CN106296569B - The method and apparatus for drawing portrait stick figure - Google Patents
The method and apparatus for drawing portrait stick figure Download PDFInfo
- Publication number
- CN106296569B CN106296569B CN201510319160.7A CN201510319160A CN106296569B CN 106296569 B CN106296569 B CN 106296569B CN 201510319160 A CN201510319160 A CN 201510319160A CN 106296569 B CN106296569 B CN 106296569B
- Authority
- CN
- China
- Prior art keywords
- portrait
- stick
- geometric transformation
- pixel
- training pattern
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000009466 transformation Effects 0.000 claims abstract description 104
- 238000012549 training Methods 0.000 claims abstract description 73
- 238000005070 sampling Methods 0.000 claims abstract description 12
- 238000012937 correction Methods 0.000 claims description 5
- 239000004744 fabric Substances 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 38
- 230000008569 process Effects 0.000 description 11
- 230000009467 reduction Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 241000208340 Araliaceae Species 0.000 description 3
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 3
- 235000003140 Panax quinquefolius Nutrition 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 235000008434 ginseng Nutrition 0.000 description 3
- 210000004709 eyebrow Anatomy 0.000 description 2
- 210000004209 hair Anatomy 0.000 description 2
- 210000004279 orbit Anatomy 0.000 description 2
- 238000005530 etching Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000010422 painting Methods 0.000 description 1
Classifications
-
- 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
- G06T3/04—Context-preserving transformations, e.g. by using an importance map
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of methods for drawing portrait stick figure, comprising: preset training pattern;Using training pattern to target portrait geometric transformation, the geometric transformation portrait with standard portrait geometric alignment is obtained;Using preset lines probability distribution area, selected pixels point obtains expectation stick figure to the pixel point sampling that lines probability distribution area is chosen in geometric transformation portrait from geometric transformation portrait;It determines the edge image of geometric transformation portrait, and using edge image amendment expectation stick figure, obtains amendment stick figure;Stick figure is corrected using the stick figure stencil matching of training pattern, generates the corresponding target stick figure of target portrait.The invention also discloses a kind of devices for drawing portrait stick figure.The present invention only handles the pixel of lines probability distribution area selection, reduces the pixel quantity of processing, improves processing speed;Expectation stick figure is corrected using edge image, improves the accuracy of stick figure.
Description
Technical field
The present invention relates to technical field of image processing, it is related specifically to the method and apparatus for drawing portrait stick figure.
Background technique
Currently, many application programs with facial image processing can be installed at the terminal, the portrait that user can be shot
Photo is converted to the image of drawing style, such as is converted to caricature, sketch images, oil colour painting, etching etc..One of conversion regime
It can be the stick figure that portrait photo is converted to portrait profile, the portrait line very much like with portrait depicted by stick figure
Item is drawn.It is to learn to obtain between portrait photo and stick figure by a large amount of sample training that existing portrait stick figure, which draws mode,
Statistical relationship, obtain stick figure template, and desired stick figure is obtained using the nonparametric method of sampling.But using non-ginseng
The number method of sampling needs to carry out neighborhood matching to each upper pixel of portrait photo, calculates the corresponding region of all sample portraits
The similarity of corresponding pixel points is gone up with portrait photo, and in order to draw the portrait stick figure of high quality, existing method needs foot
Enough sample spaces, this matched process is very time-consuming, and computation complexity is high, has seriously affected drafting efficiency.
Summary of the invention
The embodiment of the present invention provides a kind of method and apparatus for drawing portrait stick figure, being capable of Fast Drawing portrait lines
It draws, draws high-efficient.
The embodiment of the present invention proposes a kind of method for drawing portrait stick figure, comprising steps of
Preset training pattern, the training pattern include standard portrait, lines probability distribution area and stick figure mould
Version;
Geometric transformation is carried out to target portrait using training pattern, is obtained and the standard portrait in the training pattern
The geometric transformation portrait of geometric alignment;
Picture is chosen from the geometric transformation portrait using preset lines probability distribution area in the training pattern
Vegetarian refreshments, the pixel chosen in the geometric transformation portrait to the lines probability distribution area sample, it is expected that
Stick figure;
It determines the edge image of the geometric transformation portrait, and the expectation stick figure is corrected using the edge image,
Obtain amendment stick figure;
Stick figure is corrected described in stick figure stencil matching using the training pattern, it is corresponding to generate the target portrait
Target stick figure.
The embodiment of the present invention also proposes a kind of device for drawing portrait stick figure, comprising:
Model generation module, is used for preset training pattern, and the training pattern includes standard portrait, lines probability point
Cloth region and stick figure template;
Geometric transformation module, for carrying out geometric transformation to target portrait using training pattern, acquisition has been instructed with described
Practice the geometric transformation portrait of the standard portrait geometric alignment in model;
Sampling module, for using preset lines probability distribution area in the training pattern from the geometric transformation
Selected pixels point in portrait, the pixel chosen in the geometric transformation portrait to the lines probability distribution area are adopted
Sample obtains expectation stick figure;
Correction module corrects institute for determining the edge image of the geometric transformation portrait, and using the edge image
Desired stick figure is stated, amendment stick figure is obtained;
Stencil matching module, it is raw for correcting stick figure using described in the stick figure stencil matching of training pattern
At the corresponding target stick figure of the target portrait.
The embodiment of the present invention uses preset lines probability distribution area and chooses from the target portrait after geometric transformation
Pixel, and only the pixel of selection is handled, the quantity of processing pixel is greatly reduced, processing speed is improved;
Meanwhile also using edge image amendment expectation stick figure in gray value, enable correct stick figure in lines gray value more
Portrait profile accurately is sketched out, further improves the accuracy of the stick figure ultimately generated.
Detailed description of the invention
Fig. 1 is total line chart of terminal where drawing the device of portrait stick figure in the embodiment of the present invention;
Fig. 2 is the flow chart of the first embodiment for the method that the present invention draws portrait stick figure;
Fig. 3 is that the present invention draws the step process that lines probability distribution area is obtained in the embodiment of the method for portrait stick figure
Figure;
Fig. 4 is the flow chart of the second embodiment for the method that the present invention draws portrait stick figure;
Fig. 5 is the flow chart of the 3rd embodiment for the method that the present invention draws portrait stick figure;
Fig. 6 is the flow chart of the fourth embodiment for the method that the present invention draws portrait stick figure;
Fig. 7 is the module diagram of the first embodiment for the device that the present invention draws portrait stick figure;
Fig. 8 is the module diagram of the second embodiment for the device that the present invention draws portrait stick figure;
Fig. 9 is the superposition schematic diagram of lines probability distribution area in the embodiment of the present invention.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
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, Fig. 1 is total line chart of terminal where drawing the device of portrait stick figure in the embodiment of the present invention, the end
End may include: at least one processor 101, such as CPU, at least one network interface 104, user interface 103, memory
105, at least one communication bus 102.Wherein, communication bus 102 is for realizing the connection communication between these components.Wherein,
User interface 103 may include display screen (Display), keyboard (Keyboard), can also include wireline interface, the nothing of standard
Line interface.Network interface 104 may include standard wireline interface and wireless interface (such as WIFI interface).Memory 105 can be
High speed RAM memory is also possible to non-labile memory (non-volatile memory), for example, at least a disk
Memory.Memory 105 can also be that at least one is located remotely from the storage device of aforementioned processor 101.As a kind of calculating
It may include operating system, network communication module, Subscriber Interface Module SIM and drafting portrait in the memory 105 of machine storage medium
Stick figure program.
In the terminal, network interface 104 is mainly used for connection service for the device institute of drafting portrait stick figure shown in Fig. 1
Device or other terminals carry out data communication with server or other terminals;And user interface 103 is mainly used for reception user and refers to
It enables, and is interacted with user;And processor 101 can be used for calling the drafting portrait stick figure journey stored in memory 105
Sequence, and execute following operation:
Preset training pattern, the training pattern include standard portrait, lines probability distribution area and stick figure mould
Version;
Geometric transformation is carried out to target portrait using training pattern, is obtained and the standard portrait geometry in training pattern
The geometric transformation portrait of alignment;
Using lines probability distribution area selected pixels point from geometric transformation portrait preset in training pattern, to line
The pixel that probability distribution area is chosen in geometric transformation portrait is sampled, and expectation stick figure is obtained;
It determines the edge image of geometric transformation portrait, and using edge image amendment expectation stick figure, obtains amendment lines
It draws;
Stick figure is corrected using the stick figure stencil matching of training pattern, generates the corresponding target lines of target portrait
It draws.
In one embodiment, processor 101 calls in memory 105 the drafting portrait stick figure program that stores can be with
Execute following operation:
Obtain the coordinate of the black pixel point of each sample stick figure of training pattern;
The coordinate of all black pixel points is superimposed, the coordinate set of acquisition constitutes lines probability distribution area.
In one embodiment, processor 101 calls in memory 105 the drafting portrait stick figure program that stores can be with
Execute following operation:
Using lines probability distribution area from edge image selected pixels point;
Corresponding position in the pixel amendment expectation stick figure chosen in edge image using lines probability distribution area
Pixel gray value, obtain amendment stick figure.
In one embodiment, processor 101 calls in memory 105 the drafting portrait stick figure program that stores can be with
Execute following operation:
Neighborhood image matching is carried out to each pixel that lines probability distribution area is chosen in geometric transformation portrait,
Calculate the regional area corresponding to ith pixel point of n sample portrait in training pattern gray value and relative to
The similarity of ith pixel point, and obtain similarity be greater than similar threshold value k sample portrait, as with ith pixel point office
The sample portrait of portion's Region Matching, wherein 1≤n≤20,1≤k≤10;
The corresponding sample stick figure of k sample portrait is obtained, to k sample stick figure corresponding to ith pixel point
The gray value weighted sum of regional area obtains the expectation gray value of ith pixel point;
According to the expectation gray value of each pixel of the geometric transformation portrait in lines probability distribution area, the phase is generated
Hope stick figure.
