CN105976385B - A kind of image partition method based on image data field - Google Patents
A kind of image partition method based on image data field Download PDFInfo
- Publication number
- CN105976385B CN105976385B CN201610335491.4A CN201610335491A CN105976385B CN 105976385 B CN105976385 B CN 105976385B CN 201610335491 A CN201610335491 A CN 201610335491A CN 105976385 B CN105976385 B CN 105976385B
- Authority
- CN
- China
- Prior art keywords
- image
- value
- pixel
- point
- seed
- 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
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20156—Automatic seed setting
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of image partition methods based on image data field.Existing monochrome image dividing method is typically based on the discontinuity and similitude of grey scale pixel value；Existing partitioning algorithm hardly results in ideal segmentation effect under conditions of uneven illumination is even.Therefore, the present invention carries out pretreatment filtering to the image of input first, is then split with the watershed algorithm based on eight direction gradients；Then the average gray value of marginal point is obtained to the image progress edge detection after segmentation and is set as seed；Image data fields are constructed simultaneously, calculate the gesture value function of pixel；Finally seed region automatic growth is carried out according to seed obtained above and image data field.The present invention has biggish promotion to the segmentation effect of the even image of uneven illumination and sliced time, can satisfy the demand of image recognition.
Description
Technical field
The invention belongs to image segmentation fields, are related to a kind of quick Accurate Segmentation method of even image of uneven illumination.
Background technique
Image segmentation, which refers to, segments the image into many continuous sub-districts by the similitude or diversity of characteristics of image
Domain, and interested part is extracted from subregion.In practical applications, these available features mainly include gray scale,
For different image processing requirements thousands of kinds of processing methods have been proposed, but not in texture, edge, color etc. at present
A kind of general quick accurate dividing method.
In engineer application, it is a weight of image processing system that image segmentation, which is to realize the committed step of image analysis,
Component part is wanted, there is very high practical value, is the subsequent basis for carrying out image analysis, segmentation quality will directly affect last point
Analysis as a result, so the research of accuracy for improve image segmentation is necessary.
Summary of the invention
In view of the deficiencies of the prior art, the present invention provides a kind of quick Accurate Segmentation sides for the even image of uneven illumination
Method.
The technical solution adopted for solving the technical problem of the present invention are as follows:
The input of step (1) image and image preprocessing.
Step (2) carries out image segmentation with the watershed algorithm based on eight direction gradients.
Step (3) constructs image data fields, calculates the gesture value function of each pixel.
Step (4) edge detection, calculates the average gray value of marginal point and is set as seed point value.
Step (5) combination step (3), the data fields of (4) and seed point carry out automated seed region growing.
Step (6) homogeneous region fusion output segmentation effect figure.
Beneficial effects of the present invention:
(1) watershed algorithm for using eight direction gradients, takes full advantage of the marginal information of image, reduces over-segmentation
Degree.
(2) image data field is combined to carry out region growing, it is contemplated that the basic structure of image itself makes segmentation effect more
Accurately.
Detailed description of the invention
Fig. 1 is a kind of image partition method flow chart based on image data field.
Specific embodiment
The present invention is described further below in conjunction with Figure of description.
According to Figure of description (1), implementation steps are described in detail:
The input of step (1) image and image preprocessing.
In order to achieve the effect that comparison, acquire the effect picture with piece image under different illumination conditions, then respectively into
Row processing.Pretreatment to image mainly includes carrying out gray processing, filtering and noise reduction, histogram equalization, binaryzation etc. to image
Several treatment processes.
Step (2) carries out image segmentation with the watershed algorithm based on eight direction gradients.
Watershed algorithm is a kind of image partition method based on gradient, can convert " mountain for the edge in image
Homogeneous area is converted " mountain valley " by arteries and veins ", is broadly divided into sequence, floods two processes.
