CN108765426A - automatic image segmentation method and device - Google Patents
automatic image segmentation method and device Download PDFInfo
- Publication number
- CN108765426A CN108765426A CN201810463978.XA CN201810463978A CN108765426A CN 108765426 A CN108765426 A CN 108765426A CN 201810463978 A CN201810463978 A CN 201810463978A CN 108765426 A CN108765426 A CN 108765426A
- Authority
- CN
- China
- Prior art keywords
- image
- region
- image segmentation
- pixels
- segmentation
- 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
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
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of Automatic image segmentation method and devices, the described method comprises the following steps:A, it is based on Graph-theoretical Approach and carries out image pixel division, thus divide the image into different zones;B, less divided detection is carried out to the region of image segmentation, corrects image cut zone;C, the degree of approximation of the image segmentation region after the detection of analysis less divided between any two, carries out the merging in image segmentation region if enough approximations.Automatic image segmentation method and device through the invention can realize quick, accurate, reliable image region segmentation, and segmentation result is made to be more nearly the visual effect of people.
Description
Technical field
The present invention relates to digital image processing techniques, especially relate to the image segmentation field in automated graphics identification.
Background technology
Image recognition refers to being handled image, analyzed and being understood using computer, to identify various different modes
Target and technology to picture.General industry using industrial camera in use, shoot picture, and then recycling software is according to picture ash
Scale does further identifying processing, and what image recognition software foreign countries represented has Cognex etc., and what the country represented has figure intelligence etc..Separately
The technology that external geography middle finger classifies remote sensing images.
Image recognition technology is a key areas of artificial intelligence.In order to work out the simulation movable meter of mankind's image recognition
Calculation machine program, there has been proposed different image recognition models.Such as Template matching model.This model is thought, identifies some
Image, it is necessary to which the memory pattern for having this image in past experience is called template.If current stimulation can in brain
Template match, this image is also just identified.Such as there are one letter A, if there is a A templates in brain, alphabetical A's
Size, orientation, shape are all completely the same with this A template, and alphabetical A is just identified.Pattern-recognition in image recognition
(PatternRecognition), it is a kind of from bulk information and data, in expertise and on the basis of be recognized,
Identification is automatically performed to shape, pattern, curve, number, character format and figure using the method for computer and mathematical reasoning, is commented
The process of valence.Pattern-recognition includes two stages, that is, learns stage and implementation phase, the former is to carry out feature selecting to sample,
The rule of classification is found, the latter is that unknown sample collection is classified and identified according to classification rule.The mould of this pattern-recognition
Plate Matching Model is simple and clear, is also easy to get practical application.But this model emphasizes that image must be complete with the template in brain
Meeting can just be identified, and in fact people can not only identify the image completely the same with the template in brain, can also be identified and mould
The not quite identical image of plate.
It is, in general, that the first step of image recognition is divided automatically to image, i.e., by computer by picture segmentation
At different regions, the identification of image could be carried out after being then further processed.Existing image partition method includes:Based on wheel
The method of exterior feature tracking, the method based on objective contour, the method based on average drifting, is based on figure at the method based on region-competitive
The method of opinion, and the dividing method based on study.How to evaluate the quality of segmentation result, and how according to evaluation result into
It is the problem of most existing methods are ignored that row, which improves,.
To image segmentation result evaluation be presently mainly the evaluation method based on supervision, that is, by segmentation result with profit
Employment is divided obtained fact (ground true) image and is compared by hand.The main problem of supervision type evaluation method is
The acquisition of live image.First, each image is required for people to be split by delineating by hand, and workload is huge, especially
For the ever-increasing large capacity image library of picture number, this method is difficult to practical application.Further, since different people pair
The understanding of same image is variant, and the result manually divided often is not quite similar, this is also affected to image segmentation quality evaluation
Objectivity.And if using the non-supervisory type evaluation method towards gray level image segmentation result, its textural characteristics to region
Analyzed, but not consider colouring information therefore textural characteristics it is relatively simple.Also for example using standard variance method,
But this Statistics-Based Method is excessively coarse, has the practical difference of its textural characteristics of mutually homoscedastic region in many cases still
It is so larger.
