CN106340023B - The method and apparatus of image segmentation - Google Patents
The method and apparatus of image segmentation Download PDFInfo
- Publication number
- CN106340023B CN106340023B CN201610702051.8A CN201610702051A CN106340023B CN 106340023 B CN106340023 B CN 106340023B CN 201610702051 A CN201610702051 A CN 201610702051A CN 106340023 B CN106340023 B CN 106340023B
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- target area
- area
- characteristic
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
Abstract
The invention proposes a kind of methods of image segmentation, the described method includes: obtaining original target image to be split, extract the geometrical characteristic in region to be split in the original target image, obtain the color characteristic of each pixel in the original target image, each pixel in original target image is clustered according to the color characteristic of the geometrical characteristic of extraction and each pixel, is split the original target image according to the result of cluster.Each pixel in original target image is clustered by combining the color characteristic of geometrical characteristic and each pixel, accordingly even when the color characteristic in region to be split is unobvious, since region also can be accurately split by the limitation of geometrical characteristic.Furthermore, it is also proposed that a kind of device of image segmentation.
Description
Technical field
The present invention relates to field of image processings, more particularly to a kind of method and apparatus of image segmentation.
Background technique
With the development of image technique, how image is split and is become more and more important.Traditional area to image
Domain is split there are mainly two types of method, and one is the dividing methods based on color characteristic, and one is be based on edge feature and energy
Measure the dividing method of functional.Wherein, higher requirement is had based on depth of the dividing method of color characteristic to color, if segmentation
The color characteristic of target is unobvious, and it will cause segmentations to fail.For example, the very slight color when lip is almost consistent with face's colour of skin
When, with this dividing method just can not successful division go out lip region.Energy functional is substantially exactly first to establish description region collection
The parameter expression for closing feature, changes the shape in region by adjustment parameter, when the energy functional of definition is minimized, table
Up to the completely realistic edges of regions of geometric figure represented by formula, the disadvantage is that the geometrical characteristic in region is more complicated
In the case of resume parameter expression it is relatively difficult, and parameter is numerous, parameter is numerous itself will lead to again energy function solve it is minimum
The difficulty of value is very big, and the operational efficiency of algorithm also becomes very low, and its discribed curve is usually continuous and smooth
, and the boundary of actually certain cutting objects itself is not smooth, so region segmentation effect is bad.
Summary of the invention
Based on this, it is necessary in view of the above-mentioned problems, providing a kind of method and apparatus of more accurate image segmentation
A kind of method of image segmentation, which comprises
Obtain original target image to be split；
Extract the geometrical characteristic in region to be split in the original target image；
Obtain the color characteristic of each pixel in the original target image；
According to the color characteristic of the geometrical characteristic of extraction and each pixel to each in original target image
A pixel is clustered；
The original target image is split according to the result of cluster.
A kind of device of image segmentation, described device include:
Image collection module, for obtaining original target image to be split；
Extraction module, for extracting the geometrical characteristic in region to be split in the original target image；
Color characteristic obtains module, for obtaining the color characteristic of each pixel in the original target image；
First cluster module, for according to the geometrical characteristic of extraction and the color characteristic of each pixel to original
Each pixel in beginning target image is clustered；
Divide module, for being split the original target image according to the result of cluster.
The method and apparatus of above-mentioned image segmentation extract original object figure by obtaining original target image to be split
The geometrical characteristic in region to be split as in, obtains the color characteristic of each pixel in original target image, according to the several of extraction
The color characteristic of what feature and each pixel clusters each pixel in original target image.Above-mentioned image segmentation
Method, by by the color characteristic of geometrical characteristic and each pixel combine in original target image each pixel click through
Row cluster, accordingly even when the color characteristic in region to be split is unobvious, since the limitation of geometrical characteristic also can be accurately by region
It is split.Therefore, by by the geometrical characteristic in region to be split and color characteristic combine can it is more accurate to region into
Row segmentation.
