshold value determines the predetermined pixel value that the initial pictures are carried out with the translation at least two directions, and will be after translation
Image is weighted and averaged superposition, to export required analog image.
The present invention also provides a kind of deep learning algorithm training methods, use image simulation generation method institute as described above
The analog image combination initial pictures of acquisition are trained as training set and are obtained.The deep learning algorithm training method
Conventional method data amplification can be optically supplemented in the insufficient situation of data volume, and cannot to cover truth complete
The problem of portion's possibility.To provide more fully training data for the training of the deep learning algorithm, so that institute can be improved
State the effect of deep learning algorithm training.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in original of the invention
Made any modification within then, equivalent replacement and improvement etc. should all be comprising within protection scope of the present invention.
Claims (10)
1. a kind of image simulation generation method, it is characterised in that: described image simulation-generation method includes:
Step S1 obtains initial pictures, and analyzes the size characteristic and color space structure feature of the initial pictures；Specifically
Ground, can be based on the scale and color space structure of the determination initial pictures, to generate original image or input image to be processed；
Step S2, size characteristic and color space structure feature based on the initial pictures determine simulation degree threshold value；And
Step S3, the predetermined pixel value that the initial pictures are translated along preset direction, by the image after initial pictures and translation
It is weighted and averaged superposition, to export required analog image；Wherein, the predetermined pixel value is determined based on simulation degree threshold value.
2. image simulation generation method as described in the appended claim 1, it is characterised in that: above-mentioned steps S3 specifically includes following step
It is rapid:
Step S31 sets a first direction and a second direction, to the initial pictures successively along the first direction and described
Second direction translates predetermined pixel value, to obtain the first displacement images and the second displacement images respectively；
Step S32 handles first displacement images and second displacement images, to obtain and position coordinates phase
The numerical value matched；And
Step S33, the numerical value that will be matched with same position coordinate in first displacement images and second displacement images
It is weighted and averaged superposition, to export required analog image.
3. image simulation generation method as described in the appended claim 1, it is characterised in that: between above-mentioned steps S2 and step S3 also
Include:
The initial pictures are zoomed in or out processing by step S201；Wherein, the initial pictures zoom in or out processing
Multiple can be set based on the simulation degree threshold value.
4. image simulation generation method as described in the appended claim 1, it is characterised in that: after above-mentioned steps S3, further includes:
Step S4 fills up absent region using reflection model simulation or image interpolation method.
5. image simulation generation method as described in the appended claim 1, it is characterised in that: after above-mentioned steps S2, further include as
Lower step:
Step S3A, analog light source colour temperature are based on simulation degree threshold value, in conjunction with the ratio between black matrix light radiation colour temperature and rgb value
It is matched, the initial pictures is carried out with the adjustment of colour temperature；And/or
The color space structure of the initial pictures is converted to YUV color space by step S3B, analogue exposure comprising Y is logical
Road, the channel Y are luminance channel, are exposed amendment to the channel Y, wherein exposure amendment can be based on the simulation degree threshold
Value is handled；And/or
Step S3C carries out data amplification processing to the initial pictures；
Wherein, above-mentioned steps S3A, step S3B and step S3C optionally one or several and step S3 can carry out random order
Combination.
6. image simulation generation method as claimed in claim 5, it is characterised in that: in above-mentioned steps S3C, the data expand
Increasing processing further comprise: to initial pictures carry out overturning processing, rotation processing, scaling processing, cutting processing, translation processing,
Any one or several combination during noise processed, color channel are handled, image Shear Transform is handled.
7. a kind of image simulation generates system, it is characterised in that: comprising:
Image preliminary analysis module, be configurable for obtain initial pictures, and analyze the initial pictures size characteristic and
Color space structure feature；
Simulation degree threshold value obtains module, is configurable for size characteristic and color space structure based on the initial pictures
Feature determines simulation degree threshold value；And
Analog image generation module is configurable for the predetermined pixel value for translating the initial pictures along preset direction, will
Initial pictures are weighted and averaged with the image after translation and are superimposed, to export required analog image；Wherein, the predetermined pixel value
It is determined based on simulation degree threshold value.