In one embodiment, processor 101 calls in memory 105 the drafting portrait stick figure program that stores can be with
Execute following operation:
Multiple characteristic points of target portrait are extracted using training pattern;
Geometric transformation is carried out to target portrait according to characteristic point, makes the characteristic point of the geometric transformation portrait obtained and has trained
The feature point alignment of standard portrait in model.
The described device and its place terminal for drawing portrait stick figure of the present embodiment Fig. 1, uses preset lines
Probability distribution area selected pixels point from the target portrait after geometric transformation, and only the pixel of selection is handled, greatly
The quantity for reducing processing pixel greatly, improves processing speed;Meanwhile also using in edge image amendment expectation stick figure
Gray value enables the lines gray value corrected in stick figure more accurately to sketch out portrait profile, further improves final
The accuracy of the stick figure of generation.
As shown in Fig. 2, Fig. 2 is the flow chart of the first embodiment for the method that the present invention draws portrait stick figure.This implementation
The method for the drafting portrait stick figure that example is mentioned, comprising steps of
Step S09, preset training pattern, the training pattern include standard portrait, lines probability distribution area and
Stick figure template;
In the present embodiment, multiple groups sample portrait and the modeling of sample stick figure are first passed through in advance, obtains the model trained.It is building
In mold process, the feature point for calibration first on each sample portrait calculates the mean place of the characteristic point of each sample portrait, root
Standard portrait is obtained according to the mean place of characteristic point, then each group sample portrait and its corresponding sample stick figure are done into geometry change
Change, the sample portrait and sample stick figure after making geometric transformation all with standard portrait geometric alignment, then according to geometric transformation after
Each sample lines draw the control parameters of each lines, generate unified stick figure template, complete modeling.
Step S10 carries out geometric transformation to target portrait using training pattern, obtains and the standard in training pattern
The geometric transformation portrait of portrait geometric alignment;
It, can be by the photo of terminal photographic subjects portrait or from terminal when user needs terminal for target portrait stick figure
The photo of target portrait is obtained in memory.The training pattern that terminal is obtained by aforementioned modeling is to the target portrait in photo
Geometric transformation is carried out, target portrait and the standard portrait geometric alignment in training pattern are made.
Step S20 chooses picture from geometric transformation portrait using lines probability distribution area preset in training pattern
Vegetarian refreshments;
In view of target portrait is after geometric transformation, pixel required for drafting stick figure is distributed in limited sky
Between in, for example, then only needing face, eyes, nose, mouth, eyebrow, ear to the face in photo is depicted as stick figure
Piece, the profile at the positions such as hair style, therefore lines probability distribution area can be preset, will be used to draw stick figure in portrait
Necessary pixel is included in this lines probability distribution area, and uses lines probability distribution area from geometric transformation people
The selected pixels point as in.
Step S30, the pixel chosen in geometric transformation portrait to lines probability distribution area sample, and obtain the phase
Hope stick figure;
To show that gray value needed for stick figure, the present embodiment carry out geometric transformation portrait using non-parametric sampling mode
Processing obtains expectation stick figure.To improve processing speed, it is only necessary to be carried out to the pixel that lines probability distribution area is chosen non-
Parametrization sampling, advantageously reduces the quantity of processing pixel, improves processing speed.
Step S40 determines the edge image of geometric transformation portrait;
Step S50 corrects expectation stick figure using edge image, obtains amendment stick figure;
In order to further enable it is expected that the gray value of each pixel in stick figure more accurately draws out clearly line
Item is drawn, and the present embodiment is also added into boundary operator to obtain the edge image of geometric transformation portrait, obtains each main spy in portrait
Profile of sign, such as eye socket, eye benevolence, the profile of nose, nozzle type, face mask etc., and using edge image amendment expectation lines
The gray value of pixel in picture obtains amendment stick figure, the lines gray value corrected in stick figure is enable more accurately to delineate
Portrait profile out.
Step S60 corrects stick figure using the stick figure stencil matching of training pattern, generates the corresponding mesh of target portrait
Mark stick figure.
Due to obtained amendment stick figure there is also part without between the lines or lines drawn there are interdependence or
The rule of repulsion, according to the rule between the lines that stick figure template is included in model, to adjust the control ginseng of amendment stick figure
Number, clashes some unnecessary lines or gray value, and prominent some necessary lines ultimately generate optimal target lines
It draws.Further, since the present embodiment carries out geometric transformation to target portrait using training pattern in advance, therefore can also will match
Stick figure afterwards carries out the inverse transformation of geometric transformation, obtains the target stick figure with original target portrait geometric alignment, so that
The stick figure arrived is truer.