Step 1: seek image from all directions to gradient amplitude be maximized, gradient amplitude image is in the edge along object
There is higher pixel value, and have lower pixel value elsewhere, watershed algorithm can generate watershed ridge at the edge
Line；Gradient calculates and angle differentiates, set point f (x, y) represents the gray value on gray level image at pixel (x, y), uses
Robinson gradient template and original image carry out convolution:
Change rate (namely f (x+Dx, the y+dy)-f of the directional derivative of a direction i.e. f (x, y) in the direction
The value of (x, y)), gradient direction i.e. the maximum direction of directional derivative when the direction is consistent with the direction of gradient, direction is led
Several values is equal to the mould of gradient.
Step 2: sorting, flooding process.It is assumed that domain FfOn function f be image to be processed, its maximin
Respectively fmax=max (f), fmin=min (f).Gather [Ff]hThe upper limit threshold of representative function f, wherein [Ff]h={ x ∈ Ff|f
(x)≤f}。
Reg_Mink(f)={ x ∈ Ff| x=Local minimum and f (x)=f }
The watershed region of F is to gatherWithIt is obtained by interative computation:
The watershed of function f isSupplementary set,Corresponding rise process, all dams have corresponded to function f's
Watershed leads to image over-segmentation since the image of input generates many small receiving basins there are excessive very small region
Seriously, so needing to carry out similar area merging to segmentation result.
Step (3) constructs image data fields, calculates the gesture value function of each pixel.
Using the gray scale difference value of image pixel and neighborhood territory pixel as the quality of data object, corresponding image data is generated
, by image grain P (p, ε)=q | q ∈ I ∧ | | p-q | |≤ε be denoted as neighborhood territory pixel set using p as center pixel.Common
Optional potential function includes pseudo gravity, quasi- electrostatic field etc., and Mathematical Morphology is respectively as follows:
ψx(y)=G*mX/(1+(||x-y||/σ)k)；
ψ x (y)=mX/(4πε0*(1+(||x-y||/σ)k))；
Wherein mX>=0 is data object quality, represents field source intensity；, G is constant acceleration of gravity | | x-y | | represent two
Distance between person, natural number k are range indexs, and σ ∈ (0 ,+∞) is known as impact factor.
Gradient according to potential function is the field strength function in the corresponding field of force, the field intensity vector of data particle x are as follows:
Wherein | | py(t)-px(t) | | indicate the distance between x and y, quality my(t) regard the action intensity of field source as, σ >
0 is impact factor, determines the coverage of sometime particle interaction force.
Step (4) edge detection, calculates the average gray value of marginal point and is set as seed point value.
The most general method of edge detection is to detect the discontinuity of brightness value, and this discontinuity can be by the one of image
Rank and second dervative detection；Here edge detection is executed using Canny operator, progress gaussian filtering first eliminates noise, then
Calculate the gradient at pixelWith direction α (x, the y)=arctan (G at edgey/Gx)；Then
The grey level histogram for drawing marginal point pixel takes its average value labeled as seed point value.
The data fields and seed point that step (5) combination step (3), (4) obtain carry out automated seed region growing.
Region growing is a kind of image partition method of serial region segmentation.It is from some pixel, according to certain
Criterion is gradually added neighborhood pixels, and when certain conditions are met, region growing terminates.The quality of region growing is decided by 1.
The selection of initial point (seed point)；2. growing criterion；3. termination condition.The seed point pixel value that step (4) is obtained is as life
Long initial point using the gesture value function of step (3) as growth criterion, and chooses suitable growing threshold and carries out region growing.Tool
Steps are as follows for body:
One, the average gray value for obtaining step (4) is as growth initial value, and finding the 1st, there are no the initial of ownership
It is worth point, if the pixel is (x0, y0).
Two, centered on (x0, y0), 4 neighborhood territory pixels (x, y) of (x0, y0) are considered if (x0, y0) meets growth criterion
X (x-1, y)+X (x+1, y)+X (x, y-1)+X (x, y+1) > V, V=255,
X (x, y) represents pixel value when marking without seed sequence of point (x0, y0), is 0 if having.Then by (x, y) with
(x0, y0) merges (in the same area), while (x, y) is pressed into storehouse.