8. image simulation generates system as recited in claim 7, it is characterised in that:
Picture size adjustment module, for the initial pictures to be zoomed in or out processing；
Module is filled up in absent region, for being filled up using reflection model simulation or image interpolation method to absent region；
Analog light source colour temperature module, for being based on simulation degree threshold value, in conjunction with the ratio between black matrix light radiation colour temperature and rgb value
It is matched, the initial pictures is carried out with the adjustment of colour temperature；
Analogue exposure module, for the color space structure of the initial pictures to be converted to YUV color space；And/or
Data expand module, for carrying out data amplification processing to the initial pictures.
9. a kind of deep learning algorithm training method, which is characterized in that the deep learning algorithm uses such as the claims
Any one of 1-6 described image simulation-generation method analog image combination initial pictures obtained are trained as training set
And it obtains.
10. a kind of electronic equipment, it is characterised in that: the electronic equipment includes storage unit and processing unit, and the storage is single
Member is executed for storing computer program, the computer program that the processing unit is used to store by the storage unit as weighed
Benefit requires any one of 1-6 described image simulation-generation method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910634805.4A CN110335330B (en) | 2019-07-12 | 2019-07-12 | Image simulation generation method and system, deep learning algorithm training method and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910634805.4A CN110335330B (en) | 2019-07-12 | 2019-07-12 | Image simulation generation method and system, deep learning algorithm training method and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110335330A true CN110335330A (en) | 2019-10-15 |
| CN110335330B CN110335330B (en) | 2021-04-20 |
Family
ID=68146719
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910634805.4A Active CN110335330B (en) | 2019-07-12 | 2019-07-12 | Image simulation generation method and system, deep learning algorithm training method and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110335330B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111243058A (en) * | 2019-12-31 | 2020-06-05 | 河南裕展精密科技有限公司 | Object simulation image generation method and computer-readable storage medium |
| CN111259968A (en) * | 2020-01-17 | 2020-06-09 | 腾讯科技（深圳）有限公司 | Illegal image recognition method, device, equipment and computer readable storage medium |
| CN111652199A (en) * | 2020-08-04 | 2020-09-11 | 湖南莱博赛医用机器人有限公司 | Image processing method, device, equipment and medium |
| CN111931783A (en) * | 2020-08-18 | 2020-11-13 | 创新奇智(重庆)科技有限公司 | Training sample generation method, machine-readable code identification method and device |
| CN113553985A (en) * | 2021-08-02 | 2021-10-26 | 中再云图技术有限公司 | High-altitude smoke detection and identification method based on artificial intelligence, storage device and server |
| CN113781356A (en) * | 2021-09-18 | 2021-12-10 | 北京世纪好未来教育科技有限公司 | Training method of image denoising model, image denoising method, device and equipment |
| CN113781356B (en) * | 2021-09-18 | 2024-06-04 | 北京世纪好未来教育科技有限公司 | Training method of image denoising model, image denoising method, device and equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101540832A (en) * | 2009-04-24 | 2009-09-23 | 段江 | Methods for matching dynamic range of image signals |
| CN102867318A (en) * | 2012-09-25 | 2013-01-09 | 哈尔滨工业大学 | Construction method of subimage weighted stacking whole discrete degraded image of retracing movement |
| CN103310420A (en) * | 2013-06-19 | 2013-09-18 | 武汉大学 | Method and system for repairing color image holes on basis of texture and geometrical similarities |
| CN105631439A (en) * | 2016-02-18 | 2016-06-01 | 北京旷视科技有限公司 | Human face image collection method and device |
| US20180130248A1 (en) * | 2016-11-04 | 2018-05-10 | Disney Enterprises, Inc. | Noise reduction on g-buffers for monte carlo filtering |
-
2019
- 2019-07-12 CN CN201910634805.4A patent/CN110335330B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101540832A (en) * | 2009-04-24 | 2009-09-23 | 段江 | Methods for matching dynamic range of image signals |
| CN102867318A (en) * | 2012-09-25 | 2013-01-09 | 哈尔滨工业大学 | Construction method of subimage weighted stacking whole discrete degraded image of retracing movement |