The present embodiment uses preset lines probability distribution area selected pixels from the target portrait after geometric transformation
Point, and only the pixel of selection is handled, the quantity of processing pixel is greatly reduced, processing speed is improved;Meanwhile
Also using the gray value in edge image amendment expectation stick figure, keep the lines gray value corrected in stick figure more accurate
Portrait profile is sketched out, the accuracy of the stick figure ultimately generated is further improved.
As shown in figure 3, Fig. 3 is that the present invention draws acquisition lines probability distribution area in the embodiment of the method for portrait stick figure
Step flow chart.For the present embodiment for the lines probability distribution area in the step S09 of embodiment illustrated in fig. 2, use is following
Step obtains:
Step S71 obtains the coordinate of the black pixel point of each sample stick figure of training pattern;
The coordinate of all black pixel points is superimposed by step S72, and the coordinate set of acquisition constitutes lines probability distribution area
Domain.
The present embodiment improves processing speed, during modeling training, will train to reduce the pixel quantity of processing
All sample stick figures are overlapped in model, merge the coordinate of the black pixel point of each sample stick figure, are owned
The coordinate set of black pixel point in sample stick figure, as shown in figure 9, the left side is each sample stick figure before superposition in Fig. 9, it is right
Side is the coordinate set of the black pixel point obtained after being superimposed, the pixel for including, to draw in the region of this coordinate set
Pixel necessary to stick figure processed.Due to using preset lines probability distribution area from the target portrait after geometric transformation
Middle selected pixels point, and only the pixel of selection is handled, the quantity of processing pixel is greatly reduced, processing is improved
Speed.
As shown in figure 4, Fig. 4 is the flow chart of the second embodiment for the method that the present invention draws portrait stick figure.This implementation
Example includes all steps in embodiment illustrated in fig. 2, and wherein step S50 includes:
Step S51, using lines probability distribution area from edge image selected pixels point;
Step S52, the pixel amendment chosen in edge image using lines probability distribution area it is expected in stick figure
The pixel gray value of corresponding position obtains amendment stick figure.
In the present embodiment, to make it is expected that the pixel gray value in stick figure more accurately delineates portrait profile, using side
The pixel of edge image corrects the gray value of desired stick figure, makes it is expected that lines not black enough in stick figure are more clear, can
It is main to choose the black pixel point in edge image to correct the gray value of desired stick figure.In addition, in order to reduce the picture of processing
Lines probability distribution area selected pixels point from edge image can be used, only with the pixel in selected range in vegetarian refreshments quantity
Point is conducive to further increase processing speed to correct desired stick figure.
As shown in figure 5, Fig. 5 is the flow chart of the 3rd embodiment for the method that the present invention draws portrait stick figure.This implementation
Example includes all steps in embodiment illustrated in fig. 2, and wherein step S30 includes:
Step S31 carries out neighborhood to each pixel that lines probability distribution area is chosen in geometric transformation portrait
Images match, calculate the gray value of the regional area corresponding to ith pixel point of n sample portrait in training pattern with
And the similarity relative to ith pixel point, and obtain similarity be greater than similar threshold value k sample portrait, as with i-th
The matched sample portrait of pixel regional area, wherein 1≤n≤20,1≤k≤10;
Step S32 obtains the corresponding sample stick figure of k sample portrait, to k sample stick figure corresponding to i-th
The gray value weighted sum of the regional area of pixel obtains the expectation gray value of ith pixel point;
Step S33, according to the expectation gray scale of each pixel of the geometric transformation portrait in lines probability distribution area
Value generates expectation stick figure.
The present embodiment is to further increase processing speed, reduces the sample size for neighborhood matching.Due to existing
In neighborhood matching method, when sample size selection is very few, much noise can be generated, makes to generate many broken strings in the lines drawn
Part.But due to after the present embodiment extended meeting the phase that the present embodiment obtains is corrected using the edge image that boundary operator obtains
It hopes stick figure, makes it is expected that the disconnection portion of lines in stick figure gets repairing, therefore the present embodiment can reduce sample size, by sample
This quantity n is reduced within 20, or even can be reduced to 10, only by the pixel of a small amount of sample portrait and geometric transformation portrait
Point carries out neighborhood matching, obtains similarity of this portrait of various kinds relative to each pixel, and by the higher k sample people of similarity
As the matched sample portrait with the pixel.Due to the reduction of sample size, correspondingly, k value can also be reduced, such as
It is reduced within 10, or even is reduced to 3, obtained expectation stick figure will not influence most after edge image is corrected
The image effect obtained eventually.In this way, have processing speed due to the reduction of sample size and the reduction of k value
Effect improves.
As shown in fig. 6, Fig. 6 is the flow chart of the fourth embodiment for the method that the present invention draws portrait stick figure.This implementation
Example includes all steps in embodiment illustrated in fig. 2, and wherein step S10 includes:
Step S11 extracts multiple characteristic points of target portrait using training pattern;
Step S12 carries out geometric transformation to target portrait according to characteristic point, makes the characteristic point of the geometric transformation portrait obtained
With the feature point alignment of the standard portrait in training pattern.
The present embodiment, according to the characteristic point demarcated on each sample portrait, has been trained in modeling process early period
Therefore model when needing target portrait stick figure, directlys adopt the feature of the i.e. extractable target person picture of training pattern
Point.When to target portrait geometric transformation, the feature point alignment of the characteristic point of geometric transformation portrait and standard portrait obtains several
What transformation portrait.To target portrait geometric transformation by the way of feature point alignment, be conducive to reduce present in conversion process
Error reduces the error of subsequent processing, keeps the stick figure drawn more accurate.
As shown in fig. 7, Fig. 7 is the module diagram of the first embodiment for the device that the present invention draws portrait stick figure.This
The device for the drafting portrait stick figure that embodiment is mentioned, comprising:
Model generation module 110 is used for preset training pattern, and training pattern includes standard portrait, lines probability point
Cloth region and stick figure template;
Geometric transformation module 120 is obtained and has been trained for carrying out geometric transformation to target portrait using training pattern
The geometric transformation portrait of standard portrait geometric alignment in model;
Sampling module 130, for using preset lines probability distribution area in training pattern from geometric transformation portrait
Middle selected pixels point, the pixel chosen in geometric transformation portrait to lines probability distribution area sample, it is expected that
Stick figure;
Correction module 140, for determining the edge image of geometric transformation portrait, and using edge image amendment expectation lines
It draws, obtains amendment stick figure;
Stencil matching module 150 generates target for correcting stick figure using the stick figure stencil matching of training pattern
The corresponding target stick figure of portrait.
In the present embodiment, multiple groups sample portrait and the modeling of sample stick figure are first passed through in advance, obtains the model trained.It is building
In mold process, the feature point for calibration first on each sample portrait calculates the mean place of the characteristic point of each sample portrait, root
Standard portrait is obtained according to the mean place of characteristic point, then each group sample portrait and its corresponding sample stick figure are done into geometry change
Change, the sample portrait and sample stick figure after making geometric transformation all with standard portrait geometric alignment, then according to geometric transformation after
Each sample lines draw the control parameters of each lines, generate unified stick figure template, complete modeling.
It, can be by the photo of terminal photographic subjects portrait or from terminal when user needs terminal for target portrait stick figure
The photo of target portrait is obtained in memory.The training pattern that terminal is obtained by aforementioned modeling is to the target portrait in photo
Geometric transformation is carried out, target portrait and the standard portrait geometric alignment in training pattern are made.
In view of target portrait is after geometric transformation, pixel required for drafting stick figure is distributed in limited sky
Between in, for example, then only needing face, eyes, nose, mouth, eyebrow, ear to the face in photo is depicted as stick figure
Piece, the profile at the positions such as hair style, therefore lines probability distribution area can be preset, will be used to draw stick figure in portrait
Necessary pixel is included in this lines probability distribution area, and uses lines probability distribution area from geometric transformation people
The selected pixels point as in.
To show that gray value needed for stick figure, the present embodiment carry out geometric transformation portrait using non-parametric sampling mode
Processing obtains expectation stick figure.To improve processing speed, it is only necessary to be carried out to the pixel that lines probability distribution area is chosen non-
Parametrization sampling, advantageously reduces the quantity of processing pixel, improves processing speed.
In order to further enable it is expected that the gray value of each pixel in stick figure more accurately draws out clearly line
Item is drawn, and the present embodiment is also added into boundary operator to obtain the edge image of geometric transformation portrait, obtains each main spy in portrait
Profile of sign, such as eye socket, eye benevolence, the profile of nose, nozzle type, face mask etc., and using edge image amendment expectation lines
The gray value of pixel in picture obtains amendment stick figure, the lines gray value corrected in stick figure is enable more accurately to delineate
Portrait profile out.
Due to obtained amendment stick figure there is also part without between the lines or lines drawn there are interdependence or
The rule of repulsion, according to the rule between the lines that stick figure template is included in model, to adjust the control ginseng of amendment stick figure
Number, clashes some unnecessary lines or gray value, and prominent some necessary lines ultimately generate optimal target lines
It draws.Further, since the present embodiment carries out geometric transformation to target portrait using training pattern in advance, therefore can also will match
Stick figure afterwards carries out the inverse transformation of geometric transformation, obtains the target stick figure with original target portrait geometric alignment, so that
The stick figure arrived is truer.
The present embodiment uses preset lines probability distribution area selected pixels from the target portrait after geometric transformation
Point, and only the pixel of selection is handled, the quantity of processing pixel is greatly reduced, processing speed is improved;Meanwhile
Also using the gray value in edge image amendment expectation stick figure, keep the lines gray value corrected in stick figure more accurate
Portrait profile is sketched out, the accuracy of the stick figure ultimately generated is further improved.