Three, a pixel is taken out from storehouse, it is returned to step 2 as (x0, y0).
Four, step 1 is returned to when storehouse is empty.
Five, step 1 is repeated to four when each point in image has ownership, and growth terminates.
Step (6) homogeneous region fusion output segmentation effect figure.
Region merging technique, step (5) region growing obtain many zonule P later, it is assumed that its mass center is (x0,y0):
X (x, y) is the gray value of the point in formula, if p*∈ P is the center pixel of any one subregion after growth, ψε
(p*)≤α indicates its gesture value, and α is growing threshold, forIf meeting condition (a) q ∈ ξ (p1*,
ε)∧ψq(p1*)≤β；(b) work as k > 1, Then q
And p*It constitutes homogeneity to attract, wherein ε is the impact factor of gesture value function, and β is predefined gesture value threshold value.
Claims (1)
1. a kind of image partition method based on image data field, it is characterised in that this method comprises the concrete steps that:
The input of step (1) image and image preprocessing；
Step (2) carries out image segmentation with the watershed algorithm based on eight direction gradients；
Step (3) constructs image data fields, calculates the gesture value function of each pixel；
Step (4) edge detection, calculates the average gray value of marginal point and is set as seed point value；
Step (5) combination step (3), the data fields of (4) and seed point carry out automated seed region growing；
Step (6) homogeneous region fusion output segmentation effect figure；
Step (2) is specifically:
Step 1: seek image from all directions to gradient amplitude be maximized；Gradient calculates and angle differentiates, set point f (x, y) generation
Gray value on table gray level image at pixel (x, y) carries out convolution with Robinson gradient template and original image:
The directional derivative of a direction i.e. f (x, y) the direction change rate, when the direction is consistent with the direction of gradient
When, the value of directional derivative is equal to the mould of gradient；
Step 2: sorting, flooding process；It is assumed that domain FfOn function f be image to be processed, it maximin difference
For fmax=max (f), fmin=min (f)；Gather [Ff]hThe upper limit threshold of representative function f, wherein [Ff]h={ x ∈ Ff∣f(x)≤
f}；
Reg_Mink(f)={ x ∈ Ff∣ x=Local minimum and f (x)=f }
The watershed region of F is to gatherWithIt is obtained by interative computation:
The watershed of function f isSupplementary set,Corresponding rise process, all dams have corresponded to the watershed of function f；
Step (5) is specifically:
The first step, the average gray value for obtaining step (4) are as growth initial value, and finding the 1st, there are no the initial of ownership
It is worth point, if the pixel is (x0, y0)；
Second step, centered on (x0, y0), consider 4 neighborhood territory pixels (x, y) of (x0, y0), if (x0, y0) to meet growth quasi-
Then, the growth criterion is using the gesture value function in step (3)；
X (x-1, y)+X (x+1, y)+X (x, y-1)+X (x, y+1) > V, V=255,
X (x, y) represents pixel value when marking without seed sequence of point (x0, y0), is 0 if having；Then by (x, y) with (x0,
Y0) merge in the same area, while (x, y) is pressed into storehouse；
Third step takes out a pixel from storehouse, it is returned to second step as (x0, y0)；
4th step returns to the first step when storehouse is empty；
5th step repeats first to fourth step when each point in image has ownership, and growth terminates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610335491.4A CN105976385B (en) | 2016-05-19 | 2016-05-19 | A kind of image partition method based on image data field |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610335491.4A CN105976385B (en) | 2016-05-19 | 2016-05-19 | A kind of image partition method based on image data field |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105976385A CN105976385A (en) | 2016-09-28 |
| CN105976385B true CN105976385B (en) | 2019-05-10 |
Family
ID=56957172