Invention content
It is an object of the invention to overcome, image partition method workload in the prior art is larger, efficiency is low;Or not
Consider colouring information therefore the shortcomings of relatively simple textural characteristics, a kind of Automatic image segmentation method and its device are provided
In order to solve the above technical problems, the present invention adopts the following technical scheme that.
A kind of Automatic image segmentation method, the described method comprises the following steps:
A, it is based on Graph-theoretical Approach and carries out image pixel division, thus divide the image into different zones;
B, less divided detection is carried out to the region of image segmentation, corrects image cut zone;
C, the detection of analysis less divided and the degree of approximation of revised image segmentation region between any two, if enough approximations
Carry out the merging in image segmentation region.
Wherein, described to include based on Graph-theoretical Approach progress image pixel division:
A1, the color similarity between two adjacent pixels is calculated using Euclidean distance method:
R '=(R1+R2)/2;
Δ R=R1-R2;
Δ G=G1-G2;
Δ B=B1-B2;
Wherein R1And R2For two respective red values of pixel;G1And G2For two respective green values of pixel;B1And B2For
Two respective blue valves of pixel;
ScFor the color similarity between two pixels;
A2, when the color similarity of two pixels be less than the first predetermined threshold when, the two pixels are included into the same area.
In addition, the region of described image segmentation includes the step of carrying out less divided detection, correct image cut zone:
B1, multi-times random sampling is carried out to the pixel in present image cut zone, each stochastical sampling obtains two pictures
Element;
B2, two pixels are obtained to every sub-sampling, obtain the neighborhood rectangle cell domain of the two pixels;
B3, the neighborhood rectangle cell domain to the two pixels, count the color histogram of wherein all pixels respectively, calculate
The similar value of two color histograms;
B3, to two pixels after repeatedly sampling, repeatedly B2, B3 step respectively, the color histogram phase that will be obtained every time
It is ranked up like value, by the color histogram similar value of the mean value of the color histogram similar value of preceding 1/4 size and rear 1/4 size
Mean value be compared, when difference be more than all colours histogram similar value mean value half when, be judged as less divided, need
The region of described image segmentation is divided again.
And the step of region progress less divided detection that described image divides, amendment image cut zone, further includes:
B4, when needing the region divided to described image to be divided again, to described image divide region in
Pixel executes step A1, A2 again, wherein being lowered to the first predetermined threshold in step A2.
Wherein, the number that stochastical sampling is carried out in present image cut zone is divided depending on present image in step B1
Total pixel quantity in region, total pixel quantity is more, then needs the number for carrying out stochastical sampling more.
In addition, the analysis less divided detection and the degree of approximation of revised image segmentation region between any two, if foot
Reaching the step of approximation then carries out the merging in image segmentation region includes:
C1, two less divideds are detected and carry out multiple window sample respectively in revised image segmentation region, every time
Two color histograms are respectively obtained after window sample, then calculate the color histogram similar value of the two color histograms;
C2, the color histogram similar value repeatedly sampled is averaged, the population mean as two sub-regions is similar
Degree.If the average similarity is less than the second predetermined threshold, by the detection of two less divideds and revised image segmentation region into
Row merges.
The invention also includes a kind of Automatic image segmentation device, the Automatic image segmentation device includes:
Thus image pre-segmentation unit divides the image into difference for carrying out image pixel division based on Graph-theoretical Approach
Region;
Less divided detection unit carries out less divided detection for the region to image segmentation, corrects image cut zone;
Over-segmentation compensating unit, for analyzing less divided detection and the approximation of revised image segmentation region between any two
Degree carries out the merging in image segmentation region if enough approximations.