As shown in figure 8, Fig. 8 is the module diagram of the second embodiment for the device that the present invention draws portrait stick figure.This
Embodiment includes all modules in embodiment illustrated in fig. 7, and the region also added for obtaining lines probability distribution area is pre-
If module 160, which is used for:
Obtain the coordinate of the black pixel point of each sample stick figure of training pattern;
The coordinate of all black pixel points is superimposed, the coordinate set of acquisition constitutes lines probability distribution area.
The present embodiment improves processing speed, during modeling training, will train to reduce the pixel quantity of processing
All sample stick figures are overlapped in model, merge the coordinate of the black pixel point of each sample stick figure, are owned
The coordinate set of black pixel point in sample stick figure, as shown in figure 9, the left side is each sample stick figure before superposition in Fig. 9, it is right
Side is the coordinate set of the black pixel point obtained after being superimposed, the pixel for including, to draw in the region of this coordinate set
Pixel necessary to stick figure processed.Due to using preset lines probability distribution area from the target portrait after geometric transformation
Middle selected pixels point, and only the pixel of selection is handled, the quantity of processing pixel is greatly reduced, processing is improved
Speed.
Further, correction module 140 is also used to:
Using lines probability distribution area from edge image selected pixels point;
Corresponding position in the pixel amendment expectation stick figure chosen in edge image using lines probability distribution area
Pixel gray value, obtain amendment stick figure.
In the present embodiment, to make it is expected that the pixel gray value in stick figure more accurately delineates portrait profile, using side
The pixel of edge image corrects the gray value of desired stick figure, makes it is expected that lines not black enough in stick figure are more clear, can
It is main to choose the black pixel point in edge image to correct the gray value of desired stick figure.In addition, in order to reduce the picture of processing
Lines probability distribution area selected pixels point from edge image can be used, only with the pixel in selected range in vegetarian refreshments quantity
Point is conducive to further increase processing speed to correct desired stick figure.
Further, sampling module 130 is also used to:
Neighborhood image matching is carried out to each pixel that lines probability distribution area is chosen in geometric transformation portrait,
Calculate the regional area corresponding to ith pixel point of n sample portrait in training pattern gray value and relative to
The similarity of ith pixel point, and obtain similarity be greater than similar threshold value k sample portrait, as with ith pixel point office
The sample portrait of portion's Region Matching, wherein 1≤n≤20,1≤k≤10;
The corresponding sample stick figure of k sample portrait is obtained, to k sample stick figure corresponding to ith pixel point
The gray value weighted sum of regional area obtains the expectation gray value of ith pixel point;
According to the expectation gray value of each pixel of the geometric transformation portrait in lines probability distribution area, the phase is generated
Hope stick figure.
The present embodiment is to further increase processing speed, reduces the sample size for neighborhood matching.Due to existing
In neighborhood matching method, when sample size selection is very few, much noise can be generated, makes to generate many broken strings in the lines drawn
Part.But due to after the present embodiment extended meeting the phase that the present embodiment obtains is corrected using the edge image that boundary operator obtains
It hopes stick figure, makes it is expected that the disconnection portion of lines in stick figure gets repairing, therefore the present embodiment can reduce sample size, by sample
This quantity n is reduced within 20, or even can be reduced to 10, only by the pixel of a small amount of sample portrait and geometric transformation portrait
Point carries out neighborhood matching, obtains similarity of this portrait of various kinds relative to each pixel, and by the higher k sample people of similarity
As the matched sample portrait with the pixel.Due to the reduction of sample size, correspondingly, k value can also be reduced, such as
It is reduced within 10, or even is reduced to 3, obtained expectation stick figure will not influence most after edge image is corrected
The image effect obtained eventually.In this way, have processing speed due to the reduction of sample size and the reduction of k value
Effect improves.
Further, geometric transformation module 120 is also used to:
Multiple characteristic points of target portrait are extracted using training pattern;
Geometric transformation is carried out to target portrait according to characteristic point, makes the characteristic point of the geometric transformation portrait obtained and has trained
The feature point alignment of standard portrait in model.
The present embodiment, according to the characteristic point demarcated on each sample portrait, has been trained in modeling process early period
Therefore model when needing target portrait stick figure, directlys adopt the feature of the i.e. extractable target person picture of training pattern
Point.When to target portrait geometric transformation, the feature point alignment of the characteristic point of geometric transformation portrait and standard portrait obtains several
What transformation portrait.To target portrait geometric transformation by the way of feature point alignment, be conducive to reduce present in conversion process
Error reduces the error of subsequent processing, keeps the stick figure drawn more accurate.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or the device that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or device institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, method of element, article or device.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in a storage medium
In (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that a terminal device (can be mobile phone, computer, clothes
Business device or the network equipment etc.) execute method described in each embodiment of the present invention.
The above description is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all utilizations
Equivalent structure or equivalent flow shift made by description of the invention and accompanying drawing content is applied directly or indirectly in other correlations
Technical field, be included within the scope of the present invention.
