CN106503732B - The classification method and categorizing system of text image and non-textual image - Google Patents
The classification method and categorizing system of text image and non-textual image Download PDFInfo
- Publication number
- CN106503732B CN106503732B CN201610892308.0A CN201610892308A CN106503732B CN 106503732 B CN106503732 B CN 106503732B CN 201610892308 A CN201610892308 A CN 201610892308A CN 106503732 B CN106503732 B CN 106503732B
- Authority
- CN
- China
- Prior art keywords
- text
- image
- line
- line segment
- connected domain
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 239000011159 matrix material Substances 0.000 claims abstract description 85
- 238000013145 classification model Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims description 20
- 238000010276 construction Methods 0.000 claims description 18
- 239000013598 vector Substances 0.000 claims description 16
- 239000000284 extract Substances 0.000 claims description 10
- 230000009466 transformation Effects 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 description 10
- 125000000205 L-threonino group Chemical group [H]OC(=O)[C@@]([H])(N([H])[*])[C@](C([H])([H])[H])([H])O[H] 0.000 description 5
- 230000008859 change Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000012141 concentrate Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 235000015170 shellfish Nutrition 0.000 description 1
- 230000026676 system process Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- 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
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the classification methods and categorizing system of a kind of text image and non-textual image.Wherein, this method may include the two values matrix for obtaining original image；The connected domain that character point is constituted in statistics two values matrix obtains position and the size of the character connected domain for meeting the first pre-provisioning request；Line of text is divided, the number and location information of the character connected domain that the number and location information and line of text for obtaining line of text contain；The line segment in original image is extracted, and calculates the tilt angle of line segment and the median of tilt angle；According to the median of the tilt angle of line segment and tilt angle, calculating does not meet ratio shared by the line segment of the second pre-provisioning request；Count the mean value and variance of channel S in the gray space and HSV space of original image；According to abovementioned steps as a result, realizing the classification of original image using Naive Bayes Classification Model.The embodiment of the present invention solves the technical issues of how text image accurately being filtered out from large nuber of images as a result,.
Description
Technical field
The present embodiments relate to technical field of image processing, and in particular to point of a kind of text image and non-textual image
Class method and categorizing system.
Background technique
Visual basis of the image as the human perception world is that the mankind obtain information, expressing information and the weight for transmitting information
Want means.Image procossing is analyzed image, to reach the technology of required result.
Image classification in image processing techniques refers to by being analyzed picture material and is divided automatically image
The technology of class.The technology suffers from important application value at many aspects, such as in internet security, video content analysis etc.
Aspect.Image classification is a ring important in image recognition technology, which can greatly improve life in mass data processing
Produce efficiency.
In view of this, the present invention is specifically proposed.
Summary of the invention
The main purpose of the embodiment of the present invention is to provide the classification method of a kind of text image and non-textual image, until
Partially solves the technical issues of how text image accurately being filtered out from large nuber of images.In addition, also providing one kind
The categorizing system of text image and non-textual image.
To achieve the goals above, according to an aspect of the invention, there is provided following technical scheme:
A kind of classification method of text image and non-textual image, the method include at least:
Obtain the two values matrix of original image；
The connected domain that character point is constituted in the two values matrix is counted, and filters out length, width, length-width ratio and meets
The character connected domain of one pre-provisioning request, and obtain the position of the character connected domain for meeting first pre-provisioning request and big
It is small；
Line of text is divided according to the position of the character connected domain for meeting the first pre-provisioning request and size, is obtained
The number for the character connected domain that the number and location information and the line of text of the line of text contain and position letter
Breath；
Hough transformation is carried out to the two values matrix, extracts the line segment in the original image, and calculate the line segment
The median of tilt angle and the tilt angle；
For all line segments extracted, according in the tilt angle and the tilt angle of the line segment
Digit, calculating do not meet ratio shared by the line segment of the second pre-provisioning request；
Count the mean value and variance of channel S in the gray space and HSV space of the original image；
The number of the character connected domain contained according to the number of the line of text and location information, each line of text and
Ratio shared by location information, the line segment for not meeting the second pre-provisioning request and the mean value and the variance establish Piao
Plain Bayesian Classification Model, and using the Naive Bayes Classification Model realize the text image of the original image with
The classification of the non-textual image.
Further, the two values matrix for obtaining original image specifically includes:
Obtain the gray matrix of the original image；
It is filtered to obtain the Pasteur by gray matrix of Pasteur's likeness coefficient filter to the original image
Similarity matrix；
Pasteur's similarity matrix is normalized；
Numerical value in Pasteur's similarity matrix after normalization is generated into histogram according to size；
On the histogram, two-value division is carried out using OTSU method, obtains two values matrix.
Further, the position of the character connected domain that the first pre-provisioning request is met according to and size are to line of text
It is divided, the character connected domain that the number and location information and the line of text for obtaining the line of text contain
Number and location information, specifically include:
Position and size based on the character connected domain for meeting the first pre-provisioning request carry out following judgement:
If there are overlapping regions for the ordinate of any two character connected domain, the two characters connected domain is divided into
Otherwise the two characters connected domain is divided into different line of text by one text row；
If any character connected domain and current all line of text non-overlapping region on the vertical scale, are described
Character connected domain creates a new line of text；
If there are overlapping regions for the ordinate of any character connected domain and any line of text, will be described any
Character connected domain is divided into this article current row；
If there are overlapping regions for the ordinate of any character connected domain and wantonly two line of text, will be described any
Character connected domain is divided into the big line of text of overlapping region proportion；
All character connected domains are traversed, above-mentioned judging result is based on, obtain the number and position letter of the line of text
The number and location information for the character connected domain that breath and the line of text contain.
Further, described that Hough transformation is carried out to the two values matrix, the line segment in the original image is extracted, and count
The tilt angle of the line segment and the median of the tilt angle are calculated, is specifically included:
Polar coordinate transform is carried out to the character point in the two values matrix, value in polar coordinate space is greater than to the word of threshold value
Point of the symbol point as line segment alternative in image space, and the alternative line segment is inverted in described image space；
Based on the alternative line segment being inverted in described image space, the start-stop point of all alternative line segments is counted
Position, length and tilt angle, and line segment is screened according to the length of the alternative line segment；
For the line segment filtered out, the median of the line segment tilt angle is calculated.
Further, the character that number and location information, each line of text according to the line of text contains
Ratio and the mean value and institute shared by the number and location information of connected domain, the line segment for not meeting the second pre-provisioning request
Variance is stated, establishes Naive Bayes Classification Model, and realize the original image using the Naive Bayes Classification Model
The classification of the text image and the non-textual image, specifically includes:
Building includes the image set of the text image and the non-textual image；
For the text image and the non-textual image that described image is concentrated, extracts be based on the line of text respectively
Number and the number of character connected domain that contains of location information, each line of text and location information, described do not meet second
The feature of ratio shared by the line segment of pre-provisioning request, the mean value and the variance, and construction feature vector；
Based on described eigenvector, the Naive Bayes Classification Model of 2 classifications is constructed；
According to the Naive Bayes Classification Model, the text image and the non-textual image are carried out to original image
Classification.
Further, described to be based on described eigenvector, the Naive Bayes Classification Model of 2 classifications is constructed, is specifically included:
Establish training sample；
The training sample is divided into text image class sample and non-textual image class sample；
Calculate the text image class sample proportion and the non-textual image class sample proportion；
According to the text image class sample proportion and the non-textual image class sample proportion, described in estimation
The class conditional probability distribution of each dimensional feature in feature vector；
According to the class conditional probability distribution of each dimensional feature, the class conditional probability distribution of each training sample is calculated；
Naive Bayes Classification Model is established according to the following formula:
Wherein,P (the xj|ωk) indicate the class item of each dimensional feature
Part probability distribution；P (the xi|ωk) indicate the class conditional probability distribution of each training sample；The k indicates classification；It is described
ωkIndicate k-th of classification；The j indicates dimension；The xiIndicate the training sample；The ω*Expression is inferred to described
Classification belonging to training sample.