Detailed description of the invention
Fig. 1 is the schematic diagram of internal structure of terminal in one embodiment；
Fig. 2 is the method flow diagram of image segmentation in one embodiment；
Fig. 3 is the schematic diagram that the point of Face geometric eigenvector is extracted in one embodiment；
Fig. 4 is the method stream clustered according to the geometrical characteristic and color characteristic of extraction to pixel in one embodiment
Cheng Tu；
Fig. 5 A is the schematic diagram of the original target image in one embodiment；
Fig. 5 B is that lip-region fills and carries out the schematic diagram after Fuzzy Processing in one embodiment；
Fig. 6 is the method flow diagram clustered according to color characteristic and brightness value to pixel in one embodiment；
Fig. 7 is the method flow diagram for carrying out Fuzzy Processing in one embodiment to image；
Fig. 8 is the structural block diagram of the device of image segmentation in one embodiment；
Fig. 9 is the structural block diagram of the first cluster module in one embodiment；
Figure 10 is the structural block diagram of the second cluster module in one embodiment；
Figure 11 is the structural block diagram of Fuzzy Processing module in one embodiment.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
As shown in Figure 1, in one embodiment, the internal structure of terminal 100 is as shown in Figure 1, include passing through system bus
Processor, built-in storage, non-volatile memory medium, network interface, image collecting device, display screen and the input dress of connection
It sets.Wherein, the non-volatile memory medium of terminal 100 is stored with operating system, further includes a kind of device of image segmentation, the figure
As segmentation device for realizing a kind of image segmentation method.The processor supports whole for providing calculating and control ability
The operation of a terminal.Built-in storage in terminal provides ring for the operation of the device of the image segmentation in non-volatile memory medium
There is computer-readable instruction in border in the built-in storage, when which is executed by processor, may make processor
A kind of method for executing image segmentation.Network interface is communicated for being connected to network.Image collecting device is for image
Acquisition, for example carry out the typing of image.The display screen of terminal can be liquid crystal display or electric ink display screen etc., input
Device can be the touch layer covered on display screen, be also possible to the key being arranged on electronic equipment casing, trace ball or touch-control
Plate is also possible to external keyboard, Trackpad or mouse etc..The terminal can be mobile phone, tablet computer, laptop, platform
Formula computer etc..It will be understood by those skilled in the art that structure shown in Fig. 1, only portion relevant to application scheme
The block diagram of separation structure, does not constitute the restriction for the terminal being applied thereon to application scheme, and specific terminal may include
Than more or fewer components as shown in the figure, certain components are perhaps combined or with different component layouts.
As shown in Fig. 2, in one embodiment it is proposed that a kind of method of image segmentation, this method comprises:
Step 202, original target image to be split is obtained.
Specifically, original target image can be color image, it is also possible to gray level image.Original target image is carried out
The premise of segmentation is that there is in the region to be split in image certain geometry to limit.
Step 204, the geometrical characteristic in region to be split in original target image is extracted.
In the present embodiment, the geometry that region to be split is extracted according to the geometry that region to be split itself has is special
Sign.Extracting the geometrical characteristic in region to be split, there are many kinds of methods, and different methods can be taken for different images.Its
In, for facial image, the point of description Face geometric eigenvector can be extracted using Face detection technology, then by the feature of extraction
Point, which connects, is formed corresponding geometrical characteristic.For extracting the lip in face picture, Fig. 3 is in one embodiment
Extract the schematic diagram of the point of Face geometric eigenvector, wherein include the point (65-82) for describing lip-region geometrical characteristic, will retouch
The characteristic point for stating lip, which connects, just constitutes the geometrical characteristic of the lip-region.For inhuman face image, it can use and appoint
What computer picture algorithm extracts the geometrical characteristic in region to be split, does not limit here the algorithm for extracting geometrical characteristic
System.
Step 206, the color characteristic of each pixel in original target image is obtained.
Specifically, what image was made of pixel one by one, each pixel corresponds to a color characteristic.Original mesh
Logo image can be color image and be also possible to gray level image, if image is color image, multiple color space can be used
Component is combined, for example using three components of Lab color space to add the CbCr component combination of YCbCr color space is one
A includes the color characteristic vector (L, a, b, Cb, Cr) of five components, wherein the L in Lab indicates brightness, and a is indicated from carmetta
To the range of green, b indicates the range from yellow to blue；Cb in YCbCr represents the concentration excursion amount of blue, and Cr represents red
The concentration excursion amount of color.If image is gray level image, then color characteristic only uses grayscale information can.Obtain original object
The color characteristic of each pixel in image seeks to obtain the color characteristic vector for indicating pixel.
Step 208, according to the color characteristic of the geometrical characteristic of extraction and each pixel to each in original target image
A pixel is clustered.
In the present embodiment, firstly, determining target area to be split in original target image according to the geometrical characteristic of extraction
Domain and nontarget area have since the target area divided according to geometrical characteristic and nontarget area might not be accurate
When can also have relatively large deviation, so needing color characteristic further combined with each pixel to the target area of division and non-
Target area is corrected adjustment and obtains more accurately dividing.Specifically, according to geometrical characteristic determine in original target image to
Behind the target area and nontarget area of segmentation, target area and nontarget area are filled with different colors to distinguish
Filled image is carried out Fuzzy Processing, extracted in the image after carrying out Fuzzy Processing by target area and nontarget area
For characterizing the parameter of different colours feature, for example, can extract carry out Fuzzy Processing after image in each pixel it is bright
Then angle value clicks through each pixel in original target image according to the color characteristic of each pixel and corresponding brightness value
Row cluster, it should be noted that be not limited to the extraction to brightness value here, can also be that other can be with reaction color feature
Parameter, for example, chroma, saturation degree etc..
Step 210, original target image is split according to the result of cluster.
In the present embodiment, by cluster obtain the result is that each pixel is under the jurisdiction of the probability of target area, obtain
After each pixel is under the jurisdiction of the probability of target area, binary conversion treatment can be carried out to it, that is, will be greater than predetermined probabilities value
Pixel is divided into target area, and the pixel for being less than or equal to predetermined probabilities value is divided into nontarget area.
In the present embodiment, by obtaining original target image to be split, area to be split in original target image is extracted
The geometrical characteristic in domain obtains the color characteristic of each pixel in original target image, according to the geometrical characteristic of extraction and each
The color characteristic of pixel clusters each pixel in original target image.The method of above-mentioned image segmentation, passes through
The color characteristic of geometrical characteristic and each pixel is combined, each pixel in original target image is clustered, in this way
Even if the color characteristic in region to be split is unobvious, since region also can be accurately split by the limitation of geometrical characteristic.Cause
This, by combining and more accurate can be split to region the geometrical characteristic in region to be split and color characteristic.
As shown in figure 4, in one embodiment, according to the color characteristic of the geometrical characteristic of extraction and each pixel to original
The step 208 that each pixel in beginning target image is clustered includes:
Step 402, target area to be split in original target image and non-targeted is determined according to the geometrical characteristic of extraction
Region.
In the present embodiment, after being extracted the geometrical characteristic in region to be split, according to the geometry in the region to be split of extraction
Original target image has been divided into two parts by feature, and one is target area, and one is nontarget area.Wherein, target
Region can be one, be also possible to multiple.For example, by taking face picture as an example, as shown in figure 3, can be according to the face of extraction
The geometrical characteristic of (eyebrow, eye, nose, lip, face's outer profile) regard the face in face as target area, other parts conduct simultaneously
Nontarget area.It can certainly be only using some region as target area, for example, the geometrical characteristic of lip is only extracted, by lip
Portion is as target area, other are as nontarget area.
Step 404, target area is filled with nontarget area with the different color of brightness value to distinguish target area
Domain and nontarget area.
In the present embodiment, after having determined target area and nontarget area to be split in original target image, by mesh
Mark region is filled with nontarget area with the different color of brightness value to distinguish the target area and nontarget area.In order to
Preferably target area and nontarget area are distinguished, the brightness value difference of the Fill Color used is the bigger the better.Preferably,
Target area can be filled with white (brightness value 255), nontarget area is filled with black (brightness value 0)
Or be filled target area with black, nontarget area is filled with white.
Step 406, filled image is subjected to Fuzzy Processing, obtains each pixel in the image after carrying out Fuzzy Processing
The corresponding brightness value of point.
In the present embodiment, since the target area that the geometrical characteristic according to the region to be split of extraction obtains might not
Accurately, there is relatively large deviation sometimes, so need to carry out filled image Fuzzy Processing, it is subsequent in conjunction with color characteristic
Target area and nontarget area to division are corrected adjustment and obtain more accurately dividing.Specifically, carrying out mould to image
Paste processing, it is necessary first to determine blur radius, the size of blur radius depends on extracting the deviation of region geometry feature to be split
Size, deviation should then increase greatly blur radius, and deviation is small, reduces blur radius.The size of blur radius can be rule of thumb
Value is preset, and after the size for having determined blur radius, can be carried out using Gaussian Blur algorithm to filled image
Fuzzy Processing.Fig. 5 A is the original target image in one embodiment, and Fig. 5 B is to obscure after being filled lip-region
Treated schematic diagram, wherein the lip-region of extraction is filled with white, other nontarget areas are carried out with black
Filling.After filled image is carried out Fuzzy Processing, the corresponding brightness value of each pixel is calculated, the calculating of brightness value can
To use existing calculation method, any restrictions are not done to the calculating of brightness value here, for the ease of the calculating of brightness value, preferably
Can choose target area and nontarget area are filled with black and white, it is right that the calculating of such brightness value, which can simplify,
The calculating of gray value.
Step 408, according to the color characteristic of each pixel and corresponding brightness value to each in original target image
Pixel is clustered.
In the present embodiment, every pixel has unique coordinate in the picture, by coordinate same before and after image procossing
The color characteristic and brightness value of the corresponding pixel in position merge, and obtaining one indicates the characteristic vector of multidimensional of the pixel.
For example, if it is a 5 latitude color characteristic vectors (L, a, b, Cb, Cr) that each pixel is corresponding in original target image, then
One has been reformed into after brightness value B (Brightness, brightness) combination corresponding with each pixel after progress Fuzzy Processing
A 6 latitude characteristic vector (L, a, b, Cb, Cr, B).According to the corresponding multidimensional characteristic vector of obtained each pixel to original object
Pixel in image is clustered.Specifically, firstly, according to the corresponding brightness value B of pixel each in treated image
Determine the corresponding initial degree of membership of each pixel；Secondly, according to the corresponding initial degree of membership of each pixel and each pixel
The corresponding multidimensional characteristic vector of point calculates the initial centered value of target area and nontarget area；Finally, according to initial degree of membership
Multidimensional characteristic vector corresponding with initial centered value and each pixel clusters the pixel in original target image
Obtain the probability that each pixel is under the jurisdiction of target area.According to obtained each pixel be under the jurisdiction of the probability of target area into
The segmentation of row image.
In the present embodiment, the division in region is carried out by extracting the geometrical characteristic in region to be split before this, then root again
The target area marked off by geometrical characteristic is corrected according to color characteristic, accordingly even when the color characteristic of the target area of segmentation
It is unobvious, since the limitation of geometrical characteristic also accurately can mark off target area to come, by by the geometry in region to be split
Feature and color characteristic are combined and more accurate can be split to region.
As shown in fig. 6, in one embodiment, according to the color characteristic of each pixel and corresponding brightness value to original
The step 408 that each pixel in target image is clustered includes:
Step 408A, according to the corresponding brightness value of each pixel determine respectively each pixel be under the jurisdiction of target area and
The initial degree of membership of nontarget area.
In the present embodiment, degree of membership belongs to the concept in fuzzy evaluation functions, specifically, if to any in domain U
Element has several A (x) ∈ [0,1] to be corresponding to it, then A is referred to as the fuzzy set on U, and A (x) becomes x to the degree of membership of A.When
When x changes in U, A (x) is exactly a function, the referred to as membership function of A.Degree of membership A (x) indicates that x belongs to A closer to 1
Degree it is higher, degree of membership A (x) closer to 0, indicate x belong to A degree it is lower, i.e., be subordinate to letter with value interval [0,1]
Number A (x) indicates the degree height for belonging to A.Before being clustered, it is necessary first to determine that each pixel is under the jurisdiction of target area
Initial degree of membership and each pixel be under the jurisdiction of the initial degree of membership of nontarget area.The determination of initial degree of membership is to make
Finally belong to the probability of target area for each pixel of calculation of initial value.It is under the jurisdiction of the probability (degree of membership) of target area and is subordinate to
Belonging to the sum of probability (degree of membership) of nontarget area is 1.Assuming that the degree of membership that pixel is under the jurisdiction of target area is A (xi),
The degree of membership for being under the jurisdiction of nontarget area is B (xi), wherein A (xi)+B(xi)=1, xiIndicate the data that ith measurement arrives.Just
Beginning degree of membership is determined according to brightness value.Specifically, in one embodiment, the pixel that will be greater than predetermined luminance value is under the jurisdiction of mesh
The initial degree of membership in mark region is set as 1, and the pixel no more than predetermined luminance value is under the jurisdiction of to the initial degree of membership of target area
It is set as 0, then the initial degree of membership that the pixel for being greater than predetermined luminance value is under the jurisdiction of nontarget area is 0, it is bright no more than presetting
The initial degree of membership that the pixel of angle value is under the jurisdiction of target area is 1.Specifically, with reference to Fig. 5 B, xiBelong to the first of target area
The calculation method of beginning degree of membership are as follows: if xiThe brightness value on position in figure 5B is greater than 128, then initial degree of membership is 1, small
It is 0 in being equal to 128 initial degrees of membership.xiBelong to the initial degree of membership of nontarget area just and belongs to the initial of target area
Degree of membership is consistent, i.e., brightness value is greater than 128, and initial degree of membership is 0, and being less than or equal to 128 initial degrees of membership is 1.
Step 408B determines target area and non-mesh according to the color characteristic of each pixel and corresponding brightness value respectively
Mark the initial centered value in region.
In the present embodiment, before being clustered, need first to determine the cluster centre of target area and nontarget area i.e.
Initial centered value.Initial centered value is calculated, average weighted calculation method can be used.Specifically, formula can be passed throughIt is calculated, wherein cjIndicate the cluster centre of class j；xiIndicate what ith measurement arrived
Multidimensional data；uijIt is xiBelong to the degree of membership of classification j；M is a control algolithm parameter flexible, generally takes m=2；N indicates total
Data amount check.Wherein, characteristic xiIt is in the present embodiment exactly the corresponding multidimensional characteristic vector of each pixel (by color
Feature and brightness value composition)；uijThe each pixel exactly obtained belongs to the initial degree of membership of classification j.In one embodiment,
With reference to Fig. 5 B, cluster centre, that is, initial centered value calculation method of target area are as follows: all brightness values are equal to 255 in figure 5B
Pixel calculate the draw value i.e. initial centered value c of its multidimensional characteristic vector in original target image (Fig. 5 A)1.Equally
, pixel of all brightness values equal to 0 calculates its multidimensional characteristic in original target image in nontarget area in figure 5B
The draw value c of vector2。c1And c2It is the cluster centre i.e. initial centered value of target area and nontarget area respectively.
Step 408C, according to the color characteristic of initial degree of membership and initial centered value and each pixel and corresponding bright
Angle value is clustered to obtain the probability that each pixel is under the jurisdiction of target area to the pixel in original target image.
In the present embodiment, it is determined that each pixel be under the jurisdiction of target area and nontarget area initial degree of membership and
It, will be each using initial degree of membership and initial centered value as initial parameter after the cluster centre of target area and nontarget area
The color characteristic of pixel and corresponding brightness value group are combined into a multidimensional characteristic vector as input variable, then using cluster
Algorithm carries out cluster calculation.In one embodiment, it can be iterated using FCM clustering algorithm and each pixel is calculated
It is under the jurisdiction of the final degree of membership of target area, the segmentation in region is carried out according to the final degree of membership (probability).Specifically, passing through public affairs
FormulaWith
It is updated iteration and makes objective functionReach minimum.Wherein, m is greater than 1
Real number generally takes m=2, uijIt is xiBelong to the degree of membership of classification j, cjIndicate the cluster centre of class j, xiIndicate that ith measurement arrives
Multidimensional data, | | * | | indicate the pixel degree of any measurement data and cluster centre, k indicates the value range of C, and C indicates class
Other quantity.Firstly, passing through the initial degree of membership that will be determinedIt substitutes into above-mentioned formula (1) and calculates cluster centre cj, then substitute into again
The degree of membership of above-mentioned formula (2) calculating new roundIn this way continuous iteration untilWhen, iteration stopping.Wherein 0 < ε < 1 is iteration ends parameter, and n is iteration wheel number, at this
J during amConverge to a minimum.It is under the jurisdiction of the final of target area by iterating to calculate available each pixel
Degree of membership carries out the segmentation of image according to the final degree of membership (probability).