Claims (8)
1. a kind of method for drawing portrait stick figure, which is characterized in that comprising steps of
Preset training pattern, the training pattern include standard portrait, lines probability distribution area and stick figure template,
In, the lines probability distribution area is obtained using following steps: obtaining each sample stick figure of training pattern
The coordinate of all black pixel points is superimposed by the coordinate of black pixel point, and the coordinate set of acquisition constitutes the lines probability point
Cloth region;
Geometric transformation is carried out to target portrait using training pattern, is obtained and the standard portrait geometry in the training pattern
The geometric transformation portrait of alignment;
Using preset lines probability distribution area in the training pattern from the geometric transformation portrait selected pixels point,
The pixel chosen in the geometric transformation portrait according to the lines probability distribution area is sampled, expectation line is obtained
Item is drawn;
It determines the edge image of the geometric transformation portrait, and the expectation stick figure is corrected using the edge image, obtain
Correct stick figure;
Stick figure is corrected described in stick figure stencil matching using the training pattern, generates the corresponding mesh of the target portrait
Mark stick figure.
2. drawing the method for portrait stick figure as described in claim 1, which is characterized in that described to correct institute using edge image
Desired stick figure is stated, obtaining the step of correcting stick figure includes:
Using the lines probability distribution area from the edge image selected pixels point;
It is corrected in the expectation stick figure using the pixel that the lines probability distribution area is chosen in the edge image
The pixel gray value of corresponding position obtains the amendment stick figure.
3. drawing the method for portrait stick figure as claimed in claim 1 or 2, which is characterized in that described to according to lines probability
The pixel that distributed areas are chosen in the geometric transformation portrait is sampled, and is obtained the step of it is expected stick figure and is included:
Neighborhood image is carried out to each pixel that the lines probability distribution area is chosen in the geometric transformation portrait
Matching, calculate the gray value of the regional area corresponding to ith pixel point of n sample portrait in the training pattern with
And the similarity relative to ith pixel point, and obtain similarity be greater than similar threshold value k sample portrait, as with i-th
The matched sample portrait of pixel regional area, wherein 1≤n≤20,1≤k≤10;
The corresponding sample stick figure of k sample portrait is obtained, to the part corresponding to ith pixel point of k sample stick figure
The gray value weighted sum in region obtains the expectation gray value of ith pixel point;
It is raw according to the expectation gray value of each pixel of the geometric transformation portrait in the lines probability distribution area
At desired stick figure.
4. drawing the method for portrait stick figure as claimed in claim 1 or 2, which is characterized in that described using training pattern
Geometric transformation is carried out to target portrait, obtains the geometric transformation portrait with the standard portrait geometric alignment in the training pattern
The step of include:
Multiple characteristic points of the target portrait are extracted using training pattern;
Geometric transformation is carried out to the target portrait according to the characteristic point, makes the characteristic point of the geometric transformation portrait obtained
With the feature point alignment of the standard portrait in the training pattern.
5. a kind of device for drawing portrait stick figure characterized by comprising
Model generation module is used for preset training pattern, and the training pattern includes standard portrait, lines probability distribution area
Domain and stick figure template;
Geometric transformation module obtains for carrying out geometric transformation to target portrait using training pattern and has trained mould with described
The geometric transformation portrait of standard portrait geometric alignment in type;
Sampling module, for using preset lines probability distribution area in the training pattern from the geometric transformation portrait
Middle selected pixels point, adopts the pixel chosen in the geometric transformation portrait according to the lines probability distribution area
Sample obtains expectation stick figure;
Correction module corrects the phase for determining the edge image of the geometric transformation portrait, and using the edge image
It hopes stick figure, obtains amendment stick figure;
Stencil matching module generates institute for correcting stick figure using described in the stick figure stencil matching of training pattern
State the corresponding target stick figure of target portrait;
Described device further includes region presetting module, is used for: obtaining the black of each sample stick figure of training pattern
The coordinate of pixel;The coordinate of all black pixel points is superimposed, the coordinate set of acquisition constitutes lines probability distribution area
Domain.
6. drawing the device of portrait stick figure as claimed in claim 5, which is characterized in that the correction module is also used to:
Using the lines probability distribution area from the edge image selected pixels point;
It is corrected in the expectation stick figure using the pixel that the lines probability distribution area is chosen in the edge image
The pixel gray value of corresponding position obtains the amendment stick figure.
7. such as the device described in claim 5 or 6 for drawing portrait stick figure, which is characterized in that the sampling module is also used to:
Neighborhood image is carried out to each pixel that the lines probability distribution area is chosen in the geometric transformation portrait
Matching, calculate the gray value of the regional area corresponding to ith pixel point of n sample portrait in the training pattern with
And the similarity relative to ith pixel point, and obtain similarity be greater than similar threshold value k sample portrait, as with i-th
The matched sample portrait of pixel regional area, wherein 1≤n≤20,1≤k≤10;
The corresponding sample stick figure of k sample portrait is obtained, to the part corresponding to ith pixel point of k sample stick figure
The gray value weighted sum in region obtains the expectation gray value of ith pixel point;
It is raw according to the expectation gray value of each pixel of the geometric transformation portrait in the lines probability distribution area
At desired stick figure.
8. such as the device described in claim 5 or 6 for drawing portrait stick figure, which is characterized in that the geometric transformation module is also
For:
Multiple characteristic points of the target portrait are extracted using training pattern;
Geometric transformation is carried out to the target portrait according to the characteristic point, makes the characteristic point of the geometric transformation portrait obtained
With the feature point alignment of the standard portrait in the training pattern.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510319160.7A CN106296569B (en) | 2015-06-11 | 2015-06-11 | The method and apparatus for drawing portrait stick figure |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510319160.7A CN106296569B (en) | 2015-06-11 | 2015-06-11 | The method and apparatus for drawing portrait stick figure |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106296569A CN106296569A (en) | 2017-01-04 |
CN106296569B true CN106296569B (en) | 2019-08-02 |
Family
ID=57660294
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510319160.7A Active CN106296569B (en) | 2015-06-11 | 2015-06-11 | The method and apparatus for drawing portrait stick figure |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106296569B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109727299A (en) * | 2018-11-20 | 2019-05-07 | 广东智媒云图科技股份有限公司 | A kind of control mechanical arm combines the method drawn a picture, electronic equipment and storage medium |
CN113487697B (en) * | 2021-07-20 | 2024-07-19 | 维沃移动通信(杭州)有限公司 | Method and device for generating simple drawing, electronic equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6845171B2 (en) * | 2001-11-19 | 2005-01-18 | Microsoft Corporation | Automatic sketch generation |
CN101159064A (en) * | 2007-11-29 | 2008-04-09 | 腾讯科技(深圳)有限公司 | Image generation system and method for generating image |
CN104123741A (en) * | 2014-06-24 | 2014-10-29 | 小米科技有限责任公司 | Method and device for generating human face sketch |
CN104484650A (en) * | 2014-12-09 | 2015-04-01 | 北京信息科技大学 | Method and device for identifying sketch face |
-
2015
- 2015-06-11 CN CN201510319160.7A patent/CN106296569B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6845171B2 (en) * | 2001-11-19 | 2005-01-18 | Microsoft Corporation | Automatic sketch generation |
CN101159064A (en) * | 2007-11-29 | 2008-04-09 | 腾讯科技(深圳)有限公司 | Image generation system and method for generating image |
CN104123741A (en) * | 2014-06-24 | 2014-10-29 | 小米科技有限责任公司 | Method and device for generating human face sketch |
CN104484650A (en) * | 2014-12-09 | 2015-04-01 | 北京信息科技大学 | Method and device for identifying sketch face |
Non-Patent Citations (2)
Title |
---|
Example- based Facial Sketch Generation with Non-parametric Sampling;Hong Chen等;《ICCV 2001》;20010714;433-438 |
基于学习的人脸表情动画生成方法研究;刘莎;《中国优秀硕士学位论文全文数据库信息科技辑》;20140115;I138-1960 |
Also Published As
Publication number | Publication date |
---|---|
CN106296569A (en) | 2017-01-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107146199B (en) | Fusion method and device of face images and computing equipment | |
JP7129502B2 (en) | Face image processing method and device, image equipment and storage medium | |
CN109829930B (en) | Face image processing method and device, computer equipment and readable storage medium | |
CN111383232B (en) | Matting method, matting device, terminal equipment and computer readable storage medium | |
US11200680B2 (en) | Image processing method and apparatus | |
CN107564080B (en) | Face image replacement system | |
JP2020530920A (en) | Image lighting methods, devices, electronics and storage media | |
WO2019187298A1 (en) | Image processing system and image processing method | |
JP5387193B2 (en) | Image processing system, image processing apparatus, and program | |
CN106709931B (en) | Method for mapping facial makeup to face and facial makeup mapping device | |
CN107507217A (en) | Preparation method, device and the storage medium of certificate photo | |
WO2022143354A1 (en) | Face generation method and apparatus for virtual object, and device and readable storage medium | |
KR20080028282A (en) | Image processing apparatus, image processing method, and recording medium storing image processing program | |
CN111353956A (en) | Image restoration method and device, computer equipment and storage medium | |
CN106296569B (en) | The method and apparatus for drawing portrait stick figure | |
CN112347288A (en) | Character and picture vectorization method | |
JP7421652B2 (en) | How to identify the authenticity of an object | |
US10354125B2 (en) | Photograph processing method and system | |
CN110458754B (en) | Image generation method and terminal equipment | |
CN110992371A (en) | Portrait segmentation method and device based on prior information and electronic equipment | |
CN106485648A (en) | Image acquisition device with correction function and correction method of image acquisition device | |
US11900564B2 (en) | Storage medium storing program, image processing apparatus, and training method of machine learning model | |
US11200708B1 (en) | Real-time color vector preview generation | |
CN114782600A (en) | Video specific area rendering system and rendering method based on auxiliary grid | |
CN114049670A (en) | Image processing method, image processing device, electronic equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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 |