CN108596920A - A kind of Target Segmentation method and device based on coloured image - Google Patents
A kind of Target Segmentation method and device based on coloured image Download PDFInfo
- Publication number
- CN108596920A CN108596920A CN201810408454.0A CN201810408454A CN108596920A CN 108596920 A CN108596920 A CN 108596920A CN 201810408454 A CN201810408454 A CN 201810408454A CN 108596920 A CN108596920 A CN 108596920A
- Authority
- CN
- China
- Prior art keywords
- image
- target
- coarse segmentation
- segmentation
- mean
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 108
- 238000000034 method Methods 0.000 title claims abstract description 78
- 238000001914 filtration Methods 0.000 claims abstract description 44
- 230000000877 morphologic effect Effects 0.000 claims abstract description 9
- 230000007797 corrosion Effects 0.000 claims abstract description 8
- 238000005260 corrosion Methods 0.000 claims abstract description 8
- 239000003550 marker Substances 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 11
- 238000005530 etching Methods 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000012804 iterative process Methods 0.000 claims description 5
- 230000006641 stabilisation Effects 0.000 claims description 4
- 238000011105 stabilization Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 abstract description 5
- 238000000605 extraction Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 239000000284 extract Substances 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 238000003064 k means clustering Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011049 filling Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G06T5/70—
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to technical field of data processing, provide a kind of Target Segmentation method and device based on coloured image, and this method includes:Smothing filtering is carried out to input picture using mean-shift method;Cluster row coarse segmentation is carried out to the grey level histogram of image after smothing filtering using K mean cluster method, obtains coarse segmentation figure, carrying out morphologic corrosion expansive working to the coarse segmentation figure builds target label figure;Based on the target label figure, the profile of target is extracted from image after the smothing filtering using dividing ridge method.The device includes:Module is cut in image filtering module, coarse segmentation module and subdivision.The present invention can accurately be partitioned into the targets such as the aircraft under the more complex background such as the sandstone in field, meadow, woods, and realize the accurate extraction of objective contour.
Description
Technical field
The present invention relates to technical field of data processing more particularly to a kind of Target Segmentation methods and dress based on coloured image
It sets.
Background technology
In aviation field, it is an important research direction that view-based access control model, which is identified aircraft, positions and tracks,.Wherein
The key of track and localization is to need accurately to be partitioned into target in the picture.Background residing for Aircraft Targets is sandstone, meadow, tree
Lin Shi, traditional image partition method are difficult accurately to split it.
It would therefore be highly desirable to provide a kind of method that can split target from the coloured image with complex background.
Invention content
The technical problem to be solved in the present invention is, is difficult the back of the body by target from complexity for traditional image partition method
The defect accurately split in scape provides a kind of Target Segmentation method and device based on coloured image, improves complexity
The accuracy of target especially Aircraft Targets segmentation under scene.
The Target Segmentation method based on coloured image that in order to solve the above technical problem, the present invention provides a kind of, including:
Smothing filtering is carried out to input picture using mean-shift method;
Cluster row coarse segmentation is carried out to the grey level histogram of image after smothing filtering using K mean cluster method, is obtained thick
Segmentation figure carries out morphologic etching operation to the coarse segmentation figure and expansive working builds target label figure;
Based on the target label figure, the wheel of target is extracted from image after the smothing filtering using dividing ridge method
It is wide.
Optionally, described that smothing filtering is carried out to input picture using mean-shift method, including:
Input picture is transformed into CIELAB color spaces;
Each pixel on converted images is denoted as five dimension point Z=(Xs,Xy), wherein XsIndicate the space of pixel
Variable (x, y), XrThe color variables (L, a, b) for indicating pixel calculate the mean shift variable of five dimension point Z by kernel function
Value, and constantly iteration, finally will be in stable color variables (L, a, b) assignment to initial point until its stabilization.
Optionally, consider that pixel all in the kernel function spatial window around the pixel influences in iterative process, and
The threshold value that mean shift variable is arranged is 3, and the loop termination when mean shift variate-value is less than 3 determines and is in stable state.
Optionally, described that cluster row is carried out slightly to the grey level histogram of image after smothing filtering using K mean cluster method
Segmentation, including:
Divide equally the tonal range of grey level histogram using preset spacing H, and finds out ash in image in each tonal range
The maximum gray value of the frequency of angle value within this range is clustered as the initial cluster center of this tonal range, when
When the center variation of cluster is less than 0.5 pixel, then it is assumed that clustering algorithm is restrained, and iteration is terminated;
Coarse segmentation is carried out to image using cluster result, then carries out morphology closed operation, removal small holes obtain described thick
Segmentation figure.
Optionally, described that morphologic etching operation and expansive working structure target label are carried out to the coarse segmentation figure
Figure, including:
The inner marker figure of target is obtained after carrying out corrosion treatment to the coarse segmentation figure;
The external label figure of target is obtained after carrying out expansion process to the coarse segmentation figure;
By the inner marker figure and external label figure in conjunction with obtaining the target label figure.
The present invention also provides a kind of Target Segmentation device based on coloured image, including:
Image filtering module, for carrying out smothing filtering to input picture using mean-shift method;
Coarse segmentation module, for the ash using K mean cluster method to image after described image filter module smothing filtering
Degree histogram carries out cluster row coarse segmentation, obtains coarse segmentation figure, and morphologic etching operation and swollen is carried out to the coarse segmentation figure
Swollen operation builds target label figure;
Module is cut in subdivision, for being based on the target label figure, using dividing ridge method from image after the smothing filtering
In extract the profile of target.
Optionally, described image filter module includes:
Image conversion unit, for input picture to be transformed into CIELAB color spaces;
Image filtering unit, for each pixel on converted images to be denoted as five dimension point Z=(Xs,Xy), wherein Xs
Indicate the space variable (x, y) of pixel, XrThe color variables (L, a, b) for indicating pixel calculate five dimension point Z by kernel function
Mean shift variate-value, and constantly iteration until its stablize, finally by stable color variables (L, a, b) assignment to initially
Point on.
Optionally, described image filter unit considers institute in the kernel function spatial window around the pixel in an iterative process
Some pixels influence, and the threshold value that mean shift variable is arranged is 3, the loop termination when mean shift variate-value is less than 3, really
Surely it is in stable state.
Optionally, the coarse segmentation module carries out cluster row coarse segmentation in the following manner:
Divide equally the tonal range of grey level histogram using preset spacing H, and finds out ash in image in each tonal range
The maximum gray value of the frequency of angle value within this range is clustered as the initial cluster center of this tonal range, when
When the center variation of cluster is less than 0.5 pixel, then it is assumed that clustering algorithm is restrained, and iteration is terminated;
Coarse segmentation is carried out to image using cluster result, then carries out morphology closed operation, removal small holes obtain described thick
Segmentation figure.
Optionally, the coarse segmentation module builds target label figure in the following manner:
The inner marker figure of target is obtained after carrying out corrosion treatment to the coarse segmentation figure;
The external label figure of target is obtained after carrying out expansion process to the coarse segmentation figure;
By the inner marker figure and external label figure in conjunction with obtaining the target label figure.
Implement the Target Segmentation method and device provided in an embodiment of the present invention based on coloured image, at least have has as follows
Beneficial effect:The present invention is not finely divided using dividing ridge method not instead of directly and is cut, first with K- after smothing filtering
Means clustering methods carry out coarse segmentation to image, the corrosion expansive working of combining form, build the inner marker figure of target with
External label figure can position the position of area-of-interest in this way by inner marker, can be rejected by external label
The interference of noise.Connecting inner label figure and external label figure generate target label figure, as the prior information of watershed algorithm,
To realize the accurate segmentation of target.The method of the present invention can accurately be partitioned into sandstone, meadow, the woods etc. in field
The targets such as the aircraft under more complex background, and realize the accurate extraction of objective contour.
Description of the drawings
Fig. 1 is the principle flow chart that the embodiment of the present invention one provides the Target Segmentation method based on coloured image;
Fig. 2 a to 2e are respectively the change schematic diagram of the example images one of Target Segmentation method processing according to the present invention;Its
Middle Fig. 2 a are the original color image of input;Fig. 2 b are the image carried out after the disposal of gentle filter;Fig. 2 c are the figure after coarse segmentation
Picture;Fig. 2 d are target label figure;Fig. 2 e are the objective contour figure that segmentation obtains;
Fig. 3 is image segmentation algorithm implementation flow chart provided by the present invention;
Fig. 4 is the schematic diagram that the embodiment of the present invention five provides the Target Segmentation device based on coloured image;
Fig. 5 is the schematic diagram of equipment where the Target Segmentation device based on coloured image that the embodiment of the present invention is provided;
Fig. 6 is the segmentation effect figure according to the present invention;
In figure:401:Image filtering module;402:Coarse segmentation module;403:Module is cut in subdivision.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The every other embodiment that member is obtained without making creative work, shall fall within the protection scope of the present invention.
Embodiment one
As shown in Figure 1, the Target Segmentation method provided in an embodiment of the present invention based on coloured image, may include following step
Suddenly:
Step S101:Smothing filtering is carried out to input picture using mean shift (Mean Shift) method;In the step
Image is smoothly pre-processed, the sharp features in image can be made smooth-out, is convenient for subsequent cutting operation.
Step S102:The grey level histogram of image after smothing filtering is gathered using K mean values (K-means) clustering method
Class row coarse segmentation, obtains coarse segmentation figure, carries out morphologic etching operation to the coarse segmentation figure and expansive working builds target
Label figure.
Step S103:Based on the target label figure, using dividing ridge method from being extracted in image after the smothing filtering
Go out the profile of target.
Preferably, morphologic etching operation is carried out to the coarse segmentation figure in step S102 and expansive working builds mesh
The process of mark label figure, is realized especially by following steps:
1) corrosion treatment is carried out to the coarse segmentation figure, removes the noise of zonule and the interference of wisp, position target
Position, obtain the inner marker figure of target;
2) expansion process is carried out to the coarse segmentation figure, connection is realized by expanding in the region where target, limits mesh
Target area obtains the external label figure of target;
3) by the inner marker figure and external label figure in conjunction with obtaining the target label figure.Inside is marked in the step
Note figure and external label figure carry out image superposition and can be obtained the target label figure.It should be noted that inner marker figure with
External label figure gray value is different, and the marked region in inner marker figure is generally converted to high gray value (i.e. white), will be outer
Marked region in portion's label figure is converted to low gray value (i.e. black), and wherein conversion operation is realized by inverse.
Mean Shift algorithms in the prior art are generally divided into two steps, filtering and segmentation, due to noise and image gradient
Scrambling, if directly using fractional spins after the filtering, it is easy to cause over-segmentation, that is, generate a large amount of
Cut zone.Sometimes, over-segmentation can be arrived seriously so that the obtained segmentation result of algorithm is meaningless.Therefore, of the invention
For such case, after the filtering, coarse segmentation, the corruption of combining form are carried out to image first with K-means clustering methods
Expansive working is lost, the inner marker figure and external label figure of target are built.Inner marker is and interested region (object) phase
The label of contact, external label are labels associated with background.In this way by inner marker, area-of-interest can be positioned
Position can reject the interference of some noises by external label.Connecting inner label figure and external label figure generate target mark
Note figure, as the prior information of watershed algorithm, to realize the accurate segmentation of target.
If Fig. 2 a to 2e are respectively the change schematic diagram for the example images one that Target Segmentation method according to the present invention is handled.
Wherein Fig. 2 a are the original color image of input;Fig. 2 b are that step S101 carries out the image after the disposal of gentle filter;Fig. 2 c are step
Image after rapid S102 coarse segmentations;Fig. 2 d are the target label figure that step S102 is generated;Fig. 2 e are what step S103 was divided
Objective contour figure.As shown, can accurately be partitioned into sandstone in field, meadow, the woods using the method for the present invention
Etc. targets such as aircrafts under more complex background, and realize the accurate extraction of objective contour.
In the Target Segmentation method provided in an embodiment of the present invention based on coloured image, the target can be aircraft mesh
Mark is either other to need the target detached from complex background such as aircraft or vehicle.