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610335491.4A Active CN105976385B (en) | 2016-05-19 | 2016-05-19 | A kind of image partition method based on image data field |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105976385B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107481241A (en) * | 2017-08-24 | 2017-12-15 | 太仓安顺财务服务有限公司 | A kind of color image segmentation method based on mixed method |
| CN108810423B (en) * | 2018-06-20 | 2020-07-31 | 北京优尔博特创新科技有限公司 | Illumination angle adjusting method and system based on image brightness gradient |
| CN109034058B (en) * | 2018-07-25 | 2022-01-04 | 哈工大机器人（合肥）国际创新研究院 | Method and system for dividing and self-correcting region in image |
| CN111562260B (en) * | 2020-04-14 | 2022-10-28 | 江苏大学 | Lotus root mud hole detection method and device based on machine vision |
| CN115187592B (en) * | 2022-09-08 | 2022-12-09 | 山东奥洛瑞医疗科技有限公司 | Image segmentation method suitable for nuclear magnetic resonance image |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101727656A (en) * | 2008-10-31 | 2010-06-09 | 李德毅 | Image segmenting method based on data field |
| US20150287210A1 (en) * | 2014-04-03 | 2015-10-08 | Sony Corporation | Image processing system with automatic segmentation and method of operation thereof |
-
2016
- 2016-05-19 CN CN201610335491.4A patent/CN105976385B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101727656A (en) * | 2008-10-31 | 2010-06-09 | 李德毅 | Image segmenting method based on data field |
| US20150287210A1 (en) * | 2014-04-03 | 2015-10-08 | Sony Corporation | Image processing system with automatic segmentation and method of operation thereof |
Non-Patent Citations (2)
| Title |
|---|
| 《基于数据场的图像分割方法初探》;陈露等;《地理空间信息》;20071231;第5卷(第6期);79-81 * |
| 《基于标记的分水岭图像分割算法研究》;陈洁等;《软件》;20121231;第33卷(第3期);115-117 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105976385A (en) | 2016-09-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105976385B (en) | A kind of image partition method based on image data field | |
| Anjna et al. | Review of image segmentation technique | |
| CN105354865B (en) | The automatic cloud detection method of optic of multispectral remote sensing satellite image and system | |
| CN102024259B (en) | Bacterial colony automatic detection method | |
| CN100580694C (en) | Rapid multi-threshold value dividing method for gray-scale image | |
| CN111127360B (en) | Gray image transfer learning method based on automatic encoder | |
| CN109684941B (en) | Litchi fruit picking area division method based on MATLAB image processing | |
| Feng et al. | A separating method of adjacent apples based on machine vision and chain code information | |
| CN104408445A (en) | Automatic real-time human body detecting method | |
| CN102592290A (en) | Method for detecting moving target region aiming at underwater microscopic video | |
| Chen et al. | A new watershed algorithm for cellular image segmentation based on mathematical morphology | |
| CN110874825A (en) | Method for extracting binary image of water stain on surface of composite insulator | |
| CN103177244A (en) | Method for quickly detecting target organisms in underwater microscopic images | |
| CN116524269A (en) | Visual recognition detection system | |
| CN110942467A (en) | Improved watershed image segmentation method based on PSO-FCM | |
| CN104156956B (en) | A kind of multicorner edge detection operator method recognized based on Gauss wavelet one-dimensional peak value | |
| Behera et al. | Image enhancement using accelerated particle swarm optimization | |
| CN110532892B (en) | Method for detecting road vanishing point of single image of unstructured road | |
| Zhang et al. | An improved edge detection algorithm based on fuzzy theory | |
| CN114092485A (en) | Mask rcnn-based stacked coarse aggregate image segmentation method and system | |
| CN104182971B (en) | A kind of high precision image square localization method | |
| CN106815845A (en) | Color image segmentation method based on pixels probability density classification | |
| CN103632378A (en) | Multi-threshold edge detection method based on point connection drawing game idea | |
| CN106971394B (en) | A kind of image partition method of quick separating adhesion corn seed | |
| Gu et al. | Drms: Dim-light robust monocular simultaneous localization and mapping |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |