CN105279507B - A method of extraction delineation character outline - Google Patents
A method of extraction delineation character outline Download PDFInfo
- Publication number
- CN105279507B CN105279507B CN201510632941.1A CN201510632941A CN105279507B CN 105279507 B CN105279507 B CN 105279507B CN 201510632941 A CN201510632941 A CN 201510632941A CN 105279507 B CN105279507 B CN 105279507B
- Authority
- CN
- China
- Prior art keywords
- gray value
- image
- character
- multiphase
- profile
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of methods of extraction delineation character outline, including carry out the histogram analysis of optical character image, and the high gray value region of character and low gray value region are obtained using multiphase movable contour model and dual level sets method；Local histogram's analysis is carried out respectively to the high gray value region and low gray value region of acquisition, according to their intensity profile and its correspondence, greyscale transformation coefficient is determined, multiphase image is transformed to two-phase image；Using the movable contour model that can utilize local message, and using the profile of single level set method extraction two-phase image, the profile as delineation character.The character outline extracted is complete, accurate, and the feature of character can be easily extracted using its result.
Description
Technical field
The present invention relates to optics to delineate the contours extract of character more particularly to a kind of method of extraction delineation character outline.
Background technology
Delineation character is by hard alloy marking needle or to inlay the marking needle of carbonado and be directly carved into the material of metal works
The mark that material is internal and is formed, is usually used in mark and the recourse of automobile, motorcycle, engine and other parts.Optics is delineated
Character is then the delineation character picture obtained by image capture device under light source, the accurate profile for extracting optics delineation character
It is the necessary process of character recognition and character errors (stroke missing, fracture etc.) detection.
Currently, the method for extraction optical character profile has very much, the method for mainly using image segmentation is such as based on threshold
Value, the method etc. based on region, based on edge, based on graph theory, based on energy functional.But these methods are delineated for optics
Good effect cannot be all obtained when character.Especially if using strip source, on the image can with the stroke of source parallel
The pixel higher than background pixel gray value is generated, and the stroke vertical with light source will produce the picture lower than background pixel gray value
Element.Thus traditional character outline extracting method is either based on threshold value still all cannot accurately extract delineation based on edge etc.
The actual profile of character.And ideal profile can not be extracted using the method based on energy functional such as movable contour model.
Common geometric active contour model often uses Level Set Method to solve the minimum value of energy functional to obtain desired character
Profile.Partial contour can only be extracted using biphasic models and single level set, and multiphase model and dual level sets is used to extract
Two kinds of profiles, have and report to the leadship after accomplishing a task or be overlapped, and the entire profile that curve indicates a character cannot be taken with one, be subsequent processing
As character feature extraction brings inconvenience.
Invention content
The purpose of the present invention is exactly to solve the above-mentioned problems, to provide a kind of method of extraction delineation character outline, comprehensive
Delineation character is extracted with the solution of the methods of histogram analysis, greyscale transformation, movable contour model and level set
Profile, the character outline extracted is complete, accurate, and the feature of character can be easily extracted using its result.
To achieve the goals above, the present invention adopts the following technical scheme that：
A method of extraction delineation character outline includes the following steps：
Step 1 carries out the histogram analysis of optical character image, using multiphase movable contour model and dual level sets side
Method obtains the high gray value region of character and low gray value region；
Step 2 carries out local histogram's analysis respectively to the high gray value region and low gray value region of acquisition, according to
Their intensity profile and its correspondence, determines greyscale transformation coefficient, and multiphase image is transformed to two-phase image；
Step 3 extracts two-phase figure using the movable contour model that can utilize local message, and using single level set method
The profile of picture, the profile as delineation character.
Multiphase movable contour model uses multiphase CV (Chan-vese) model in the step 1.
Background area is without greyscale transformation when multiphase image being transformed to two-phase image in the step 2, using high ash
The linear transformation method that angle value is hinted obliquely to low gray value, the region for needing to convert using element marking law regulation.
The specific method that multiphase image is transformed to two-phase image is,
High gray value region ΩH, gray average and variance are respectively mHAnd δH；Low gray value region ΩL, gray scale is equal
Value and variance are respectively mLAnd δL；Background area is ΩB, gray average and variance are respectively mBAnd δB；Image u0A certain picture
Plain u0The gray value of (i, j) is h (i, j), and the gray value after transformation is h'(i, j), greyscale transformation process is as follows：
IfAnd h (i, j) ＞ mH-δH；
So h'(i, j)=- (δL/δH)*h(i,j)+mH*(δL/δH)-mL。
Movable contour model in the step 3 uses LCV (Local Chan-vese) model.
LCV models comprehensively utilize global and local information so that segmentation can not be influenced by gradation of image is non-uniform；LCV
The energy functional E of modelLCVIt is made of global keys, local entity and regularization term：ELCV=α EG+β·EL+ER, wherein EL、ERWith
EGIt is global keys, local entity and regularization term respectively, α and β are greater than 0 constant；
It is expressed as after introducing level set function：
R is real number, u'0It is two-phase character picture, C indicates the contour curve of smooth closure, c1And c2It is respectively
Gradation of image mean value inside and outside evolution curve C；φ (x, y) is level set function, and H (z) and δ (z) are Hai Shi respectively
(HeaViside) the regularization form of function H (z) and dirac (Dirac) function δ (z)；gkIt is window size being averaged for k
Convolution operator, d1And d2It is the difference of the image and source images after convolution respectively；
Level set movements equation is accordingly：
Non trivial solution is exactly the profile finally extracted.
Beneficial effects of the present invention：
The character outline that the present invention extracts is accurate, and completely, and segmentation result is indicated with a closed curve, is follow-up
Character feature extraction bring great convenience.
Description of the drawings
Fig. 1 (a) is original delineation character picture；The profile that edge extracting methods of the Fig. 1 (b) based on Canny operators obtains；
It is the delineation character outline that Fig. 1 (c) is extracted based on multiphase C-V models and multilevel collection；It is the word that the present invention extracts for Fig. 1 (d)
Accord with profile；
Fig. 2 is the histogram for being character " G ".
Specific implementation mode
The invention will be further described with embodiment below in conjunction with the accompanying drawings.
It is original delineation character picture, in order to extract the profile of character " G " as shown in Fig. 1 (a).The present invention includes：
Step 1 carries out the histogram analysis of optical character image, using multiphase movable contour model and dual level sets side
Method obtains the high gray value region of character and low gray value region；
Histogram analysis is by determining that multiphase movable contour model and dual level sets method, which may be used, obtains the height of character
Gray value and low gray value region.
In delineating character picture, a complete stroke may include two class pixel of low gray value and high gray value.Such as
It is the histogram of character " G " shown in Fig. 2.The intensity profile of pixel is analyzed it is found that background gray scale concentrates on 80-100, low gray scale
The gray scale of value pixel concentrates on 20-60, and the gray scale of high gray-value pixel concentrates on 200-250, background, high gray value region, low
The gray scale in gray value region is mutual " fitting ".High gray value area is obtained using multiphase movable contour model such as multiphase CV models
The edge in domain and gray value region.
Step 2 carries out local histogram's analysis respectively to the high gray value region and low gray value region of acquisition, according to
Their intensity profile and its correspondence, determines greyscale transformation coefficient, and multiphase image is transformed to two-phase image；
If high gray value region ΩH, gray average and variance are respectively mHAnd δH；Low gray value region ΩL, gray scale
Mean value and variance are respectively mLAnd δL, background area ΩB, gray average and variance are respectively mBAnd δB.Background area is not required to
Greyscale transformation, and the linear transformation method hinted obliquely to low gray value using high gray value are carried out, using element marking law regulation
Need the region Ω convertedH。
If image u0A certain pixel u0The gray value of (i, j) is h (i, j), and the gray value after transformation is h'(i, j), transformation
Image u ' afterwards0It indicates.Greyscale transformation process is as follows：
IfAnd h (i, j) ＞ mH-δH；
So h'(i, j)=- (δL/δH)*h(i,j)+mH*(δL/δH)-mL。
Step 3 extracts two-phase figure using the movable contour model that can utilize local message, and using single level set method
The profile of picture, the profile as delineation character.
LCV models comprehensively utilize global and local information so that segmentation can not be influenced by gradation of image is non-uniform.LCV
The energy functional of model is made of global keys, local entity and regularization term：ELCV=α EG+β·EL+ER,
EL、ERAnd EGIt is global keys, local entity and regularization term respectively, α and β are greater than 0 constant.
It is expressed as after introducing level set function：
u'0It is two-phase character picture, C indicates the contour curve of smooth closure, c1And c2It is evolution curve C respectively
Inside and outside gradation of image mean value；φ (x, y) is level set function, and H (z) and δ (z) are Hai Shi (HeaViside) respectively
The regularization form of function H (z) and dirac (Dirac) function δ (z)；gkIt is the average convolution operator that window size is k, d1With
d2It is the difference of the image and source images after convolution respectively.
Level set movements equation is accordingly：
Non trivial solution is exactly final segmentation result.
It is compared with traditional method (such as Canny operators) based on edge extracting.
It is the profile that the edge extracting method based on Canny operators obtains, thick line mark is not word as shown in Fig. 1 (b)
The real profile of symbol, this is because Canny operators are fundamentally based on the edge extracting method of gradient.In the centre of a stroke
Since the difference of direction of illumination produces two kinds of pixels of low gray-value pixel and high gray value, ash is produced in their intersection
Degree significantly changes, and exactly completely is captured by the non-maximum value suppressing method of the gradient of Canny operators, becomes Canny meanings
Edge in justice.
Compared with multiphase CV models.
It is the delineation character outline based on multiphase CV models and multilevel collection extraction, two kinds of gray areas as shown in Fig. 1 (c)
Domain obtains two kinds of profiles, is indicated respectively with solid line and dotted line, either division or juxtaposition between them, cannot table strictly according to the facts
Up to the profile of character, and result is expressed as subsequent processing with two curves and makes troubles (such as extraction character feature).
As shown in Fig. 1 (d), the character outline of the character outline that the present invention extracts, extraction is accurate, completely, and divides knot
Fruit is indicated with a closed curve, is brought great convenience for the extraction of subsequent character feature.For example, may be used to profile
Curve carries out wavelet transformation and obtains its fractal dimension, and base is established to carry out the character feature extraction detection based on Contour tracing
Plinth.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (2)
1. a kind of method of extraction delineation character outline, characterized in that include the following steps：
Step 1 is carried out the histogram analysis of optical character image, is obtained using multiphase movable contour model and dual level sets method
Take the high gray value region of character and low gray value region；
Step 2 carries out local histogram's analysis, according to them respectively to the high gray value region and low gray value region of acquisition
Intensity profile and its correspondence, determine greyscale transformation coefficient, multiphase image be transformed to two-phase image；
Step 3, using the movable contour model that can utilize local message, and using single level set method extraction two-phase image
Profile, the profile as delineation character；
Background area is without greyscale transformation when multiphase image being transformed to two-phase image in the step 2, using high gray value
The linear transformation method hinted obliquely to low gray value, the region for needing to convert using element marking law regulation；
The specific method that multiphase image is transformed to two-phase image is,
High gray value region ΩH, gray average and variance are respectively mHAnd δH；Low gray value region ΩL, gray average and side
Difference is respectively mLAnd δL；Background area is ΩB, gray average and variance are respectively mBAnd δB；Image u0A certain pixel u0(i,
J) gray value is h (i, j), and the gray value after transformation is h'(i, j), greyscale transformation process is as follows：
IfAnd h (i, j) ＞ mH-δH；
So h'(i, j)=- (δL/δH)*h(i,j)+mH*(δL/δH)-mL；
Movable contour model in the step 3 uses LCV models；
The energy functional E of LCV modelsLCVIt is made of global keys, local entity and regularization term：ELCV=α EG+β·EL+ER, wherein
EL、ERAnd EGIt is global keys, local entity and regularization term respectively, α and β are greater than 0 constant；
It is expressed as after introducing level set function：
u′0It is two-phase character picture, C indicates the contour curve of smooth closure, c1And c2It is inside evolution curve C respectively
With external gradation of image mean value；φ (x, y) is level set function, and H (z) and δ (z) are Hai Shi function H (z) and dirac respectively
The regularization form of function δ (z)；gkIt is the average convolution operator that window size is k, d1And d2Image after convolution respectively with
The difference of source images；
Level set movements equation is accordingly：
Non trivial solution is exactly the profile finally extracted.
2. a kind of method of extraction delineation character outline as described in claim 1, characterized in that multiphase activity in the step 1
Skeleton pattern uses multiphase CV models.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510632941.1A CN105279507B (en) | 2015-09-29 | 2015-09-29 | A method of extraction delineation character outline |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510632941.1A CN105279507B (en) | 2015-09-29 | 2015-09-29 | A method of extraction delineation character outline |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105279507A CN105279507A (en) | 2016-01-27 |
| CN105279507B true CN105279507B (en) | 2018-07-17 |
Family
ID=55148490
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510632941.1A Active CN105279507B (en) | 2015-09-29 | 2015-09-29 | A method of extraction delineation character outline |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105279507B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106023191B (en) * | 2016-05-16 | 2018-11-27 | 山东建筑大学 | A kind of optics delineation character edge extraction and edge fitting method based on structure feature |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101119429A (en) * | 2006-08-01 | 2008-02-06 | 北京北大方正电子有限公司 | Digital watermark embedded and extracting method and device |
| CN101398894A (en) * | 2008-06-17 | 2009-04-01 | 浙江师范大学 | Automobile license plate automatic recognition method and implementing device thereof |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7519209B2 (en) * | 2004-06-23 | 2009-04-14 | Vanderbilt University | System and methods of organ segmentation and applications of same |
| US7620226B2 (en) * | 2004-10-28 | 2009-11-17 | Siemens Medical Solutions Usa, Inc. | Semi-local active contour for a faint region detection in knee CT images |
-
2015
- 2015-09-29 CN CN201510632941.1A patent/CN105279507B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101119429A (en) * | 2006-08-01 | 2008-02-06 | 北京北大方正电子有限公司 | Digital watermark embedded and extracting method and device |
| CN101398894A (en) * | 2008-06-17 | 2009-04-01 | 浙江师范大学 | Automobile license plate automatic recognition method and implementing device thereof |
Non-Patent Citations (2)
| Title |
|---|
| "MULTI-PHASE ACTIVE CONTOUR MODEL FOR IMAGE SEGMENTATION BASED ON LEVEL SETS";Zheng Gang，Wang Huinan;《Transactions of Nanjing University of Aeronautics & Astronautics》;20060630;第23卷(第2期);第132页-136页 * |
| "基于MDMP-LSM算法的黄瓜叶片病斑分割方法";江海洋 等;《农业工程学报》;20121130;第28卷(第21期);第142页-146页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105279507A (en) | 2016-01-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105046196B (en) | Front truck information of vehicles structuring output method based on concatenated convolutional neutral net | |
| CN109859227B (en) | Method and device for detecting flip image, computer equipment and storage medium | |
| Fang et al. | The study on an application of otsu method in canny operator | |
| CN109242791B (en) | Batch repair method for damaged plant leaves | |
| CN102790841B (en) | Method of detecting and correcting digital images of books in the book spine area | |
| CN101673338B (en) | Fuzzy license plate identification method based on multi-angle projection | |
| CN104751142B (en) | A kind of natural scene Method for text detection based on stroke feature | |
| CN102629322B (en) | Character feature extraction method based on stroke shape of boundary point and application thereof | |
| CN106156684B (en) | A kind of two-dimensional code identification method and device | |
| CN110276279B (en) | Method for detecting arbitrary-shape scene text based on image segmentation | |
| CN105046252A (en) | Method for recognizing Renminbi (Chinese currency yuan) crown codes | |
| CN111259878A (en) | Method and equipment for detecting text | |
| CN110751154B (en) | Complex environment multi-shape text detection method based on pixel-level segmentation | |
| CN109961416B (en) | Business license information extraction method based on morphological gradient multi-scale fusion | |
| CN109886935A (en) | A kind of road face foreign matter detecting method based on deep learning | |
| Kanter | Color Crack: Identifying Cracks in Glass | |
| CN102609903B (en) | A kind of method of the movable contour model Iamge Segmentation based on marginal flow | |
| CN103914829B (en) | Method for detecting edge of noisy image | |
| CN102147867A (en) | Method for identifying traditional Chinese painting images and calligraphy images based on subject | |
| CN106504225A (en) | A kind of recognition methodss of regular polygon and device | |
| CN111754538A (en) | Threshold segmentation method for USB surface defect detection | |
| CN106778766B (en) | Positioning point-based rotating number identification method and system | |
| Getahun et al. | A robust lane marking extraction algorithm for self-driving vehicles | |
| CN107330434A (en) | Electrical symbol recognition methods in a kind of circuit diagram based on PHOG features | |
| CN109766850A (en) | Fingerprint image matching method based on Fusion Features |
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 |