Embodiment two
On the basis of embodiment one provides the Target Segmentation method based on coloured image, used described in step S101
Mean-shift method carries out input picture the process of smothing filtering, can specifically be achieved by the steps of:
1) coloured image of input is transformed into CIELAB color spaces, in order to which subsequent Mean Shift are smooth.
2) regard each pixel on image after format conversion as five dimension point Z=(Xs,Xy), wherein XsIndicate picture
The space variable (x, y) of vegetarian refreshments, XrThe color variables (L, a, b) for indicating pixel are calculated by using the kernel function of homogeneous nucleus
Mean shift (Mean Shift) variate-value of five dimension point Z, and constantly iteration finally becomes stable color until its stabilization
It measures in (L, a, b) assignment to initial point.For example, the frequency domain threshold point that Mean Shift variables are arranged in iteration is 3, when calculating
Iteration ends when Mean Shift variate-values are less than 3 determine and are in stable state.Consider when the iteration around this point
All points influence in kernel function spatial window.After handling in this way, all adjacent, color Euclidean distances are less than definition on picture
Frequency domain threshold point will link together, and just eliminate some details of image in this way, subsequent step is facilitated to execute segmentation again
Operation.Subsequent singulation process is actually region merging technique, merges similar neighboring region and rejects zonule.
Embodiment three
On the basis of embodiment one or two provides the Target Segmentation method based on coloured image, described in step S102
The process for carrying out cluster row coarse segmentation to the grey level histogram of image after smothing filtering using K mean cluster method, specifically can be with
It is achieved by the steps of:
1) spacing H is preset, respectively the tonal range of grey level histogram, and finds out gray value in image in each tonal range
The maximum gray value of frequency within this range is clustered as the initial cluster center of this tonal range, works as cluster
Center variation be less than 0.5 pixel when, then it is assumed that clustering algorithm restrain, terminate iteration.Since the least unit of pixel is 1 picture
Element, and 0.5 pixel is the precision of the sub-pix of image procossing, thus the present invention using 0.5 as cluster iteration convergence threshold.
2) coarse segmentation is carried out to image using cluster result, then carries out morphology closed operation, removal small holes obtain described
Coarse segmentation figure.
Example IV
On the basis of any of embodiment one to three provides the Target Segmentation method based on coloured image, step
S103 extracts the process of the profile of target using dividing ridge method from image after the smothing filtering, specifically can be by such as
Under type is realized:
By means of target label figure, watershed algorithm proceeds by " water filling " from a group echo pixel set, just obtains
Then profile information, removal interference are analyzed in " watershed " in marked region, i.e. objective contour.When the identical basin of two labels
When the water on ground crosses, it is not generate watershed, also just reduces over-segmentation in this way.
As shown in figure 3, for according to the image segmentation flow chart of the present invention.There are two the watershed algorithm of step S103 has altogether
Input, one is image after mean shift smothing filterings, the other is target label figure.Target label figure is equivalent to
The prior information of watershed algorithm, algorithm are iterated search by means of prior information, finally extract and are present in target label
Objective contour in graph region.
Embodiment five
As shown in figure 4, the Target Segmentation device provided in an embodiment of the present invention based on coloured image, may include:Image
Module 403 is cut in filter module 401, coarse segmentation module 402 and subdivision.
Image filtering module 401 is used to carry out smothing filtering to input picture using mean-shift method.The image filtering
The operation that module 401 executes is identical as step S101 in preceding method.
Coarse segmentation module 402 is used for using K mean cluster method to image after 401 smothing filtering of described image filter module
Grey level histogram carry out cluster row coarse segmentation, obtain coarse segmentation figure, morphologic etching operation carried out to the coarse segmentation figure
Target label figure is built with expansive working.The operation that the coarse segmentation module 402 executes is identical as step S102 in preceding method.Its
Middle coarse segmentation module 402 builds target label figure in the following manner:Mesh is obtained after carrying out corrosion treatment to the coarse segmentation figure
Target inner marker figure;The external label figure of target is obtained after carrying out expansion process to the coarse segmentation figure;By the internal mark
Note figure and external label figure are in conjunction with obtaining the target label figure.