First, Automatic image segmentation method and apparatus through the invention can realize quick, accurate, reliable image
Region segmentation, and segmentation result is made to be more nearly the visual effect of people.Secondly, Automatic image segmentation method and apparatus of the invention
The process of grab sample ensure that covering under the premise of smaller calculation amount to whole region, what each sub-sampling calculated is
The statistical value of all pixels color in sampling window, therefore the Automatic image segmentation method and apparatus of the present invention can combine face
The advantages of color and the method for two kinds of features of texture.
Description of the drawings
Fig. 1 is the flow diagram according to the Automatic image segmentation method of the specific embodiment of the invention.
Fig. 2 is the signal that adjacent pixel method is determined in the Automatic image segmentation method according to the specific embodiment of the invention
Figure.
Fig. 3 is the structural schematic diagram of the Automatic image segmentation device of the specific embodiment of the invention.
Specific implementation mode
Below in conjunction with the accompanying drawings, it elaborates to the present invention.
The detailed example embodiment of following discloses.However, specific structure and function details disclosed herein merely for the sake of
The purpose of example embodiment is described.
It should be appreciated, however, that the present invention is not limited to disclosed particular exemplary embodiment, but covering falls into disclosure model
Enclose interior all modifications, equivalent and alternative.In the description to whole attached drawings, identical reference numeral indicates identical member
Part.
Refering to attached drawing, structure, ratio, size etc. depicted in this specification institute accompanying drawings, only coordinating specification
Revealed content is not limited to the enforceable restriction item of the present invention so that those skilled in the art understands and reads
Part, therefore do not have technical essential meaning, the modification of any structure, the change of proportionate relationship or the adjustment of size are not influencing
Under the effect of present invention can be generated and the purpose that can reach, it should all still fall and obtain and can contain in disclosed technology contents
In the range of lid.Meanwhile cited position restriction term in this specification, it is merely convenient to being illustrated for narration, rather than to
It limits the scope of the invention, relativeness is altered or modified, in the case where changing technology contents without essence, when being also considered as
The enforceable scope of the present invention.
It will also be appreciated that term "and/or" includes the arbitrary of one or more relevant list items as used in this
With all combinations.It will further be appreciated that when component or unit are referred to as " connecting " or when " coupled " to another component or unit, it
Other component or unit are can be directly connected or coupled to, or there may also be intermediate member or units.In addition, for describing
Between component or unit other words of relationship should understand in the same fashion (for example, " between " to " directly between ",
" adjacent " is to " direct neighbor " etc.).
As shown in Figure 1, include a kind of Automatic image segmentation method in the specific embodiment of the invention, the method includes
Following steps:
A, it is based on Graph-theoretical Approach and carries out image pixel division, thus divide the image into different zones;
B, less divided detection is carried out to the region of image segmentation, corrects image cut zone;
C, the detection of analysis less divided and the degree of approximation of revised image segmentation region between any two, if enough approximations
Carry out the merging in image segmentation region.
So-called less divided should as be divided into the part of different zones because its with certain similitude and
Do not distinguish;And so-called over-segmentation, it should as belong to the part of the same area, because of excessively stringent screening criteria
And divided for different zones, either less divided or over-segmentation by mistake, all it is the error result in Automatic image segmentation.
Using the Automatic image segmentation method in the specific embodiment of the invention, image can be accurately distinguished, and can be by owing to divide
The monitoring with over-segmentation is cut, to avoid less divided and over-segmentation, therefore there is higher accuracy compared with the existing technology.
In a specific embodiment, described to include based on Graph-theoretical Approach progress image pixel division:
A1, the color similarity between two adjacent pixels is calculated using Euclidean distance method:
R '=(R1+R2)/2;
Δ R=R1-R2;
Δ G=G1-G2;
Δ B=B1-B2;
Wherein R1And R2For two respective red values of pixel;G1And G2For two respective green values of pixel;B1And B2For
Two respective blue valves of pixel;
ScFor the color similarity between two pixels;
A2, when the color similarity of two pixels be less than the first predetermined threshold when, the two pixels are included into the same area,
Wherein described first predetermined threshold rule of thumb obtains.