| CN103310420A (en) * | 2013-06-19 | 2013-09-18 | 武汉大学 | Method and system for repairing color image holes on basis of texture and geometrical similarities |
| CN105631439A (en) * | 2016-02-18 | 2016-06-01 | 北京旷视科技有限公司 | Human face image collection method and device |
| US20180130248A1 (en) * | 2016-11-04 | 2018-05-10 | Disney Enterprises, Inc. | Noise reduction on g-buffers for monte carlo filtering |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111243058A (en) * | 2019-12-31 | 2020-06-05 | 河南裕展精密科技有限公司 | Object simulation image generation method and computer-readable storage medium |
| CN111243058B (en) * | 2019-12-31 | 2024-03-22 | 富联裕展科技(河南)有限公司 | Object simulation image generation method and computer readable storage medium |
| CN111259968A (en) * | 2020-01-17 | 2020-06-09 | 腾讯科技（深圳）有限公司 | Illegal image recognition method, device, equipment and computer readable storage medium |
| CN111652199A (en) * | 2020-08-04 | 2020-09-11 | 湖南莱博赛医用机器人有限公司 | Image processing method, device, equipment and medium |
| CN111931783A (en) * | 2020-08-18 | 2020-11-13 | 创新奇智(重庆)科技有限公司 | Training sample generation method, machine-readable code identification method and device |
| CN113553985A (en) * | 2021-08-02 | 2021-10-26 | 中再云图技术有限公司 | High-altitude smoke detection and identification method based on artificial intelligence, storage device and server |
| CN113781356A (en) * | 2021-09-18 | 2021-12-10 | 北京世纪好未来教育科技有限公司 | Training method of image denoising model, image denoising method, device and equipment |
| CN113781356B (en) * | 2021-09-18 | 2024-06-04 | 北京世纪好未来教育科技有限公司 | Training method of image denoising model, image denoising method, device and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110335330B (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110335330A (en) | Image simulation generation method and its system, deep learning algorithm training method and electronic equipment | |
| US20210248355A1 (en) | Face key point detection method and apparatus, storage medium, and electronic device | |
| CN112989904B (en) | Method for generating style image, method, device, equipment and medium for training model | |
| CN106157273B (en) | Method and device for generating composite picture | |
| EP4137991A1 (en) | Pedestrian re-identification method and device | |
| CN107407554A (en) | Polyphaser imaging system is emulated | |
| CN111598111B (en) | Three-dimensional model generation method, device, computer equipment and storage medium | |
| CN112241933A (en) | Face image processing method and device, storage medium and electronic equipment | |
| CN108701355A (en) | GPU optimizes and the skin possibility predication based on single Gauss online | |
| CN114418069A (en) | Method and device for training encoder and storage medium | |
| CN110619334A (en) | Portrait segmentation method based on deep learning, architecture and related device | |
| CN115082322B (en) | Image processing method and device, and training method and device of image reconstruction model | |
| CN113298931B (en) | Reconstruction method and device of object model, terminal equipment and storage medium | |
| CN108665498B (en) | Image processing method, device, electronic equipment and storage medium | |
| CN113706400A (en) | Image correction method, image correction device, microscope image correction method, and electronic apparatus | |
| CN110852132B (en) | Two-dimensional code space position confirmation method and device | |
| CN112614110A (en) | Method and device for evaluating image quality and terminal equipment | |
| CN110047126B (en) | Method, apparatus, electronic device, and computer-readable storage medium for rendering image | |
| CN109658360B (en) | Image processing method and device, electronic equipment and computer storage medium | |
| CN106530286A (en) | Method and device for determining definition level | |
| CN110288691A (en) | Render method, apparatus, electronic equipment and the computer readable storage medium of image | |
| CN115147577A (en) | VR scene generation method, device, equipment and storage medium | |
| CN113781653B (en) | Object model generation method and device, electronic equipment and storage medium | |
| CN114663570A (en) | Map generation method and device, electronic device and readable storage medium | |
| CN110196638B (en) | Mobile terminal augmented reality method and system based on target detection and space projection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |