CN101334896A - Processing method for measuring sub-pixel rim of digital picture - Google Patents
Processing method for measuring sub-pixel rim of digital picture Download PDFInfo
- Publication number
- CN101334896A CN101334896A CNA200810150465XA CN200810150465A CN101334896A CN 101334896 A CN101334896 A CN 101334896A CN A200810150465X A CNA200810150465X A CN A200810150465XA CN 200810150465 A CN200810150465 A CN 200810150465A CN 101334896 A CN101334896 A CN 101334896A
- Authority
- CN
- China
- Prior art keywords
- edge
- gradient
- image
- point
- normal
- 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.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a digital image measurement sub-pixel edge processing method which carries out a bulk processing on the image edge by adopting a gradient model, takes a gradient mean as the threshold of the binaryzation control, realizes a search of binaryzation images in the biggest connected region on the basis of a flood filling principle and searches consecutive points of which a segment of gradient values exceeds the gradient threshold at all location points in four direction; the direction which has the smallest number of consecutive points is the direction of normal; a gradient peak value along the direction of normal is calculated on base of a spline interpolation subdivision algorithm; edge points contained on a fitting curve is marked from the collected target rough edges; after multiple rounds of iterative computations, if all points on the collected target rough edges are marked, the precise sub-pixel edge computation is finished. The processing method of the invention bases on traditional gradient edge extraction, flood filling algorithm and spline interpolation subdivision algorithm; the processing course is realized through a crude image processing and a refined image processing; the target edge is searched on purpose; the computational efficiency is high and the algorithm course takes the requirement of the noise suppression into consideration and increases the immunity of the algorithm.
Description
Technical field
The present invention relates to a kind of image measurement disposal route, be specifically related to a kind of processing method for measuring sub-pixel rim of digital picture.
Background technology
Along with the development of current digitizing infotech, it is more and more important that digital image processing techniques are just becoming, and all be used widely at a lot of productive life scenes.The current digital image measuring technique has also occupied important status in the Photoelectric Detection field, and its most important content is exactly the accurate location of realizing the image object edge.This measures in the application particularly important in digital picture, we can say that accurate rapid extraction image border is extract photographed image-related information basic and crucial in the digital picture measurement is used.
Present digital picture all is a two-dimensional image.According to the difference of CCD sensitive chip, target image is by the storage of being dispersed with different segmentation density, and the minimum dimension unit on the image is a pixel, has stored the image color information of each location of pixels on the file layout.Digital Image Processing is exactly to adopt certain calculation method that these memory contentss are carried out correlation computations to extract the process of relevant information of interest in the image.
At present adopt basic processing algorithm mainly to be based on the computing method of gray scale or gradient principle to Flame Image Process.They all are certain eigenwerts (gray scale or gradient) by computed image, according to given eigenwert control threshold values figure are carried out binary conversion treatment then.The main thought of its institute's foundation is exactly that the edge and the background of target object in the image has tangible visual impact, and is promptly visually very big as content change at the edge of the object location drawing.The main computation model of weighing this picture material variation at present is gray variance or gradient.If on the image one with point (i j) is 3 * 3 image-regions at center.
i-1 j+1 | i j+1 | i+1 j+1 |
i-1 j | i j | i+1 j |
i-1 j-1 | i-1 j-1 | i+1 j-1 |
The gray variance computation model of every bit is
I, j, m, n are nonnegative integer, (i, j), (i, j), (i is that ((i j) is point (i, gray variance j) to V to point for i, RGB j) (Red Green Blue) component value j) to b to g to r.
The gradient calculation model is
Wherein
I, j, m, n are nonnegative integer, r (m, n), g (m, n), (m is that ((i j) is point (i, gradient j) to G to point for m, RGB n) (Red Green Blue) component value n) to b, (i, j) (i, j=0,1,2) is Sobel operator correspondence (i, j) element value of position to Sobel.
Traditional image processing algorithm all is based on this two kinds of computation models at present.Such as the Moravec operator is exactly a kind of operator that utilizes the gray variance extract minutiae, also has Roberts, Sobel, Laplace constant gradient operator in addition.The gradient calculation model can also reach different Flame Image Process purposes by changing the coherent element value of calculating in the operator.Also has a kind of Hough mapping algorithm from image space to the parameter space conversion that realize the image object extraction by in addition.Traditional algorithm generally also adopts auxiliary process algorithms such as noise filtering, level and smooth and sharpening before Flame Image Process with after handling.
These image processing algorithms have treatment effect preferably for the extraordinary object of target image quality, but in practical engineering application, especially also have many application difficult points in realtime graphic is handled:
One,, and is unfavorable for the realization of computing machine auto-programming because the binaryzation result is influenced by eigenwert control threshold values, and the processing result image artifical influence factor is too big, and algorithm does not possess versatility for different application occasion image;
Two, there is multiple noise in the processing image,, is unfavorable for the realtime graphic processing though traditional algorithm hope is very little for the different images effect by changing relational operator or auxiliary filtering algorithm to reach the purpose of squelch;
Three, to location, target image edge out of true, result is too rough, is difficult to satisfy precision profile and measures application demand, has the application scenario of precision profile processing requirements also to have significant limitation many simultaneously.
Summary of the invention
Purpose of the present invention provides a kind of efficient height, and has taken into account the requirement of squelch in the application demand of measuring at realtime graphic, has improved the processing method for measuring sub-pixel rim of digital picture of the immunity of algorithm.
For achieving the above object, the technical solution used in the present invention is:
1) at first adopts gradient former that the image edge is carried out bulk processing, control threshold values as binaryzation with gradient mean value;
2) realize the largest connected range searching of binary image based on flood filling principle;
By reference position is to all the winds spread the size that mode obtains whole connected region, search out largest connected zone, edge, and write down these edge data, and being taken as noise, other little connected region suppressed;
3) search comprises the continuity point of one section Grad of each location point above gradient control threshold values on four direction, the minimum direction of counting is direction of normal, the calculated direction of exact image edge location just, position, image edge has unimodality along the direction of normal Grad at edge, and the gradient peak position is exactly position, accurate edge;
4) based on the gradient peak on the spline interpolation algorithm of subdivision calculating edge direction of normal
Because object is in the Grad maximum of edge, then utilize spline-fitting to obtain a curve at edge direction of normal and the Grad that comprises the continuous image vegetarian refreshments of this marginal point to one section, on the curve that match obtains, carry out 10 times of segmentations, be that bearing accuracy reaches 0.1 pixel, relatively the gradient magnitude of each point position, segmentation back is got maximum value position and is accurate marginal point;
5) from the thick edge of collecting of target, the edge point that is comprised in the matched curve is carried out mark;
6),, show that then accurate sub-pixel rim calculating finishes if the thick edge point of collecting of target all is labeled through too much taking turns iterative computation.
The present invention is based on the extraction of traditional gradient edge, flood filling algorithm and spline interpolation algorithm of subdivision, and processing procedure is by thick, smart two steps Flame Image Process realization.Carrying out for the target edge has the purpose search, the counting yield height, and algorithmic procedure takes into account the requirement of squelch, improved the immunity of algorithm.
Description of drawings
Fig. 1 is a process flow diagram of the present invention;
Fig. 2 is the target edge extraction figure that the present invention is based on largest connected range searching, and wherein Fig. 2 (a) is an original image, and Fig. 2 (b) is gradient binary conversion treatment result's a image, and Fig. 2 (c) is based on the target edge figure in largest connected zone;
Fig. 3 was point (x
0, y
0) image edge direction of normal search graph, wherein Fig. 3 (a) is that the edge curve is cut and vowed and method arrow figure that Fig. 3 (b) is a target edge image, Fig. 3 (c) is that target edge image method is vowed the search synoptic diagram;
Fig. 4 is a SPL interpolation subdividing synoptic diagram, and wherein horizontal ordinate and ordinate are pixel value.
Embodiment
Below in conjunction with accompanying drawing the present invention is described in further detail.
Referring to Fig. 1, the present invention at first carries out first step acquisition of image data, collection result such as Fig. 2 (a).
First step institute images acquired is carried out the gradient operator universe search of second step.Adopt the gradient operator model to carry out the bulk processing of image edge, control threshold values as binaryzation, prevent that the target side information is lost among the binary conversion treatment result, also be beneficial to Automatic Program simultaneously and realize with gradient mean value;
Utilize the result of gradient operator universe search, the processing of images acquired being carried out the 3rd step is the area search of flood fill area.Owing to have many factor jamming target images in the realtime graphic, after handling by the search of gradient operator universe, there is multiple random noise in image.This noise like generally is difficult to prove effective by traditional filtering mode.According to the ultimate principle of Flame Image Process, in image measurement, generally once only be concerned about a certain concrete target edge, in the target area that only comprises the target edge, the target edge after the binary conversion treatment should have maximum image area.By the flood searching algorithm reference position is to all the winds spread the size that mode obtains whole connected region, search out largest connected zone, edge, and write down these edges data set P={pi, and 0=<i<=n n is a positive integer }, and being taken as noise, other little connected region suppressed.
The effect of squelch as can be seen from Fig. 2 (b) and (c) contrast: except that the target side information, all the other noises are suppressed fully in the result.Set different Flame Image Process zones, just can obtain institute interested different target edge.
Then carried out for the 4th step---choose pi, search pi place edge direction of normal.The image edge testing result that first three step obtains is a thick edge also, is generally 3 to 5 pixel wide, and this also can't satisfy measurement requirement in precision profile is measured, must accurately locate the image edge.As Fig. 3 (a), if from continuous angle, position (x is put at certain in the edge
0, y
0) calculating of the direction of normal located should at first calculate one section mistake (x
0, y
0) point curve y=f (x), then by differentiate, ask the mode of cutting arrow to realize point (x
0, y
0) indirect calculation of the direction of normal located.But this account form is too complicated in realtime graphic is handled, and calculated amount is too big.Here we provide following simplification computation model.
Owing to be that unit stores with the pixel in the real image,, thought point (x in therefore calculating as Fig. 3 (b) by discrete back
0, y
0) direction of normal have only four (it is positive and negative to be regardless of direction), shown in Fig. 3 (c).
Search comprises point (x on four direction
0, y
0) one section gradient surpass the continuity point of gradient control threshold values, the minimum direction of counting is promptly thought and is direction of normal.This is the calculated direction of location, exact image edge just.
Direction of normal search through edge, pi place obtained for the 5th step---and get direction of normal point set Pij, Pij represents the direction of normal search gained point set that corresponding pi is ordered, 1=<j<=n, and preserve this point set.
Here only be concerned about accurate image border data, so will carry out for the 6th step---from point set P, delete point (P ∩ Pij), guarantee that the image point set of handling is the exact image marginal date;
Utilize the 6th step result, image carried out for the 7th step handle---to the corresponding Grad of point set Pij is that functional value carries out poplar bar curve fitting and 10 times of interpolation subdividings of curve and gets new point set Pim, prepares for realizing that the image edge is accurately located.Because object is in the Grad maximum of edge, then to one section at the edge direction of normal and comprise point (x
0, y
0) the Grad of continuous image vegetarian refreshments (comprising the whole edges point on this direction) utilize spline-fitting to obtain a curve, on the curve that match obtains, carry out 10 times of segmentations, promptly bearing accuracy reaches 0.1 pixel, specifically as shown in Figure 4.Utilizing the 7th step result, next carried out for the 8th step---Grad maximum point pim among the search Pim, the position of corresponding point is the accurate sub-pixel edge on this edge direction of normal.According to Flame Image Process ultimate principle and gradient calculation principle, position, image edge has unimodality along the direction of normal Grad at edge, and the gradient peak position is exactly position, accurate edge.Relatively the gradient magnitude of each point position, segmentation back is got maximum value position and is accurate marginal point.
From the thick edge of collecting of target, mark the edge point that is comprised in the current matched curve in order to avoid carry out double counting.And carried out for the 9th step and judge---whether point set is searched for finishes, and continues to handle if be output as " N " and changeed for the 4th step then explanation also has left point.
Through too much taking turns iterative computation, if the 9th step was judged---whether point set is searched for finishes, if be output as " Y ", then the thick edge point of target of explanation collection all is labeled, and shows that then accurate sub-pixel rim calculates finish then last the tenth EOS program and output---and exporting the maximum point set Pm={pim} of whole Grad is accurate sub-pixel edge.
Claims (1)
1, processing method for measuring sub-pixel rim of digital picture is characterized in that:
1) at first adopts gradient former that the image edge is carried out bulk processing, control threshold values as binaryzation with gradient mean value;
2) realize the largest connected range searching of binary image based on flood filling principle;
By reference position is to all the winds spread the size that mode obtains whole connected region, search out largest connected zone, edge, and write down these edge data, and being taken as noise, other little connected region suppressed;
3) search comprises the continuity point of one section Grad of each location point above gradient control threshold values on four direction, the minimum direction of counting is direction of normal, the calculated direction of exact image edge location just, position, image edge has unimodality along the direction of normal Grad at edge, and the gradient peak position is exactly position, accurate edge;
4) based on the gradient peak on the spline interpolation algorithm of subdivision calculating edge direction of normal
Because object is in the Grad maximum of edge, then utilize spline-fitting to obtain a curve at edge direction of normal and the Grad that comprises the continuous image vegetarian refreshments of this marginal point to one section, on the curve that match obtains, carry out 10 times of segmentations, be that bearing accuracy reaches 0.1 pixel, relatively the gradient magnitude of each point position, segmentation back is got maximum value position and is accurate marginal point;
5) from the thick edge of collecting of target, the edge point that is comprised in the matched curve is carried out mark;
6),, show that then accurate sub-pixel rim calculating finishes if the thick edge point of collecting of target all is labeled through too much taking turns iterative computation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200810150465XA CN101334896B (en) | 2008-07-25 | 2008-07-25 | Processing method for measuring sub-pixel rim of digital picture |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200810150465XA CN101334896B (en) | 2008-07-25 | 2008-07-25 | Processing method for measuring sub-pixel rim of digital picture |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101334896A true CN101334896A (en) | 2008-12-31 |
CN101334896B CN101334896B (en) | 2010-08-11 |
Family
ID=40197479
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200810150465XA Expired - Fee Related CN101334896B (en) | 2008-07-25 | 2008-07-25 | Processing method for measuring sub-pixel rim of digital picture |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101334896B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101833748A (en) * | 2010-03-29 | 2010-09-15 | 中山大学 | Subdivision method-based image amplifying method |
CN101841641A (en) * | 2010-03-29 | 2010-09-22 | 中山大学 | Video amplification method and system based on subdivision method |
CN102104779A (en) * | 2011-03-11 | 2011-06-22 | 深圳市融创天下科技发展有限公司 | 1/4 sub-pixel interpolation method and device |
CN108648205A (en) * | 2018-05-07 | 2018-10-12 | 广州大学 | A kind of sub-pixel edge detection method |
CN108989551A (en) * | 2018-06-27 | 2018-12-11 | Oppo广东移动通信有限公司 | Position indicating method, device, storage medium and electronic equipment |
CN109003258A (en) * | 2018-06-15 | 2018-12-14 | 广东工业大学 | A kind of high-precision sub-pix circular pieces measurement method |
CN109269416A (en) * | 2017-07-17 | 2019-01-25 | 成都唐源电气股份有限公司 | A kind of contact line conducting wire measurement of wear method and device |
CN109269415A (en) * | 2017-07-17 | 2019-01-25 | 成都唐源电气股份有限公司 | A kind of contact line conducting wire measurement of wear method and device based on multichannel camera |
CN109269403A (en) * | 2017-07-17 | 2019-01-25 | 成都唐源电气股份有限公司 | A kind of contact line conducting wire measurement of wear method and device based on machine vision |
CN112700431A (en) * | 2021-01-11 | 2021-04-23 | 武汉大学 | Flood coverage surface extraction method based on remote sensing image |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101137003B (en) * | 2007-10-15 | 2010-06-23 | 北京航空航天大学 | Gray associated analysis based sub-pixel fringe extracting method |
-
2008
- 2008-07-25 CN CN200810150465XA patent/CN101334896B/en not_active Expired - Fee Related
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101833748A (en) * | 2010-03-29 | 2010-09-15 | 中山大学 | Subdivision method-based image amplifying method |
CN101841641A (en) * | 2010-03-29 | 2010-09-22 | 中山大学 | Video amplification method and system based on subdivision method |
CN101833748B (en) * | 2010-03-29 | 2013-01-02 | 中山大学 | Subdivision method-based image amplifying method |
CN101841641B (en) * | 2010-03-29 | 2013-01-02 | 中山大学 | Video amplification method and system based on subdivision method |
CN102104779A (en) * | 2011-03-11 | 2011-06-22 | 深圳市融创天下科技发展有限公司 | 1/4 sub-pixel interpolation method and device |
WO2012122729A1 (en) * | 2011-03-11 | 2012-09-20 | 深圳市融创天下科技股份有限公司 | A 1/4 sub-pixel interpolation method and device |
CN102104779B (en) * | 2011-03-11 | 2015-10-28 | 深圳市云宙多媒体技术有限公司 | A kind of 1/4 sub-pixel interpolation method and device |
CN109269403A (en) * | 2017-07-17 | 2019-01-25 | 成都唐源电气股份有限公司 | A kind of contact line conducting wire measurement of wear method and device based on machine vision |
CN109269416A (en) * | 2017-07-17 | 2019-01-25 | 成都唐源电气股份有限公司 | A kind of contact line conducting wire measurement of wear method and device |
CN109269415A (en) * | 2017-07-17 | 2019-01-25 | 成都唐源电气股份有限公司 | A kind of contact line conducting wire measurement of wear method and device based on multichannel camera |
CN109269403B (en) * | 2017-07-17 | 2020-02-04 | 成都唐源电气股份有限公司 | Contact net lead abrasion measuring method and device based on machine vision |
CN109269415B (en) * | 2017-07-17 | 2020-02-18 | 成都唐源电气股份有限公司 | Contact net lead abrasion measuring method and device based on multi-path camera |
CN108648205A (en) * | 2018-05-07 | 2018-10-12 | 广州大学 | A kind of sub-pixel edge detection method |
CN109003258A (en) * | 2018-06-15 | 2018-12-14 | 广东工业大学 | A kind of high-precision sub-pix circular pieces measurement method |
CN109003258B (en) * | 2018-06-15 | 2021-10-19 | 广东工业大学 | High-precision sub-pixel circular part measuring method |
CN108989551A (en) * | 2018-06-27 | 2018-12-11 | Oppo广东移动通信有限公司 | Position indicating method, device, storage medium and electronic equipment |
CN108989551B (en) * | 2018-06-27 | 2020-12-01 | Oppo广东移动通信有限公司 | Position prompting method and device, storage medium and electronic equipment |
CN112700431A (en) * | 2021-01-11 | 2021-04-23 | 武汉大学 | Flood coverage surface extraction method based on remote sensing image |
CN112700431B (en) * | 2021-01-11 | 2022-08-05 | 武汉大学 | Flood coverage surface extraction method based on remote sensing image |
Also Published As
Publication number | Publication date |
---|---|
CN101334896B (en) | 2010-08-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101334896B (en) | Processing method for measuring sub-pixel rim of digital picture | |
CN102654902B (en) | Contour vector feature-based embedded real-time image matching method | |
US10521694B2 (en) | 3D building extraction apparatus, method and system | |
CN108197583B (en) | Building change detection method based on graph cut optimization and image structure characteristics | |
CN107292310B (en) | Visual positioning and automatic reading method for circular pointer type dial plate | |
CN107092871B (en) | Remote sensing image building detection method based on multiple dimensioned multiple features fusion | |
CN109060836A (en) | High-pressure oil pipe joint external screw thread detection method based on machine vision | |
CN109035276A (en) | A kind of image edge extraction method, device and automated driving system | |
CN111027446B (en) | Coastline automatic extraction method of high-resolution image | |
CN106529593B (en) | Pavement disease detection method and system | |
CN107358161B (en) | Coastline extraction method and coastline extraction system based on remote sensing image classification | |
CN111931696A (en) | Lake and reservoir water area remote sensing automatic extraction method based on space constraint | |
CN105809673A (en) | SURF (Speeded-Up Robust Features) algorithm and maximal similarity region merging based video foreground segmentation method | |
CN104866852B (en) | Extract the method and device of windy and sandy soil information in remote sensing image | |
CN106940782A (en) | High score SAR based on variogram increases construction land newly and extracts software | |
Singh et al. | A hybrid approach for information extraction from high resolution satellite imagery | |
CN105205801A (en) | Method and device for extracting sea reclamation information based on change detection | |
CN113240735B (en) | Slope displacement activity monitoring method | |
CN111882573B (en) | Cultivated land block extraction method and system based on high-resolution image data | |
Dahiya et al. | Object oriented approach for building extraction from high resolution satellite images | |
Ekhtari et al. | Automatic building extraction from LIDAR digital elevation models and WorldView imagery | |
CN104392209A (en) | Evaluation model for image complexity of target and background | |
CN104123569A (en) | Video person number information statistics method based on supervised learning | |
CN115984687A (en) | River work moving bed model test water boundary measuring method, device, equipment and medium | |
Taherzadeh et al. | A comparison of hyperspectral data and WorldView-2 images to detect impervious surface |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100811 Termination date: 20160725 |