It is, in general, that can have larger color distortion between the pixel of different zones, therefore, using based on Graph-theoretical Approach
Image pixel division is carried out, can different zones simply and effectively be carried out to initial division using rgb value.
When executing above step A1, A2, need to be included in each pixel in the way of " snowball " same or different
Region, specifically it is first determined two adjacent pixels.
As shown in Fig. 2, the method for determining adjacent pixel is to be based on a source pixel, the pixel within the scope of its distance R is looked for,
Such as first since A pixels, neighborhood is square region, the pixels such as including B, C, D.If true by step A1, A2
After determining A, B pixel and should being included into the same area, because A pixels are source pixels and B pixels are edge pixels, then according to the side of Fig. 2
Formula looks for next analysis pixel using the neighborhood of the other side of B pixels as range, and that whether A, B pixel should be included into is same
Region.And if after by step A1, A2, B pixels should not be included into the same area with A pixels, then again using B pixels as source image
Element finds the pixel in its neighborhood, analyzes whether it should be included into the same area with B pixels.
By above method, all pixels on image can be divided into different regions automatically, realized quick, accurate
Really, reliable image region segmentation, and image segmentation result is made to be more nearly the visual effect of people.
In addition, in one specific implementation mode of this patent, the region of described image segmentation carries out less divided detection, corrects
The step of image segmentation region includes:
B1, multi-times random sampling is carried out to the pixel in present image cut zone, each stochastical sampling obtains two pictures
Element;
B2, two pixels are obtained to every sub-sampling, obtain the neighborhood rectangle cell domain of the two pixels;
B3, the neighborhood rectangle cell domain to the two pixels, count the color histogram of wherein all pixels respectively, calculate
The similar value of two color histograms;
B3, to two pixels after repeatedly sampling, repeatedly B2, B3 step respectively, the color histogram phase that will be obtained every time
It is ranked up like value, by the color histogram similar value of the mean value of the color histogram similar value of preceding 1/4 size and rear 1/4 size
Mean value be compared, when difference be more than all colours histogram similar value mean value half when, be judged as less divided, need
Again divided in the region that described image is divided.
Specifically, to the neighborhood rectangle cell domain of a pixel, the color histogram of statistics wherein all pixels
Histi, in order to improve computational efficiency, the color value of Histi is one dimensional numerical, be by corresponding to hsv color space coloration H,
What saturation degree S and intensity V was obtained by way of bit combination.
The judgement of less divided is the similar Distribution value of color histogram according to gross sample, is obtained by statistical analysis.
The color histogram similar value of the N sampled first to n times carries out ascending ascending order arrangement, obtains:d1,d2…dn.It is right
Preceding N/4 and rear N/4 color histogram similar value calculate its average value and are:
Judging whether the foundation of less divided is:
The above distinguishing rule is obtained using many experiments, the color histogram similar value of current 1/4 size it is equal
The mean value of the color histogram similar value of value and rear 1/4 size is compared, when difference is more than all colours histogram similar value
Mean value half when, be judged as less divided, 90% or more less divided can be eliminated in this way, there is very high efficiency and accurate
Property.
And in another specific implementation mode, less divided detection is carried out to the region that described image divides, is corrected
The step of image segmentation region further includes:
B4, when needing the region divided to described image to be divided again, to described image segmentation region in
Pixel executes step A1, A2 again, wherein being lowered to the first predetermined threshold in step A2.
When needing to divide again, illustrate that region division is too rough in step A, it should the region after dividing step A
Further subdivision, therefore should adjust the first predetermined threshold in A2, such as make its drop by half, can allow more pictures in this way
Can be considered as similarity between element cannot meet the first predetermined threshold, different regions should be classified as, such that by former
The region division of first less divided is more zonules.
Wherein, the number that stochastical sampling is carried out in present image cut zone is divided depending on present image in step B1
Total pixel quantity in region, total pixel quantity is more, then needs the number for carrying out stochastical sampling more.
In addition, the analysis less divided detection and the degree of approximation of revised image segmentation region between any two, if foot
Reaching the step of approximation then carries out the merging in image segmentation region includes:
C1, two less divideds are detected and carry out multiple window sample respectively in revised image segmentation region, every time
Two color histograms are respectively obtained after window sample, then calculate the color histogram similar value of the two color histograms;
C2, the color histogram similar value repeatedly sampled is averaged, the population mean as two sub-regions is similar
Degree.If the average similarity is less than the second predetermined threshold, by the detection of two less divideds and revised image segmentation region into
Row merges.
Above C1, C2 step is the similar processes of step B in fact, is only sampled from adjacent subregion, and it is straight to calculate color
Square figure similar value, therefore to determine whether after step B, whether the result that drawing is divided is excessively in disorder.
As shown in figure 3, corresponding with the Automatic image segmentation method in the specific embodiment of the invention, of the invention is specific
Further include a kind of Automatic image segmentation device in embodiment, the Automatic image segmentation device includes:
Thus image pre-segmentation unit divides the image into difference for carrying out image pixel division based on Graph-theoretical Approach
Region;
Less divided detection unit carries out less divided detection for the region to image segmentation, corrects image cut zone;
Over-segmentation compensating unit, for analyzing less divided detection and the approximation of revised image segmentation region between any two
Degree carries out the merging in image segmentation region if enough approximations.
By described above as it can be seen that the random of Automatic image segmentation method and apparatus in the specific embodiment of the invention takes
The process of sample ensure that the covering to whole region under the premise of smaller calculation amount, and what each sub-sampling calculated is sampling window
The statistical value of interior all pixels color, therefore the Automatic image segmentation method and apparatus of the present invention can combine color and texture
The advantages of method of two kinds of features.
It should be noted that the above embodiment is only the preferable embodiment of the present invention, cannot be understood as to this
The limitation of invention protection domain, under the premise of without departing from present inventive concept, to any minor variations and modifications of the invention done
It all belongs to the scope of protection of the present invention.
Claims (7)
1. a kind of Automatic image segmentation method, the described method comprises the following steps:
A, it is based on Graph-theoretical Approach and carries out image pixel division, thus divide the image into different zones;
B, less divided detection is carried out to the region of image segmentation, corrects image cut zone;
C, the detection of analysis less divided and the degree of approximation of revised image segmentation region between any two carry out if enough approximations
The merging of described image cut zone.
2. according to the Automatic image segmentation method described in claim 1, which is characterized in that described to carry out figure based on Graph-theoretical Approach
As pixel division includes:
A1, the color similarity between two adjacent pixels is calculated using Euclidean distance method:
R '=(R1+R2)/2;
Δ R=R1-R2;
Δ G=G1-G2;
Δ B=B1-B2;
Wherein R1And R2For two respective red values of pixel;G1And G2For two respective green values of pixel;B1And B2It is two
The respective blue valve of pixel;
ScFor the color similarity between two pixels;
A2, when the color similarity of two pixels be less than the first predetermined threshold when, the two pixels are included into the same area.
3. according to the Automatic image segmentation method described in claim 2, which is characterized in that the region of described image segmentation carries out
Less divided detect, correct image cut zone the step of include:
B1, multi-times random sampling is carried out to the pixel in present image cut zone, each stochastical sampling obtains two pixels;
B2, two pixels are obtained to every sub-sampling, obtain the neighborhood rectangle cell domain of the two pixels;
B3, the neighborhood rectangle cell domain to the two pixels count the color histogram of wherein all pixels respectively, calculate two
The similar value of color histogram;
B3, to two pixels after repeatedly sampling, repeatedly B2, B3 step respectively, the color histogram similar value that will be obtained every time
It is ranked up, by the equal of the color histogram similar value of the mean value of the color histogram similar value of preceding 1/4 size and rear 1/4 size
Value is compared, and when difference is more than the half of the mean value of all colours histogram similar value, is judged as less divided, is needed to institute
Again divided in the region for stating image segmentation.
4. according to the Automatic image segmentation method described in claim 3, which is characterized in that the region that described image divides carries out
Less divided detect, correct image cut zone the step of further include:
B4, when needing the region divided to described image to be divided again, to described image divide region in pixel
Step A1, A2 is executed again, wherein being lowered to the first predetermined threshold in step A2.
5. according to the Automatic image segmentation method described in claim 3, which is characterized in that divide present image in step B1
The number of stochastical sampling is carried out in region depending on total pixel quantity in present image cut zone, total pixel quantity is got over
It is more, then need the number for carrying out stochastical sampling more.
6. according to the Automatic image segmentation method described in claim 3, which is characterized in that the analysis less divided is detected and repaiied
The degree of approximation of image segmentation region between any two after just carries out the merging of described image cut zone if enough approximations
Step includes:
C1, two less divideds are detected and carry out multiple window sample, each window respectively in revised image segmentation region
Two color histograms are respectively obtained after sampling, then calculate the color histogram similar value of the two color histograms;
C2, the color histogram similar value repeatedly sampled is averaged, the population mean similarity as two sub-regions.If
The average similarity is less than the second predetermined threshold, then detects two less divideds and revised image segmentation region is closed
And.
7. a kind of Automatic image segmentation device, including:
Thus image pre-segmentation unit divides the image into different zones for carrying out image pixel division based on Graph-theoretical Approach;
Less divided detection unit carries out less divided detection for the region to image segmentation, corrects image cut zone;
Over-segmentation compensating unit, for analyzing less divided detection and the degree of approximation of revised image segmentation region between any two,
The merging of described image cut zone is carried out if enough approximations.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810463978.XA CN108765426A (en) | 2018-05-15 | 2018-05-15 | automatic image segmentation method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810463978.XA CN108765426A (en) | 2018-05-15 | 2018-05-15 | automatic image segmentation method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108765426A true CN108765426A (en) | 2018-11-06 |
Family
ID=64007852
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810463978.XA Pending CN108765426A (en) | 2018-05-15 | 2018-05-15 | automatic image segmentation method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108765426A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977745A (en) * | 2018-12-25 | 2019-07-05 | 深圳云天励飞技术有限公司 | Face image processing process and relevant apparatus |
CN110910396A (en) * | 2019-10-18 | 2020-03-24 | 北京量健智能科技有限公司 | Method and device for optimizing image segmentation result |
CN112274110A (en) * | 2020-10-10 | 2021-01-29 | 苏州万微光电科技有限公司 | Pore detection system, device and method based on skin fluorescence image |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002093450A1 (en) * | 2001-05-16 | 2002-11-21 | Cellavision Ab | Information processing for distinguishing an object |
CN101710387A (en) * | 2009-10-29 | 2010-05-19 | 中国科学院对地观测与数字地球科学中心 | Intelligent method for classifying high-resolution remote sensing images |
CN102968822A (en) * | 2012-08-23 | 2013-03-13 | 华南理工大学 | Three-dimensional medical image segmentation method based on graph theory |
CN103778413A (en) * | 2014-01-16 | 2014-05-07 | 华东师范大学 | Remote-sensing image under-segmentation object automatic recognition method |
CN105335965A (en) * | 2015-09-29 | 2016-02-17 | 中国科学院遥感与数字地球研究所 | High-resolution remote sensing image multi-scale self-adaptive decision fusion segmentation method |
CN107767383A (en) * | 2017-11-01 | 2018-03-06 | 太原理工大学 | A kind of Road image segmentation method based on super-pixel |
-
2018
- 2018-05-15 CN CN201810463978.XA patent/CN108765426A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002093450A1 (en) * | 2001-05-16 | 2002-11-21 | Cellavision Ab | Information processing for distinguishing an object |
CN101710387A (en) * | 2009-10-29 | 2010-05-19 | 中国科学院对地观测与数字地球科学中心 | Intelligent method for classifying high-resolution remote sensing images |
CN102968822A (en) * | 2012-08-23 | 2013-03-13 | 华南理工大学 | Three-dimensional medical image segmentation method based on graph theory |
CN103778413A (en) * | 2014-01-16 | 2014-05-07 | 华东师范大学 | Remote-sensing image under-segmentation object automatic recognition method |
CN105335965A (en) * | 2015-09-29 | 2016-02-17 | 中国科学院遥感与数字地球研究所 | High-resolution remote sensing image multi-scale self-adaptive decision fusion segmentation method |
CN107767383A (en) * | 2017-11-01 | 2018-03-06 | 太原理工大学 | A kind of Road image segmentation method based on super-pixel |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109977745A (en) * | 2018-12-25 | 2019-07-05 | 深圳云天励飞技术有限公司 | Face image processing process and relevant apparatus |
CN109977745B (en) * | 2018-12-25 | 2021-09-14 | 深圳云天励飞技术有限公司 | Face image processing method and related device |
CN110910396A (en) * | 2019-10-18 | 2020-03-24 | 北京量健智能科技有限公司 | Method and device for optimizing image segmentation result |
CN112274110A (en) * | 2020-10-10 | 2021-01-29 | 苏州万微光电科技有限公司 | Pore detection system, device and method based on skin fluorescence image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN115082683B (en) | Injection molding defect detection method based on image processing | |
CN106651872B (en) | Pavement crack identification method and system based on Prewitt operator | |
Funck et al. | Image segmentation algorithms applied to wood defect detection | |
CN103035013B (en) | A kind of precise motion shadow detection method based on multi-feature fusion | |
CN109978822B (en) | Banana maturity judging modeling method and judging method based on machine vision | |
CN102426649B (en) | Simple steel seal digital automatic identification method with high accuracy rate | |
US9740957B2 (en) | Learning pixel visual context from object characteristics to generate rich semantic images | |
EP3343440A1 (en) | Identifying and excluding blurred areas of images of stained tissue to improve cancer scoring | |
CN116205919B (en) | Hardware part production quality detection method and system based on artificial intelligence | |
Ilonen et al. | Comparison of bubble detectors and size distribution estimators | |
CN108181316B (en) | Bamboo strip defect detection method based on machine vision | |
CN111815564B (en) | Method and device for detecting silk ingots and silk ingot sorting system | |
CN107610114A (en) | Optical satellite remote sensing image cloud snow mist detection method based on SVMs | |
CN108319973A (en) | Citrusfruit detection method on a kind of tree | |
CN111242896A (en) | Color printing label defect detection and quality rating method | |
Niu et al. | Image segmentation algorithm for disease detection of wheat leaves | |
Galsgaard et al. | Circular hough transform and local circularity measure for weight estimation of a graph-cut based wood stack measurement | |
CN107154044B (en) | Chinese food image segmentation method | |
CN117253024B (en) | Industrial salt quality inspection control method and system based on machine vision | |
CN108765426A (en) | automatic image segmentation method and device | |
CN112580647A (en) | Stacked object oriented identification method and system | |
Yingthawornsuk et al. | Automatic Thai Coin Calculation System by Using SIFT | |
CN111695373A (en) | Zebra crossing positioning method, system, medium and device | |
Saidane et al. | Robust binarization for video text recognition | |
CN108073940A (en) | A kind of method of 3D object instance object detections in unstructured moving grids |
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: 20181106 |
|
RJ01 | Rejection of invention patent application after publication |