As shown in fig. 7, in one embodiment, filled image being carried out Fuzzy Processing, calculates and carries out Fuzzy Processing
The step 406 of the brightness value of each pixel includes: in image afterwards
Step 406A obtains preset blur radius, carries out Fuzzy Processing to filled image according to blur radius.
In the present embodiment, it to the target area in original target image and after nontarget area is filled, needs pair
Filled image carries out Fuzzy Processing, this is because not necessarily by the target area of geometrical characteristic division and nontarget area
Accurately, in fact it could happen that deviation, so need to filled image carry out Fuzzy Processing, it is subsequent in conjunction with color characteristic to division
Target area and nontarget area be corrected adjustment and obtain more accurately dividing.Before carrying out Fuzzy Processing, it is necessary first to really
Determine blur radius, the size of blur radius is can to carry out pre-set, and the size of pre-set blur radius depends on
Deviation size.Deviation size is an empirical value, and the deviation for extracting the algorithm of geometrical characteristic can be evaluated by repeatedly test
Then range can determine that blur radius, blur radius generally take the value greater than maximum deviation according to deviation range.Such as deviation
Range is 0-10, then blur radius may be greater than 10 value, in this way convenient for it is subsequent can be more accurate to target area
It is divided.
Step 406B calculates the corresponding brightness value of each pixel in the image after carrying out Fuzzy Processing.
In the present embodiment, in order to distinguish target area and nontarget area, respectively by target area and non-mesh
Mark region is filled with the different color of brightness value, is then carried out Fuzzy Processing to filled image, is thus obtained
One new target image calculates the corresponding brightness value of each pixel in the new target image.For the ease of brightness value
It calculates, white filling generally is carried out to target area, nontarget area carries out filled black, not only can preferably distinguish in this way
Target area and nontarget area can also simplify the calculating of brightness value, only need when because filling by black and white and calculate brightness value
Carrying out gray count can be obtained by the corresponding brightness value of each pixel.If be filled with colour, calculate bright
The corresponding brightness value of pixel can just be obtained by needing the value to three kinds of colors to be weighted and averaged calculating when angle value.
In one embodiment, region is carried out according to the probability that each pixel that cluster obtains is under the jurisdiction of target area
Segmentation, wherein the pixel that will be greater than predetermined probabilities value is divided into target area, will be less than or equal to the pixel of predetermined probabilities value
Point is divided into nontarget area.
In the present embodiment, by cluster iteration at the end of obtain each pixel and belong to the degree of membership of target area i.e. most
Each pixel belongs to the probability of target area eventually.Obtained probability is subjected to a binary conversion treatment, i.e., according to probabilistic determination
Whether each pixel belongs to target area.Specifically, a probability value can be set, for example probability value is set as 0.6, by probability
Pixel greater than 0.6 is divided into target area, and the pixel by probability value less than or equal to 0.6 is divided into nontarget area.
As shown in figure 8, in one embodiment it is proposed that a kind of device of image segmentation, the device include:
Image collection module 802, for obtaining original target image to be split.
Extraction module 804, for extracting the geometrical characteristic in region to be split in original target image.
Color characteristic obtains module 806, for obtaining the color characteristic of each pixel in original target image.
First cluster module 808, for according to the geometrical characteristic of extraction and the color characteristic of each pixel to original mesh
Each pixel in logo image is clustered.
Divide module 810, for being split original target image according to the result of cluster.
As shown in figure 9, in one embodiment, the first cluster module 808 includes:
Determining module 902 determines target area to be split in original target image for the geometrical characteristic according to extraction
The nontarget area and.
Module 904 is filled, for being filled target area with the different color of brightness value with area with nontarget area
Partial objectives for region and nontarget area.
Fuzzy Processing module 906 calculates the figure after carrying out Fuzzy Processing for filled image to be carried out Fuzzy Processing
The corresponding brightness value of each pixel as in.
Second cluster module 908, for the color characteristic and corresponding brightness value according to each pixel to original object
Each pixel in image is clustered.
As shown in Figure 10, in one embodiment, the second cluster module 908 includes:
Initial degree of membership determining module 908A, for determining each pixel respectively according to the corresponding brightness value of each pixel
Point is under the jurisdiction of the initial degree of membership of target area and nontarget area.
Initial centered value determining module 908B, for according to the color characteristic of each pixel and corresponding brightness value difference
Determine the initial centered value of target area and nontarget area.
Third cluster module 908C, for special according to initial degree of membership and initial centered value and the color of each pixel
Corresponding brightness value of seeking peace is clustered to obtain each pixel to the pixel in original target image is under the jurisdiction of target area
Probability.
As shown in figure 11, Fuzzy Processing module 906 includes:
Blur radius obtains module 906A, for obtaining preset blur radius, according to blur radius to filled figure
As carrying out Fuzzy Processing.
Brightness value computing module 906B, for calculating the corresponding brightness of each pixel in the image after carrying out Fuzzy Processing
Value.
In one embodiment, segmentation module is also used to be under the jurisdiction of target area according to each pixel that cluster obtains
The segmentation in probability progress region, wherein the pixel that will be greater than predetermined probabilities value is divided into target area, will be less than or equal to pre-
If the pixel of probability value is divided into nontarget area.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, which can be stored in a computer-readable storage and be situated between
In matter, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, storage medium above-mentioned can be
The non-volatile memory mediums such as magnetic disk, CD, read-only memory (Read-Only Memory, ROM) or random storage note
Recall body (Random Access Memory, RAM) etc..
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
Limitations on the scope of the patent of the present invention therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention
Protect range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (10)
1. a kind of method of image segmentation, which comprises
Obtain original target image to be split；
Extract the geometrical characteristic in region to be split in the original target image；
Obtain the color characteristic of each pixel in the original target image；
According to the color characteristic of the geometrical characteristic of extraction and each pixel to each picture in original target image
Vegetarian refreshments is clustered, and determines target area to be split in the original target image and non-according to the geometrical characteristic of extraction
The target area and nontarget area are filled with different colors to distinguish the target area and non-by target area
Filled image is carried out Fuzzy Processing by target area, is extracted and is used to characterize difference in the image after carrying out Fuzzy Processing
The extraction characteristic parameter of color characteristic, according to the color characteristic of each pixel and corresponding extraction characteristic parameter to original object
Each pixel in image is clustered；
The original target image is split according to the result of cluster.
2. the method according to claim 1, wherein the geometrical characteristic according to extraction and described each
The step of color characteristic of pixel clusters each pixel in original target image include:
The target area is filled with the nontarget area with the different color of brightness value to distinguish the target area
Domain and the nontarget area；
Filled image is subjected to Fuzzy Processing, calculates the corresponding brightness of each pixel in the image after carrying out Fuzzy Processing
Value；
According to the color characteristic of each pixel and the corresponding brightness value to each in the original target image
Pixel is clustered.
3. according to the method described in claim 2, it is characterized in that, the color characteristic according to each pixel and right
The step of brightness value answered clusters each pixel in the original target image include:
Determine that each pixel is under the jurisdiction of the target area and described respectively according to the corresponding brightness value of each pixel
The initial degree of membership of nontarget area；
Target area and non-target area are determined respectively according to the color characteristic of each pixel and the corresponding brightness value
The initial centered value in domain；
According to the color characteristic and corresponding institute of the initial degree of membership and the initial centered value and each pixel
It states brightness value and the pixel in original target image is clustered to obtain each pixel and be under the jurisdiction of the general of the target area
Rate.
4. according to the method described in claim 2, it is characterized in that, described carry out Fuzzy Processing, acquisition for filled image
The step of brightness value of each pixel, includes: in image after carrying out Fuzzy Processing
Preset blur radius is obtained, Fuzzy Processing is carried out to filled image according to the blur radius；
Calculate the corresponding brightness value of each pixel in the image after carrying out the Fuzzy Processing.
5. the method according to claim 1, wherein the result according to cluster is by the original target image
The step of being split include:
The segmentation in region is carried out according to the probability that each pixel that cluster obtains is under the jurisdiction of the target area, wherein will be big
It is divided into target area in the pixel of predetermined probabilities value, the pixel for being less than or equal to predetermined probabilities value is divided into non-targeted
Region.
6. a kind of device of image segmentation, which is characterized in that described device includes:
Image collection module, for obtaining original target image to be split；
Extraction module, for extracting the geometrical characteristic in region to be split in the original target image；
Color characteristic obtains module, for obtaining the color characteristic of each pixel in the original target image；
First cluster module, for according to the geometrical characteristic of extraction and the color characteristic of each pixel to original mesh
Each pixel in logo image is clustered, and is determined in the original target image according to the geometrical characteristic of extraction wait divide
The target area and nontarget area are filled with different colors to distinguish by the target area and nontarget area cut
Filled image is carried out Fuzzy Processing, extracts the image after carrying out Fuzzy Processing by the target area and nontarget area
In for characterizing the extraction characteristic parameter of different colours feature, according to the color characteristic of each pixel and corresponding extractions spy
Sign parameter clusters each pixel in original target image；
Divide module, for being split the original target image according to the result of cluster.
7. device according to claim 6, which is characterized in that first cluster module includes:
Determining module, for determining target area to be split in the original target image according to the geometrical characteristic of extraction
The nontarget area and；
Module is filled, for being filled the target area with the different color of brightness value with area with the nontarget area
Divide the target area and the nontarget area；
Fuzzy Processing module calculates every in the image after carrying out Fuzzy Processing for filled image to be carried out Fuzzy Processing
The corresponding brightness value of a pixel；
Second cluster module, for the color characteristic and the corresponding brightness value according to each pixel to described original
Each pixel in target image is clustered.
8. device according to claim 7, which is characterized in that second cluster module includes:
Initial degree of membership determining module, for determining that each pixel is subordinate to respectively according to the corresponding brightness value of each pixel
Belong to the initial degree of membership of the target area and the nontarget area；
Initial centered value determining module, for being distinguished according to the color characteristic and the corresponding brightness value of each pixel
Determine the initial centered value of target area and nontarget area；
Third cluster module, for the face according to the initial degree of membership and the initial centered value and each pixel
Color characteristic and the corresponding brightness value are clustered to obtain each pixel to the pixel in original target image to be under the jurisdiction of
The probability of the target area.
9. device according to claim 7, which is characterized in that the Fuzzy Processing module includes:
Blur radius obtain module, for obtaining preset blur radius, according to the blur radius to filled image into
Row Fuzzy Processing；
Brightness value computing module, for calculating the corresponding brightness value of each pixel in the image after carrying out the Fuzzy Processing.
10. device according to claim 6, which is characterized in that the segmentation module is also used to be obtained according to cluster every
The probability that a pixel is under the jurisdiction of the target area carries out the segmentation in region, wherein will be greater than the pixel of predetermined probabilities value
It is divided into target area, the pixel for being less than or equal to predetermined probabilities value is divided into nontarget area.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610702051.8A CN106340023B (en) | 2016-08-22 | 2016-08-22 | The method and apparatus of image segmentation |
| PCT/CN2017/098417 WO2018036462A1 (en) | 2016-08-22 | 2017-08-22 | Image segmentation method, computer apparatus, and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610702051.8A CN106340023B (en) | 2016-08-22 | 2016-08-22 | The method and apparatus of image segmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106340023A CN106340023A (en) | 2017-01-18 |
| CN106340023B true CN106340023B (en) | 2019-03-05 |
Family
ID=57824596
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610702051.8A Active CN106340023B (en) | 2016-08-22 | 2016-08-22 | The method and apparatus of image segmentation |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN106340023B (en) |
| WO (1) | WO2018036462A1 (en) |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106340023B (en) * | 2016-08-22 | 2019-03-05 | 腾讯科技（深圳）有限公司 | The method and apparatus of image segmentation |
| CN109147011B (en) * | 2018-08-27 | 2023-11-14 | 平安科技（深圳）有限公司 | License plate image generation method, license plate image generation device, computer equipment and storage medium |
| CN110290426B (en) * | 2019-06-24 | 2022-04-19 | 腾讯科技（深圳）有限公司 | Method, device and equipment for displaying resources and storage medium |
| CN112308938A (en) * | 2019-07-30 | 2021-02-02 | 西安诺瓦星云科技股份有限公司 | Image processing method and image processing apparatus |
| CN110852938B (en) * | 2019-10-28 | 2024-03-19 | 腾讯科技（深圳）有限公司 | Display picture generation method, device and storage medium |
| CN110910400A (en) * | 2019-10-29 | 2020-03-24 | 北京三快在线科技有限公司 | Image processing method, image processing device, storage medium and electronic equipment |
| CN111026641B (en) * | 2019-11-14 | 2023-06-20 | 北京云聚智慧科技有限公司 | Picture comparison method and electronic equipment |
| CN111079637B (en) * | 2019-12-12 | 2023-09-08 | 武汉轻工大学 | Method, device, equipment and storage medium for segmenting rape flowers in field image |
| CN112991357B (en) * | 2019-12-18 | 2023-04-18 | 中国船舶集团有限公司第七一一研究所 | Image segmentation method, system, computer device, readable storage medium and ship |
| CN111325691B (en) * | 2020-02-20 | 2023-11-10 | Oppo广东移动通信有限公司 | Image correction method, apparatus, electronic device, and computer-readable storage medium |
| CN111667553A (en) * | 2020-06-08 | 2020-09-15 | 北京有竹居网络技术有限公司 | Head-pixelized face color filling method and device and electronic equipment |
| WO2021253373A1 (en) * | 2020-06-19 | 2021-12-23 | Alibaba Group Holding Limited | Probabilistic geometric partitioning in video coding |
| CN113012188A (en) * | 2021-03-23 | 2021-06-22 | 影石创新科技股份有限公司 | Image fusion method and device, computer equipment and storage medium |
| CN113344961B (en) * | 2021-06-01 | 2023-09-26 | 中国平安人寿保险股份有限公司 | Image background segmentation method, device, computing equipment and storage medium |
| WO2023108444A1 (en) * | 2021-12-14 | 2023-06-22 | 深圳传音控股股份有限公司 | Image processing method, intelligent terminal, and storage medium |
| CN114495236B (en) * | 2022-02-11 | 2023-02-28 | 北京百度网讯科技有限公司 | Image segmentation method, apparatus, device, medium, and program product |
| CN115439846B (en) * | 2022-08-09 | 2023-04-25 | 北京邮电大学 | Image segmentation method and device, electronic equipment and medium |
| CN116152562B (en) * | 2023-02-23 | 2023-08-01 | 北京朗视仪器股份有限公司 | Method and system for rapid color classification of color structured light image |
| CN116563312B (en) * | 2023-07-11 | 2023-09-12 | 山东古天电子科技有限公司 | Method for dividing display image of double-screen machine |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020975A (en) * | 2012-12-29 | 2013-04-03 | 北方工业大学 | Wharf and ship segmentation method combining multi-source remote sensing image characteristics |
| CN103034872A (en) * | 2012-12-26 | 2013-04-10 | 四川农业大学 | Farmland pest recognition method based on colors and fuzzy clustering algorithm |
| CN103903257A (en) * | 2014-02-27 | 2014-07-02 | 西安电子科技大学 | Image segmentation method based on geometric block spacing symbiotic characteristics and semantic information |
| CN105869175A (en) * | 2016-04-21 | 2016-08-17 | 北京邮电大学 | Image segmentation method and system |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101925916B (en) * | 2007-11-21 | 2013-06-19 | 高通股份有限公司 | Method and system for controlling electronic device based on media preferences |
| CN202110564U (en) * | 2011-06-24 | 2012-01-11 | 华南理工大学 | Intelligent household voice control system combined with video channel |
| CN104134080B (en) * | 2014-08-01 | 2018-09-11 | 重庆大学 | A kind of road foundation collapses automatic testing method and system with slope failure |
| JP6529246B2 (en) * | 2014-11-28 | 2019-06-12 | キヤノン株式会社 | Feature extraction method, feature extraction device, and program |
| CN105469356B (en) * | 2015-11-23 | 2018-12-18 | 小米科技有限责任公司 | Face image processing process and device |
| CN106340023B (en) * | 2016-08-22 | 2019-03-05 | 腾讯科技（深圳）有限公司 | The method and apparatus of image segmentation |
-
2016
- 2016-08-22 CN CN201610702051.8A patent/CN106340023B/en active Active
-
2017
- 2017-08-22 WO PCT/CN2017/098417 patent/WO2018036462A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103034872A (en) * | 2012-12-26 | 2013-04-10 | 四川农业大学 | Farmland pest recognition method based on colors and fuzzy clustering algorithm |
| CN103020975A (en) * | 2012-12-29 | 2013-04-03 | 北方工业大学 | Wharf and ship segmentation method combining multi-source remote sensing image characteristics |
| CN103903257A (en) * | 2014-02-27 | 2014-07-02 | 西安电子科技大学 | Image segmentation method based on geometric block spacing symbiotic characteristics and semantic information |
| CN105869175A (en) * | 2016-04-21 | 2016-08-17 | 北京邮电大学 | Image segmentation method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2018036462A1 (en) | 2018-03-01 |
| CN106340023A (en) | 2017-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106340023B (en) | The method and apparatus of image segmentation | |
| Oh et al. | Approaching the computational color constancy as a classification problem through deep learning | |
| Ban et al. | Face detection based on skin color likelihood | |
| WO2020082577A1 (en) | Seal anti-counterfeiting verification method, device, and computer readable storage medium | |
| Shuhua et al. | The application of improved HSV color space model in image processing | |
| US9483835B2 (en) | Depth value restoration method and system | |
| WO2021169161A1 (en) | Image recognition method, recognition model training method and apparatuses related thereto, and device | |
| WO2020199475A1 (en) | Facial recognition method and apparatus, computer device and storage medium | |
| CN103035013B (en) | A kind of precise motion shadow detection method based on multi-feature fusion | |
| Wang et al. | Background-driven salient object detection | |
| CN107392968B (en) | The image significance detection method of Fusion of Color comparison diagram and Color-spatial distribution figure | |
| CN109952594A (en) | Image processing method, device, terminal and storage medium | |
| Renugambal et al. | Application of image processing techniques in plant disease recognition | |
| CN105335719A (en) | Living body detection method and device | |
| Shih et al. | Automatic reference color selection for adaptive mathematical morphology and application in image segmentation | |
| CN110059722A (en) | Checking method, device, equipment and the readable storage medium storing program for executing of seal image | |
| CN110298829A (en) | A kind of lingual diagnosis method, apparatus, system, computer equipment and storage medium | |
| Huo et al. | Semisupervised learning based on a novel iterative optimization model for saliency detection | |
| CN111339932B (en) | Palm print image preprocessing method and system | |
| Paul et al. | Rotation invariant multiview face detection using skin color regressive model and support vector regression | |
| Duan et al. | Visual saliency detection using information contents weighting | |
| Raval et al. | Color image segmentation using FCM clustering technique in RGB, L* a* b, HSV, YIQ color spaces | |
| CN111160194A (en) | Static gesture image recognition method based on multi-feature fusion | |
| CN106557771A (en) | Skin disease color of image feature extracting method based on Naive Bayes Classifier | |
| Nguyen et al. | Using contextual information to classify nuclei in histology images |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210924 Address after: 518057 Tencent Building, No. 1 High-tech Zone, Nanshan District, Shenzhen City, Guangdong Province, 35 floors Patentee after: TENCENT TECHNOLOGY (SHENZHEN) Co.,Ltd. Patentee after: TENCENT CLOUD COMPUTING (BEIJING) Co.,Ltd. Address before: 2, 518000, East 403 room, SEG science and Technology Park, Zhenxing Road, Shenzhen, Guangdong, Futian District Patentee before: TENCENT TECHNOLOGY (SHENZHEN) Co.,Ltd. |
|
| TR01 | Transfer of patent right |