CN107909610A - A kind of gray scale target perimeter evaluation method based on image grain and sub-pix border detection - Google Patents
A kind of gray scale target perimeter evaluation method based on image grain and sub-pix border detection Download PDFInfo
- Publication number
- CN107909610A CN107909610A CN201711045758.7A CN201711045758A CN107909610A CN 107909610 A CN107909610 A CN 107909610A CN 201711045758 A CN201711045758 A CN 201711045758A CN 107909610 A CN107909610 A CN 107909610A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- boundary
- gray scale
- sub
- 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
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- 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
- G06T7/13—Edge detection
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of gray scale target perimeter evaluation method based on image grain and sub-pix border detection, mainly include:Input picture, chooses appropriate prospect threshold value and background threshold;Image is normalized, then asks for object boundary width parameter;Granular processing is carried out to image, and carries out gray scale morphology dividing processing;Then object boundary is extracted, and accurate boundary position is calculated with Zernike squares operator;The accurate girth result of destination object is asked for finally by B-spline curves interpolation fitting.The beneficial effects of the invention are as follows:A kind of girth evaluation method detected based on image grain and sub-pix is proposed, this method make use of image grain to simplify the boundary information of destination object in image, be obscured suitable for image, the situation of high resolution;Profit can obtain the estimation result of more accurate image object object girth in this way, have robustness to the image estimation process obtained under different situations.
Description
1. technical field
The present invention relates to image partition method, image procossing, pattern-recognition, artificial intelligence field, and in particular to Yi Zhongji
In image grain and the gray scale target perimeter evaluation method of sub-pix border detection.
Background technology
The problem of girth estimation of destination object is one of image characteristics extraction very challenging in image.Reality is raw
In work, people can be easy to calculate the girth of real world object, but be it is difficult to object Zhou Changjin from the aspect of mathematical image
The accurate estimation of row.In addition, the different loss that may cause important edges and intensity profile of the image quality of digital picture are not
Uniformly, so that the estimation of boundary length is difficult to.
Usually used target perimeter evaluation method includes:Believe based on target object geometric characteristic, based on gray level
Breath and the method based on curve interpolation fitting.Wherein the method based on geometric characteristic is based primarily upon bianry image, has
Representational method has digitlization straightway method and minimum polygon method, and this kind of method characterizes target by finding out in digital picture
The boundary point of the geometric properties of object estimates its girth.And the girth evaluation method based on gray-scale information is based primarily upon
Original image is converted into pixel and covered by the pixel covering digitized image that Sladoje and Lindblad et al. are proposed, this method
Digitized image, obtains the half-tone information of the neighborhood of boundary pixel, replaces curve with digitlization straight line to calculate the length on border
Degree.Later, it is a kind of to be suggested with cubic spline interpolation boundary method, improve the precision of calculating.Based on curve interpolation
Method is all different from both approaches, and this kind of method needs first to extract object boundary, is then intended with corresponding curve
Border is closed, girth is estimated by the length of digital simulation boundary curve, representative method has Suhadolnik et al.
The polynomial interpolation evaluation method that evaluation method based on B-spline curves and Sameh for proposing et al. propose.First kind method needs
Image is converted into bianry image, conversion process can give up a large amount of useful boundary informations;Second class method is needed to image
Pixel covering segmentation is carried out, image type is limited very high;Three classes method only considered the situation of extraction ideal boundary, and not have
The problem of can be potentially encountered in view of Boundary Extraction process.
The present invention proposes a kind of evaluation method for being different from traditional girth evaluation method, be broadly divided into image granularityization and
Boundary Extraction and estimating stage.First, image to be detected is inputted, the threshold value of prospect threshold value and background is determined, by image normalization;
Then destination object border width is sought, image is carried out by granular according to border width, grayscale morphologic credit is carried out to new images
Cut, the boundary position of sub-pix is obtained using Zernike square operators;Finally with asking for final mesh using B-spline interpolation method
Mark girth.
The content of the invention
It is a primary object of the present invention to propose a kind of first to image granularity, simplified boundary information, and combine and be based on
The sub-pix boundary detection method of Zernike squares obtains accurate boundary position, and target is estimated by B-spline curves interpolation method
Girth, obtains more accurate girth estimation result.
The object of the invention to solve the technical problems is that the technical solution used is:One kind is based on image grain and sub-pix
The gray scale target perimeter evaluation method of border detection, including herein below:
1) image to be detected is inputted, determines the threshold value of prospect threshold value and background;
2) by image normalization;
3) destination object border width is estimated;
4) image is subjected to granular, gray scale morphology segmentation is carried out to new images;
5) boundary position of sub-pix is obtained using Zernike square operators;
6) final target perimeter is asked for using B-spline interpolation method.
Specifically, consider the image involved by practical application, all pictures be converted into gray level image, threshold value to meet with
Lower condition:Image intensity value is less than background parts energy complete representation background parts, and gray value, which is more than foreground part, completely to be included
Destination object to be measured, and background threshold is less than prospect threshold value.
The background threshold b and prospect threshold value f determined according to previous step, traversing graph is as all pixels, for each with (i, j)
Centered on pixel:If pij≤ b, then pijRepresent background pixel,If pij>=f, then pijRepresent destination object pixel,If b < pij< f, then pijRepresent object boundary pixel,Wherein,Represent image pair after normalizing
The pixel value of position is answered, normalization makes the background of image and target form sharp contrast
By boundary pixel as background pixel, the bianry image in foreground and background region is obtained, extracts the side of target at this time
Boundary is as inner boundary.Then boundary pixel is obtained the bianry image in foreground and background region, extracted as destination object pixel
The border of target is as outer boundary at this time.Each outer boundary pixel is traveled through, the inner boundary pixel minimum with its distance is found, obtains
Its Euclidean distance liAnd record, last border width w is defined as:
Wherein, n is the number of outer boundary pixel,Expression rounds up x, finally obtaining the result is that one is more than 0
Integer.
By in image per w × w potting gum into an image grain, the image grain after merging is converted into new pixel,
New pixel value just takes summation to take the strategy of average.When image boundary cannot be divided exactly by w, it is necessary to expand border until its length
It can be divided exactly by d, the pixel value newly supplemented takes the pixel value identical with background, and the unit length represented by each pixel is former at this time
W times come.Then gray scale morphology segmentation is carried out with 3 × 3 template to image.
To previous step processing after image carry out Boundary Extraction, the position on Pixel-level border with Zernike squares operator into
Row sub-pixel positioning, chooses 7 × 7 Zernike convolution masks, the distance of template center and border is limited in 1 unit picture
In plain width, being unsatisfactory for the boundary point of this condition will be rejected.
The sequence for the sub-pix boundary point that previous step is obtained utilizes B-spline curves as the control point of B-spline curves
Carry out interpolation fitting to it, circular is according to being defined as below:
Wherein, piRepresenting the position of the obtained boundary point of previous step, d is exponent number,For spline base function.Calculating
When, exponent number controls the smooth degree of matched curve, and usual d is taken as 3.
Brief description of the drawings
Fig. 1 is the evaluation method flow chart of the present invention;
Fig. 2 is the sub-pix border detection process schematic diagram of the present invention;
Fig. 3 is the spline interpolation procedural image partial schematic diagram of the present invention;
Embodiment
As shown in Figure 1, overall procedure of the present invention is as follows:First, input picture, chooses appropriate prospect threshold value and background threshold
Value;Secondly, image is normalized, then asks for object boundary width parameter;Then granular processing is carried out to image, and
Morphological segment processing is carried out, border width is limited in unit pixel width;Then object boundary is extracted, is used in combination
Zernike squares operator calculates accurate boundary position, and it is accurately all to ask for destination object finally by B-spline curves interpolation fitting
Long result.
The present invention comprises the following steps that:
Step 1:Input picture, determines prospect threshold value and background threshold
The type of image in practical application is considered, it is necessary to input picture be changed into gray level image, and choose prospect
Threshold value and background threshold.Selected threshold value will ensure that image intensity value only includes background, gray value less than background threshold part
Prospect is only included more than prospect threshold portion.
Step 2:By image normalization
According to upper definite background threshold b and prospect threshold value f, traversing graph is with (i, j) as all pixels, for each
The pixel of the heart:If pij≤ b, then pijRepresent background pixel,If pij>=f, then pijRepresent destination object pixel,If b < pij< f, then pijRepresent object boundary pixel,Wherein,Represent image pair after normalizing
The pixel value of position is answered, normalization makes the background of image and target form sharp contrast.
Step 3:Seek destination object border width
1) opposite side border width model is defined
To meeting that the region of certain condition is defined in image obtained in the previous step, the outer boundary and inner edge of target are defined
Boundary:By gray value in section (0,1), and its 4 neighborhood territory pixel is inner boundary there are the pixel definition that gray value is 1;By ash
Angle value is in section (0,1), and its 4 neighborhood territory pixel is inner boundary there are the pixel definition that gray value is 0.
2) destination object border width is estimated
Each outer boundary pixel is traveled through, the inner boundary pixel minimum with its distance is found, obtains its Euclidean distance liAnd remember
Record, last border width w according to the following formula gained, it is finally obtaining the result is that one be more than 0 integer.
Step 4:Image is subjected to granular, gray scale morphology segmentation is carried out to new images
1) image granularity
By in image per w × w potting gum into an image grain, the image grain after merging is converted into new pixel,
New pixel value just takes summation to take the strategy of average.When image boundary cannot be divided exactly by w, it is necessary to expand border until its length
It can be divided exactly by d, the pixel value newly supplemented takes the pixel value identical with background, and the unit length represented by each pixel is former at this time
W times come.
2) gray scale morphology is split
The method for the gray scale morphology segmentation that Sladoje and Lindblad is proposed can be tight by the object boundary in gray level image
Lattice are limited within a unit pixel width.Input, obtain using the image after granular as gray scale morphology partitioning algorithm
Object boundary is the wide image of unit pixel.
Step 5:The boundary position of sub-pix is obtained using Zernike square operators
Behind classical boundary tracking algorithm extraction object pixel level border, by method as shown in Figure 2, to each border picture
Plain position uses Zernike square operators, calculates the sub-pix point position of object boundary and records.The Zernike squares of utilization
Convolution mask uses 7 × 7 size, and the distance of template center and border is limited in 1 unit pixel width, and will be discontented
The boundary point of this condition of foot is given up, and the calculation of final sub-pix point is:
Wherein,For, newly to the line of object boundary point and the angle of horizontal direction, l is convolution in Zernike convolution masks
The distance of object boundary, A are newly arrived in templatenmFor the Zernike squares of n ranks m weights, A'nmFor rotationZernike squares after angle.
The absolute position of boundary point in the picture can be calculated according to parameter.
Step 6:Final target perimeter is asked for using B-spline interpolation method
The sequence for the sub-pix boundary point that previous step is obtained carries out B-spline as the control point of B-spline curves to it
Curve interpolation is fitted.Curve fitting process is calculated by following equation:
Wherein, piRepresenting the position of the obtained boundary point of previous step, d is exponent number,For spline base function.Calculating
When, exponent number controls the smooth degree of matched curve, and usual d is taken as 3.The girth of destination object is calculated by interpolation method, specifically
Computational methods define as the following formula:
Wherein, nk is need to supplement starting by the obtained interpolation point quantity of Cox-de Boor algorithms, final calculation result
With the wire length of ending boundary point.The local state of image is as shown in Figure 3 in Interpolation Process.
Claims (5)
- A kind of 1. gray scale target perimeter evaluation method based on image grain and sub-pix border detection, it is characterised in that including with Lower step:Step 1, input image to be detected, determine the threshold value of prospect threshold value and background;Step 2, by image normalization;Step 3, estimation destination object border width;Step 4, by image carry out granular, and gray scale morphology segmentation is carried out to new images;Step 5, the boundary position for obtaining using Zernike square operators sub-pix;Step 6, ask for final target perimeter using B-spline interpolation method.
- 2. the gray scale target perimeter evaluation method according to claim 1 based on image grain and sub-pix border detection, institute State in step 2, by image normalization, it is characterised in that:The background threshold b and prospect threshold value f that are determined according to step 1, traversing graph are with (i, j) as all pixels, for each The pixel of the heart:If pij≤ b, then pijRepresent background pixel,If pij>=f, then pijRepresent destination object pixel, If b < pij< f, then pijRepresent object boundary pixel,Wherein,Represent image correspondence position after normalizing Pixel value, normalization make image background and target formed sharp contrast.
- 3. the gray scale target perimeter evaluation method according to claim 1 based on image grain and sub-pix border detection, institute State in step 3, seek destination object border width, it is characterised in that:Boundary pixel is obtained the bianry image in foreground and background region, extract the side of target at this time as background pixel first Boundary is as inner boundary.Then boundary pixel is obtained the bianry image in foreground and background region, extracted as destination object pixel The border of target is as outer boundary at this time.Each outer boundary pixel is traveled through, the inner boundary pixel minimum with its distance is found, obtains Its Euclidean distance liAnd record, last border width w is defined as:Wherein, n is the number of outer boundary pixel,Expression rounds up x, finally obtaining the result is that one whole more than 0 Number.
- 4. the gray scale target perimeter evaluation method according to claim 1 based on image grain and sub-pix border detection, institute State in step 4, image be subjected to granular, gray scale morphology segmentation is carried out to new images, it is characterised in that:1) image granularity is carried outImage grain after merging, will be converted into new by the image of a given width per w × w potting gum into an image grain Pixel, new pixel value just take summation to take the strategy of average.When image boundary cannot be divided exactly by w, it is necessary to border expand until Its length can be divided exactly by d, and the pixel value newly supplemented takes the pixel value identical with background, at this time the unit length represented by each pixel Degree is original w times.2) gray scale morphology segmentation is carried out to image.
- 5. the gray scale target perimeter evaluation method according to claim 1 based on image grain and sub-pix border detection, institute State in step 5, the boundary position of sub-pix obtained using Zernike square operators, it is characterised in that:Boundary Extraction is carried out to the image after previous step processing, Asia is carried out with Zernike squares operator in the position on Pixel-level border Pixel positions, and chooses 7 × 7 Zernike convolution masks, it is wide that the distance of template center and border is limited in 1 unit pixel In degree, the boundary point for being unsatisfactory for this condition is not counted in border point set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711045758.7A CN107909610A (en) | 2017-10-31 | 2017-10-31 | A kind of gray scale target perimeter evaluation method based on image grain and sub-pix border detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711045758.7A CN107909610A (en) | 2017-10-31 | 2017-10-31 | A kind of gray scale target perimeter evaluation method based on image grain and sub-pix border detection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107909610A true CN107909610A (en) | 2018-04-13 |
Family
ID=61842216
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711045758.7A Pending CN107909610A (en) | 2017-10-31 | 2017-10-31 | A kind of gray scale target perimeter evaluation method based on image grain and sub-pix border detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107909610A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109872365A (en) * | 2019-02-20 | 2019-06-11 | 上海鼎盛汽车检测设备有限公司 | 3D four-wheel position finder destination disk image-recognizing method |
CN110517318A (en) * | 2019-08-28 | 2019-11-29 | 昆山国显光电有限公司 | Localization method and device, storage medium |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103323209A (en) * | 2013-07-02 | 2013-09-25 | 清华大学 | Structural modal parameter identification system based on binocular stereo vision |
-
2017
- 2017-10-31 CN CN201711045758.7A patent/CN107909610A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103323209A (en) * | 2013-07-02 | 2013-09-25 | 清华大学 | Structural modal parameter identification system based on binocular stereo vision |
Non-Patent Citations (5)
Title |
---|
ALOJZ SUHADOLNIK: "Digital Curve Length Calculation by Using B-spline", 《JOURNAL OF MATHEMATICAL IMAGING & VISION》 * |
JOAKIM LINDBLAD 等: "Coverage segmentation based on linear unmixing and minimization of perimeter and boundary thickness", 《PATTERN RECOGNITION LETTERS》 * |
吴秦 等: "灰度级信息的目标边界精确周长估算", 《中国图象图形学报》 * |
周琪 等: "基于自适应图像粒的目标对象边界周长估算", 《计算机工程》 * |
田春苗 等: "基于 Zernike矩的亚像素边缘检测算法", 《中国科技论文在线》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109872365A (en) * | 2019-02-20 | 2019-06-11 | 上海鼎盛汽车检测设备有限公司 | 3D four-wheel position finder destination disk image-recognizing method |
CN110517318A (en) * | 2019-08-28 | 2019-11-29 | 昆山国显光电有限公司 | Localization method and device, storage medium |
CN110517318B (en) * | 2019-08-28 | 2022-05-17 | 昆山国显光电有限公司 | Positioning method and device, and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111539273B (en) | Traffic video background modeling method and system | |
CN107644429B (en) | Video segmentation method based on strong target constraint video saliency | |
CN110232389B (en) | Stereoscopic vision navigation method based on invariance of green crop feature extraction | |
CN103886589B (en) | Object-oriented automated high-precision edge extracting method | |
CN109741356B (en) | Sub-pixel edge detection method and system | |
CN108961158B (en) | Image synthesis method and device | |
CN101976436B (en) | Pixel-level multi-focus image fusion method based on correction of differential image | |
CN103413120A (en) | Tracking method based on integral and partial recognition of object | |
CN102298773B (en) | Shape-adaptive non-local mean denoising method | |
CN106611416B (en) | Method and device for segmenting lung in medical image | |
CN102693426A (en) | Method for detecting image salient regions | |
CN106373128B (en) | Method and system for accurately positioning lips | |
CN109711268B (en) | Face image screening method and device | |
CN105869174B (en) | A kind of Sky Scene image partition method | |
CN106204617B (en) | Adapting to image binarization method based on residual image histogram cyclic shift | |
CN107578430A (en) | A kind of solid matching method based on adaptive weight and local entropy | |
CN107154044B (en) | Chinese food image segmentation method | |
CN115908371B (en) | Plant leaf disease and pest degree detection method based on optimized segmentation | |
CN107909610A (en) | A kind of gray scale target perimeter evaluation method based on image grain and sub-pix border detection | |
CN105913391A (en) | Defogging method based on shape variable morphological reconstruction | |
CN103985113B (en) | Tongue is as dividing method | |
CN108062762A (en) | A kind of method for tracking target based on Density Estimator | |
CN109993090B (en) | Iris center positioning method based on cascade regression forest and image gray scale features | |
CN117092647A (en) | Method and system for manufacturing regional satellite-borne optical and SAR image DOM | |
JP5080416B2 (en) | Image processing apparatus for detecting an image of a detection object from an input image |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180413 |
|
WD01 | Invention patent application deemed withdrawn after publication |