Subdivision cut module 403 for be based on the target label figure, schemed after the smothing filtering using dividing ridge method
The profile of target is extracted as in.The operation that the execution of module 403 is cut in the subdivision is identical as step S103 in preceding method.
Optionally, image filtering module 401 includes:Image conversion unit and image filtering unit.Image conversion unit is used
In input picture is transformed into CIELAB color spaces.
Image filtering unit, for each pixel on converted images to be denoted as five dimension point Z=(Xs,Xy), wherein Xs
Indicate the space variable (x, y) of pixel, XrThe color variables (L, a, b) for indicating pixel calculate five dimension point Z by kernel function
Mean shift variate-value, and constantly iteration until its stablize, finally by stable color variables (L, a, b) assignment to initially
Point on.The image filtering unit considers pixel shadow all in the kernel function spatial window around the pixel in an iterative process
The threshold value rung, and mean shift variable is arranged is 3, the loop termination when mean shift variate-value is less than 3, determines to be in and stablizes shape
State.
Optionally, coarse segmentation module 402 carries out cluster row coarse segmentation in the following manner:
1) divide equally the tonal range of grey level histogram using preset spacing H, and found out in image in each tonal range
The maximum gray value of the frequency of gray value within this range is clustered as the initial cluster center of this tonal range,
When the center of cluster, variation is less than 0.5 pixel, then it is assumed that clustering algorithm is restrained, and iteration is terminated;
2) coarse segmentation is carried out to image using cluster result, then carries out morphology closed operation, removal small holes obtain described
Coarse segmentation figure.
It is further to note that the Target Segmentation device provided in an embodiment of the present invention based on coloured image, Ke Yitong
Software realization is crossed, can also be realized by way of hardware or software and hardware combining.For hardware view, as shown in figure 5, being
A kind of hardware structure diagram of equipment where fine motion feature acquisition device provided in an embodiment of the present invention, in addition to processing shown in fig. 5
Except device, memory, network interface and nonvolatile memory, the equipment in embodiment where device usually can also include
Other hardware, such as it is responsible for the forwarding chip of processing message.For implemented in software, as shown in Figure 1, anticipating as a logic
Device in justice is to be read corresponding computer program instructions in nonvolatile memory by the CPU of equipment where it
Operation is formed in memory.For example, Target Segmentation method proposed by the present invention is realized using C Plus Plus programming, in Microsoft
It compiles and runs under Visual Studio 2012.
Fig. 6 is a series of pictures effect handled according to methods and apparatus of the present invention.The original of wherein the first behavior input
Beginning coloured image, the second behavior carries out the image after Mean Shift the disposal of gentle filter, after third behavior K-means coarse segmentations
Image, fourth line is target label figure, fifth line be according to the objective contour figure that is partitioned into of the present invention, the 6th performance-based objective
Actual profile figure.As shown, the present invention has extraordinary segmentation effect for the target under complex background.
In conclusion the Target Segmentation method and device provided in an embodiment of the present invention based on coloured image, mainly for
The segmentations of targets such as the aircraft under the more complex backgrounds such as sandstone, meadow, the woods in field and design.In visible images
In, first coloured image is smoothed using Mean Shift methods, later, using K-means clustering methods to gray scale
Image carries out coarse segmentation, and combining form corrodes expansive working structure label figure, the input as watershed algorithm.It calculates in watershed
Method extracts the profile of target according to label figure from smooth pretreated figure, realizes Target Segmentation.
Finally it should be noted that:The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
Present invention has been described in detail with reference to the aforementioned embodiments, it will be understood by those of ordinary skill in the art that:It still may be used
With technical scheme described in the above embodiments is modified or equivalent replacement of some of the technical features;
And these modifications or replacements, various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of Target Segmentation method based on coloured image, which is characterized in that including:
Smothing filtering is carried out to input picture using mean-shift method;
Cluster row coarse segmentation is carried out to the grey level histogram of image after smothing filtering using K mean cluster method, obtains coarse segmentation
Figure carries out morphologic etching operation to the coarse segmentation figure and expansive working builds target label figure;
Based on the target label figure, the profile of target is extracted from image after the smothing filtering using dividing ridge method.
2. according to the method described in claim 1, it is characterized in that, described put down input picture using mean-shift method
Sliding filtering, including:
Input picture is transformed into CIELAB color spaces;
Each pixel on converted images is denoted as five dimension point Z=(Xs,Xy), wherein XsIndicate the space variable of pixel
(x, y), XrThe color variables (L, a, b) for indicating pixel calculate the mean shift variate-value of five dimension point Z by kernel function, and
And constantly iteration, finally will be in stable color variables (L, a, b) assignment to initial point until its stabilization.
3. according to the method described in claim 2, it is characterized in that, considering that the kernel function around the pixel is empty in iterative process
Between pixel all in the window threshold value that influences, and mean shift variable is set be 3, followed when mean shift variate-value is less than 3
Ring terminates, and determines and is in stable state.
4. method described in any one of claim 1 to 3, which is characterized in that described to use K mean cluster method to flat
The grey level histogram of sliding filtered image carries out cluster row coarse segmentation, including:
Divide equally the tonal range of grey level histogram using preset spacing H, and finds out gray value in image in each tonal range
The maximum gray value of frequency within this range is clustered as the initial cluster center of this tonal range, works as cluster
Center variation be less than 0.5 pixel when, then it is assumed that clustering algorithm restrain, terminate iteration;
Coarse segmentation is carried out to image using cluster result, then carries out morphology closed operation, removal small holes obtain the coarse segmentation
Figure.
5. method described in any one of claim 1 to 3, which is characterized in that described to carry out shape to the coarse segmentation figure
The etching operation of state and expansive working build target label figure, including:
The inner marker figure of target is obtained after carrying out corrosion treatment to the coarse segmentation figure;
The external label figure of target is obtained after carrying out expansion process to the coarse segmentation figure;
By the inner marker figure and external label figure in conjunction with obtaining the target label figure.
6. a kind of Target Segmentation device based on coloured image, which is characterized in that including:
Image filtering module, for carrying out smothing filtering to input picture using mean-shift method;
Coarse segmentation module, for straight to the gray scale of image after described image filter module smothing filtering using K mean cluster method
Square figure carries out cluster row coarse segmentation, obtains coarse segmentation figure, carries out morphologic etching operation to the coarse segmentation figure and expansion is grasped
Make structure target label figure;
Module is cut in subdivision, for being based on the target label figure, using dividing ridge method from being carried in image after the smothing filtering
Take out the profile of target.
7. device according to claim 6, which is characterized in that described image filter module includes:
Image conversion unit, for input picture to be transformed into CIELAB color spaces;
Image filtering unit, for each pixel on converted images to be denoted as five dimension point Z=(Xs,Xy), wherein XsIt indicates
The space variable (x, y) of pixel, XrThe color variables (L, a, b) for indicating pixel calculate the equal of five dimension point Z by kernel function
It is worth offset variable value, and constantly iteration is until its stabilization, finally by stable color variables (L, a, b) assignment to initial point
On.
8. device according to claim 7, which is characterized in that described image filter unit considers the picture in an iterative process
All pixels influence in kernel function spatial window around vegetarian refreshments, and the threshold value that mean shift variable is arranged is 3, when mean value is inclined
Loop termination when variate-value is less than 3 is moved, determines and is in stable state.
9. the device according to any one of claim 6~8, which is characterized in that the coarse segmentation module passes through with lower section
Formula carries out cluster row coarse segmentation:
Divide equally the tonal range of grey level histogram using preset spacing H, and finds out gray value in image in each tonal range
The maximum gray value of frequency within this range is clustered as the initial cluster center of this tonal range, works as cluster
Center variation be less than 0.5 pixel when, then it is assumed that clustering algorithm restrain, terminate iteration;
Coarse segmentation is carried out to image using cluster result, then carries out morphology closed operation, removal small holes obtain the coarse segmentation
Figure.
10. the device according to any one of claim 6~8, which is characterized in that the coarse segmentation module passes through with lower section
Formula builds target label figure:
The inner marker figure of target is obtained after carrying out corrosion treatment to the coarse segmentation figure;
The external label figure of target is obtained after carrying out expansion process to the coarse segmentation figure;
By the inner marker figure and external label figure in conjunction with obtaining the target label figure.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810408454.0A CN108596920A (en) | 2018-05-02 | 2018-05-02 | A kind of Target Segmentation method and device based on coloured image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810408454.0A CN108596920A (en) | 2018-05-02 | 2018-05-02 | A kind of Target Segmentation method and device based on coloured image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108596920A true CN108596920A (en) | 2018-09-28 |
Family
ID=63619513
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810408454.0A Pending CN108596920A (en) | 2018-05-02 | 2018-05-02 | A kind of Target Segmentation method and device based on coloured image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108596920A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110399840A (en) * | 2019-05-22 | 2019-11-01 | 西南科技大学 | A kind of quick lawn semantic segmentation and boundary detection method |
CN110992376A (en) * | 2019-11-28 | 2020-04-10 | 北京推想科技有限公司 | CT image-based rib segmentation method, device, medium and electronic equipment |
CN111210452A (en) * | 2019-12-30 | 2020-05-29 | 西南交通大学 | Certificate photo portrait segmentation method based on graph segmentation and mean shift |
CN111462143A (en) * | 2020-03-22 | 2020-07-28 | 华中科技大学 | Watershed algorithm-based insect body recognition and counting method and system |
CN112183556A (en) * | 2020-09-27 | 2021-01-05 | 长光卫星技术有限公司 | Port ore heap contour extraction method based on spatial clustering and watershed transformation |
CN112215852A (en) * | 2020-09-29 | 2021-01-12 | 忻州师范学院 | Digital image segmentation method based on cluster learning device integration |
CN112308077A (en) * | 2020-11-02 | 2021-02-02 | 中科麦迪人工智能研究院(苏州)有限公司 | Sample data acquisition method, image segmentation method, device, equipment and medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853495A (en) * | 2010-06-03 | 2010-10-06 | 浙江工业大学 | Cell separation method based on morphology |
CN102222234A (en) * | 2011-07-14 | 2011-10-19 | 苏州两江科技有限公司 | Image object extraction method based on mean shift and K-means clustering technology |
CN104732229A (en) * | 2015-03-16 | 2015-06-24 | 华南理工大学 | Segmentation method for overlapping cells in cervical smear image |
CN104966085A (en) * | 2015-06-16 | 2015-10-07 | 北京师范大学 | Remote sensing image region-of-interest detection method based on multi-significant-feature fusion |
CN106296675A (en) * | 2016-08-04 | 2017-01-04 | 山东科技大学 | A kind of dividing method of the uneven image of strong noise gray scale |
CN106599793A (en) * | 2016-11-21 | 2017-04-26 | 江苏大学 | Marked watershed segmentation-based steel grain boundary automatic extraction method |
-
2018
- 2018-05-02 CN CN201810408454.0A patent/CN108596920A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853495A (en) * | 2010-06-03 | 2010-10-06 | 浙江工业大学 | Cell separation method based on morphology |
CN102222234A (en) * | 2011-07-14 | 2011-10-19 | 苏州两江科技有限公司 | Image object extraction method based on mean shift and K-means clustering technology |
CN104732229A (en) * | 2015-03-16 | 2015-06-24 | 华南理工大学 | Segmentation method for overlapping cells in cervical smear image |
CN104966085A (en) * | 2015-06-16 | 2015-10-07 | 北京师范大学 | Remote sensing image region-of-interest detection method based on multi-significant-feature fusion |
CN106296675A (en) * | 2016-08-04 | 2017-01-04 | 山东科技大学 | A kind of dividing method of the uneven image of strong noise gray scale |
CN106599793A (en) * | 2016-11-21 | 2017-04-26 | 江苏大学 | Marked watershed segmentation-based steel grain boundary automatic extraction method |
Non-Patent Citations (3)
Title |
---|
DINGDING LIU 等: "Robust interactive image segmentation with automatic boundary refinement", 《2010 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING》 * |
LIU, D 等: "A Review of Computer Vision Segmentation Algorithms", 《COURSES.CS.WASHINGTON.EDU》 * |
SHENGYANG DAI 等: "Color image segmentation with watershed on color histogram and Markov random fields", 《FOURTH INTERNATIONAL CONFERENCE ON INFORMATION, COMMUNICATIONS AND SIGNAL PROCESSING》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110399840A (en) * | 2019-05-22 | 2019-11-01 | 西南科技大学 | A kind of quick lawn semantic segmentation and boundary detection method |
CN110399840B (en) * | 2019-05-22 | 2024-04-02 | 西南科技大学 | Rapid lawn semantic segmentation and boundary detection method |
CN110992376A (en) * | 2019-11-28 | 2020-04-10 | 北京推想科技有限公司 | CT image-based rib segmentation method, device, medium and electronic equipment |
CN111210452A (en) * | 2019-12-30 | 2020-05-29 | 西南交通大学 | Certificate photo portrait segmentation method based on graph segmentation and mean shift |
CN111210452B (en) * | 2019-12-30 | 2023-04-07 | 西南交通大学 | Certificate photo portrait segmentation method based on graph segmentation and mean shift |
CN111462143A (en) * | 2020-03-22 | 2020-07-28 | 华中科技大学 | Watershed algorithm-based insect body recognition and counting method and system |
CN111462143B (en) * | 2020-03-22 | 2022-12-02 | 华中科技大学 | Watershed algorithm-based insect body recognition and counting method and system |
CN112183556A (en) * | 2020-09-27 | 2021-01-05 | 长光卫星技术有限公司 | Port ore heap contour extraction method based on spatial clustering and watershed transformation |
CN112183556B (en) * | 2020-09-27 | 2022-08-30 | 长光卫星技术股份有限公司 | Port ore heap contour extraction method based on spatial clustering and watershed transformation |
CN112215852A (en) * | 2020-09-29 | 2021-01-12 | 忻州师范学院 | Digital image segmentation method based on cluster learning device integration |
CN112308077A (en) * | 2020-11-02 | 2021-02-02 | 中科麦迪人工智能研究院(苏州)有限公司 | Sample data acquisition method, image segmentation method, device, equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108596920A (en) | A kind of Target Segmentation method and device based on coloured image | |
CN108537239B (en) | Method for detecting image saliency target | |
CN105139015B (en) | A kind of remote sensing images Clean water withdraw method | |
Jin et al. | Left atrial appendage segmentation using fully convolutional neural networks and modified three-dimensional conditional random fields | |
CN108876820B (en) | Moving target tracking method under shielding condition based on mean shift | |
CN108664838A (en) | Based on the monitoring scene pedestrian detection method end to end for improving RPN depth networks | |
CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
CN111160407A (en) | Deep learning target detection method and system | |
CN109886128A (en) | A kind of method for detecting human face under low resolution | |
CN109345536B (en) | Image super-pixel segmentation method and device | |
CN106446925A (en) | Dolphin identity recognition method based on image processing | |
CN107016680B (en) | A kind of pest image background minimizing technology based on conspicuousness detection | |
Wu et al. | Cartoon image segmentation based on improved SLIC superpixels and adaptive region propagation merging | |
CN110070545A (en) | A kind of method that textural characteristics density in cities and towns automatically extracts cities and towns built-up areas | |
Katkar et al. | A novel approach for medical image segmentation using PCA and K-means clustering | |
CN111027586A (en) | Target tracking method based on novel response map fusion | |
JP2021517281A (en) | Multi-gesture fine division method for smart home scenes | |
CN110473224B (en) | Automatic RSF level set image segmentation method based on KL entropy | |
CN113379789B (en) | Moving target tracking method in complex environment | |
CN109508674A (en) | Airborne lower view isomery image matching method based on region division | |
CN109359653A (en) | A kind of cotton leaf portion adhesion scab image partition method and system | |
Yin et al. | Study and application of improved level set method with prior graph cut in PCB image segmentation | |
Lin et al. | In defense of iterated conditional mode for hyperspectral image classification | |
Orrite et al. | Portrait segmentation by deep refinement of image matting | |
Zhang et al. | Visual tracking algorithm for aircrafts in airport |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180928 |
|
RJ01 | Rejection of invention patent application after publication |