To achieve the goals above, according to another aspect of the present invention, additionally provide a kind of text image with it is non-textual
The categorizing system of image, the system include at least:
First obtains module, for obtaining the two values matrix of original image；
Second obtain module, for counting the connected domain that character point is constituted in the two values matrix, and filter out length,
Width, length-width ratio meet the character connected domain of the first pre-provisioning request, and obtain the character for meeting first pre-provisioning request
The position of connected domain and size；
Division module, the position and size for meeting the character connected domain of the first pre-provisioning request according to are to line of text
It is divided, the character connected domain that the number and location information and the line of text for obtaining the line of text contain
Number and location information；
First computing module, for extracting the line segment in the original image to two values matrix progress Hough transformation,
And calculate the tilt angle of the line segment and the median of the tilt angle；
Second computing module, for being directed to all line segments extracted, according to the tilt angle of the line segment
And the median of the tilt angle, calculating do not meet ratio shared by the line segment of the second pre-provisioning request；
Statistical module, the mean value and variance of channel S in the gray space and HSV space for counting the original image；
Categorization module, the character contained for number and location information, each line of text according to the line of text
Ratio and the mean value and institute shared by the number and location information of connected domain, the line segment for not meeting the second pre-provisioning request
Variance is stated, establishes Naive Bayes Classification Model, and realize the original image using the Naive Bayes Classification Model
The classification of the text image and the non-textual image.
Further, the first acquisition module specifically includes:
Acquiring unit, for obtaining the gray matrix of the original image；
Filter unit, for being filtered by gray matrix of Pasteur's likeness coefficient filter to the original image
Obtain Pasteur's similarity matrix；
Normalization unit, for Pasteur's similarity matrix to be normalized；
Processing unit generates histogram according to size for the numerical value in Pasteur's similarity matrix after normalizing；
Two-value division unit, for carrying out two-value division using OTSU method, obtaining two-value square in the histogram
Battle array.
Further, first computing module specifically includes:
Converter unit will take for carrying out polar coordinate transform to the character point in the two values matrix in polar coordinate space
Value is greater than point of the character point of threshold value as line segment alternative in image space, and the alternative line segment is inverted to described image sky
Between in；
Screening unit, for counting all described standby based on the alternative line segment being inverted in described image space
Start-stop point position, length and the tilt angle of route selection section, and line segment is screened according to the length of the alternative line segment；
Computing unit, for calculating the median of the line segment tilt angle for the line segment filtered out.
Further, the categorization module specifically includes:
First construction unit, for constructing the image set including the text image and the non-textual image；
Second construction unit, the text image and the non-textual image for being concentrated for described image, respectively
Extract the number and position letter of the character connected domain that number and location information, each line of text based on the line of text contain
Ratio, the feature of the mean value and the variance shared by breath, the line segment for not meeting the second pre-provisioning request, and construction feature
Vector；
Third construction unit constructs the Naive Bayes Classification Model of 2 classifications for being based on described eigenvector；
Taxon, for according to the Naive Bayes Classification Model, to original image carry out the text image with
The classification of the non-textual image.
The embodiment of the present invention provides the classification method and categorizing system of a kind of text image and non-textual image.Wherein, should
Method may include the two values matrix for obtaining original image；The connected domain that character point is constituted in statistics two values matrix, and screen
Length, width, length-width ratio meet the character connected domain of the first pre-provisioning request out, and obtain the character company for meeting the first pre-provisioning request
The position in logical domain and size；Line of text is drawn according to the position for the character connected domain for meeting the first pre-provisioning request and size
Point, the number and location information of the character connected domain that the number and location information and line of text for obtaining line of text contain；To two
Value matrix carries out Hough transformation, extracts the line segment in original image, and calculate the tilt angle of line segment and the middle position of tilt angle
Number；For all line segments extracted, according to the median of the tilt angle of line segment and tilt angle, it is pre- that calculating does not meet second
Ratio shared by the line segment of provisioning request；Count the mean value and variance of channel S in the gray space and HSV space of original image；Root
The number and location information of the character connected domain contained according to the number and location information of line of text, each line of text do not meet
Ratio shared by the line segment of two pre-provisioning requests and mean value and variance establish Naive Bayes Classification Model, and utilize simple shellfish
This disaggregated model of leaf realizes the classification of the text image and non-textual image of original image.The embodiment of the present invention solves as a result,
The technical issues of how text image accurately being filtered out from large nuber of images, can complete in mass image data to text diagram
The screening operation of picture can also carry out the related works such as subsequent OCR identification to the text image filtered out again.
Other features and advantages of the present invention will be illustrated in the following description, also, partly becomes from specification
It obtains it is clear that understand through the implementation of the invention.Objectives and other advantages of the present invention can be by written explanation
Specifically noted method is achieved and obtained in book, claims and attached drawing.
Detailed description of the invention
Attached drawing is as a part of the invention, and for providing further understanding of the invention, of the invention is schematic
Examples and descriptions thereof are used to explain the present invention, but does not constitute an undue limitation on the present invention.Obviously, the accompanying drawings in the following description
Only some embodiments to those skilled in the art without creative efforts, can be with
Other accompanying drawings can also be obtained according to these attached drawings.In the accompanying drawings:
Fig. 1 is to be illustrated according to the process of the classification method of text image and non-textual image shown in an exemplary embodiment
Figure；
Fig. 2 is to be shown according to the structure of text image and the categorizing system of non-textual image shown in another exemplary embodiment
It is intended to.
Specific embodiment
With reference to the accompanying drawing and the present invention is described in detail in specific embodiment.Obviously, described embodiment
Only a part of the embodiment of the application is not whole embodiments.Based on the embodiment in the application, the common skill in this field
Without creative efforts, all other equivalent or obvious variant the embodiment obtained all falls within this to art personnel
In the protection scope of invention.The embodiment of the present invention can be come specific according to the multitude of different ways being defined and covered by claim
Change.In the following description, understand for convenience, give many details.However, it will be apparent that realization of the invention can be with
Without these details.It should also be noted that, in the absence of clear limitations or conflicts, it is each in the present invention
Embodiment and technical characteristic therein can be combined with each other and form technical solution.
It should be noted that, although can provide the example of the parameter comprising particular value herein, it is to be understood that parameter is without true
It cuts and is equal to corresponding value, but be similar to be worth accordingly in acceptable error margin or design constraint.
The basic thought of the embodiment of the present invention is: randomly selecting several width images from data source, is built into image set, connects
The image set being drawn into is subjected to manual sort, be divided into two class of text image and non-textual image, then extract respectively these two types of
The feature vector of other image, and the class conditional probability distribution in each feature space is counted, then, utilize Bayesian formula, root
According to the class conditional probability distribution in the prior probability and each feature space of classification, a unknown images can be calculated and belonged to
The probability of every one kind finally takes the biggish one kind of probability as the prediction to the image category.
In practical applications, in order to filter out text image from mass image data, the embodiment of the present invention proposes one kind
The classification method of text image and non-textual image.As shown in Figure 1, this method can be by step S100 to step S160 come real
It is existing.Wherein:
Step S100: the two values matrix of original image is obtained.
Specifically, this step can be realized further by step S101 to step S105.
Step S101: the gray matrix of original image is obtained.
As an example, the value range of original image gray matrix element typically [0,255].Therefore gray matrix
The data type of element is generally 8 signless integers, i.e. 256 gray level images.Wherein, " 0 " indicates ater, and " 255 " indicate
Pure white, intermediate number are indicated from small to large by black to white intermediate color.
Step S102: it is filtered to obtain Pasteur by gray matrix of Pasteur's likeness coefficient filter to original image
Similarity matrix.
Wherein, Pasteur's likeness coefficient filter can be realized by following formula:
Wherein, x0Indicate the pixel coordinate in original image gray matrix；y0Indicate the pixel in original image gray matrix
Coordinate；The parameter of i expression Gaussian kernel；The parameter of j expression Gaussian kernel；The width of W expression Pasteur's likeness coefficient filter；H table
Show the height of Pasteur's likeness coefficient filter；BS(x0,y0) indicate Pasteur's likeness coefficient matrix；G (i, j) indicates Gaussian kernel,
AndThe standard deviation of δ expression Gaussian Profile；p(x0+i,y0+ j) it indicates to be located in gray matrix
(x0+i,y0+ j) numerical value；L (i, j) indicates background template
In practical applications, it can configure the size of background template to as the size of Filtering Template.Such as: background
The size of template can be 5 × 5, and define θmean-thres≤θ(i)≤θmean+ thres i.e. in gray matrix any position,
Background template is all indicated with the non-zero constant value matrix of identical numerical value, wherein θmeanIndicate the median of tilt angle, thres
Indicate the threshold value of tilt angle, θ (i) indicates tilt angle.Here, the embodiment of the present invention is using non-zero constant value matrix come to part
Background carry out approximate representation be because are as follows: from the point of view of whole image, the variation of background may be bigger, with constant value matrix come approximate table
Show and unreasonable, but from the point of view of size is 5 × 5 part, grey scale change is not when background variation is in this one small range
It can be very big.
In some embodiments, x0It can indicate the pixel abscissa in original image gray matrix；y0It can indicate original
Pixel ordinate in image grayscale matrix.
Step S103: Pasteur's similarity matrix is normalized.
Such as: numerical value all in Pasteur's similarity matrix are zoomed between [0,1] in proportion.
Step S104: the numerical value in Pasteur's similarity matrix after normalization is generated into histogram according to size.
In practical applications, the numerical value in Pasteur's similarity matrix after normalization can be generated one according to size
The histogram in 1000 channels.
Step S105: on the histogram, two-value division is carried out using OTSU method, obtains two values matrix.
Wherein, OTSU method is carrying out image threshold segmentation method.On two values matrix, 1 indicates character point, and 0 indicates background dot.
Step S110: the connected domain that character point is constituted in statistics two values matrix, and filter out length, width, length-width ratio
Meet the character connected domain of the first pre-provisioning request, and obtains the position of the character connected domain for meeting the first pre-provisioning request and big
It is small.
This step counts the connected domain of character point in two values matrix, records the height and width of the connected domain of each character point
Degree, and the average height and mean breadth of the connected domain of character point are calculated, it filters out length, width, length-width ratio and meets first
The character connected domain of pre-provisioning request.
Illustratively, the first pre-provisioning request can be set as:
5≤height (i)≤50 and 5≤width (i)≤50, and
0.3 × mean_height≤height (i)≤2 × mean_height, and
0.5 × mean_width≤width (i)≤2 × mean_width, and
Wherein, height (i) indicates the height of the connected domain of character point；Width (i) indicates the width of the connected domain of character point
Degree；1≤i≤CN；CN indicates the number of the connected domain of character point；Mean_height indicates the mean height of the connected domain of character point
Degree；Mean_width indicates the mean breadth of the connected domain of character point.
Step S120: line of text is drawn according to the position for the character connected domain for meeting the first pre-provisioning request and size
Point, the number and location information of the character connected domain that the number and location information and each line of text for obtaining line of text contain.
As an example, this step can be according to the starting ordinate and end ordinate of character connected domain, to obtain text
Capable location information, the location information may include the information of the starting column and end column of line of text.
Specifically, this step can be realized further by step S121 to step S122.
Step S121: position and size based on the character connected domain for meeting the first pre-provisioning request carry out following judgement:
If the ordinate of any two character connected domain has overlapping region, which is divided into same
Otherwise two character connected domains are divided into different line of text by a line of text；
If any one character connected domain and current all line of text non-overlapping region on the vertical scale, for character connection
Domain creates a new line of text；
If the ordinate of any one character connected domain and any line of text has overlapping region, which is divided
To this article current row；
If the ordinate of any character connected domain and any two line of text has overlapping region, which is connected to
Domain is divided into that big line of text of overlapping region proportion.
Step S122: traversing all character connected domains, is based on above-mentioned judging result, obtains the number and position letter of line of text
The number and location information for the character connected domain that breath and line of text contain.
For example, in practical applications, the connection of each character can be handled in a manner of a in accordance with the following steps to step d
Domain:
Step a: when the ordinate of character connected domain x (i) range [upRow (i), downRow (i)] not with any one
When range [lineUpRow (j), lineDownRow (the j)] intersection of the ordinate of line of text, it may be assumed that
Then
A new line of text k is created, and enables lineUpRow (k)=upRow (i), lineDownRow (k)=downRow (i).
Step b: if range [upRow (i), downRow (i)] and a certain text when the ordinate of character connected domain x (i)
Row j intersection, it may be assumed that
The character connected domain is then divided into line of text j, and updates the information of line of text j in the following manner:
LineUpRow (j)=min (lineUpRow (j), upRow (i))
LineDownRow (j)=max (lineDownRow (j), downRow (i)).
Step c: if the range [upRow (i), downRow (i)] of the ordinate of character connected domain x (i) and multiple line of text
The character connected domain, then be divided into that maximum line of text of intersecting ranges, and update this in the way of mode (2) by intersection
The information of line of text.
Wherein, i indicates the serial number of character connected domain；The line number of j expression line of text；X (i) indicates character connected domain；upRow
(i) the starting ordinate of character connected domain is indicated；The end ordinate of downRow (i) expression character connected domain；Line (j) table
Show line of text；The starting ordinate of lineUpRow (j) expression line of text；LineDownRow (j) indicates that the end of line of text is vertical
Coordinate.
Step d: statistics obtains the information of all line of text, comprising number of characters less than certain predetermined quantity (such as: 3)
Line of text is deleted.
Step S130: Hough transformation is carried out to two values matrix, extracts the line segment in original image, and calculate inclining for the line segment
The median of rake angle and tilt angle.
Specifically, this step may further include:
Step S131: carrying out polar coordinate transform to the character point in two values matrix, and value in polar coordinate space is greater than threshold
Point of the character point of value as line segment alternative in image space, and the alternative line segment is inverted in image space.
Such as: in two values matrix, polar coordinate transform only is carried out to 1 point (character point), by value in polar coordinate space
Point of the point as line segment alternative in image space greater than 10, and be inverted in the corresponding position of image space.
Step S132: based on the alternative line segment being inverted in image space, the start-stop point of all alternative line segments is counted
It sets, length and tilt angle, and line segment is screened according to the length of alternative line segment.
In the specific implementation process, this step is inverted to the straight line in image space for each, finds straight across this
The all pixels point of line, the start-stop position of line segment is found by the abscissa of these pixels, and then can calculate the line segment
Length, then length be less than length threshold (such as: line segment 20) is deleted.
Step S133: for the line segment filtered out, the median of the line segment tilt angle is calculated.
Step S140: for all line segments extracted, according to the median of the tilt angle of the line segment and tilt angle,
Calculating does not meet ratio shared by the line segment of the second pre-provisioning request.
Wherein, the second pre-provisioning request can be set are as follows:
θmean-thres≤θ(i)≤θmean+thres
Wherein, θ (i) indicates tilt angle；θmeanIndicate the median of tilt angle；The threshold of thres expression tilt angle
Value.
In some embodiments, θ (i) can be calculated according to the following formula:
Wherein, (x_left, y_left) indicates the location information of each line segment left end point；(x_right,y_right)
Indicate the location information of each line segment right endpoint.
If meeting θmean-thres≤θ(i)≤θmean+ thres, then the line segment is the line segment of one " meeting expection "；It is no
It then, is the line segment of one " not meeting expection ".Then, the line segment of " not meeting expection " (namely second pre-provisioning request) is counted
Ratio.
The embodiment of the present invention is not using this index of ratio shared by expected line segment is met, mainly due to examining below
Consider:
In the actual implementation process, tolerance interval can be considered, for example, the tolerance interval is [θmean- 10 °, θmean+
10 °], it is to meet expected line segment that tilt angle, which is in the range, it is on the contrary then not meet expected line segment, can then unite
It counts out and does not meet ratio shared by expected line segment.For text image, most of line segment of acquisition is that parallel lines are disconnected, inclination
Angular deviation will not be very big, therefore does not meet shared by expected line segment than regular meeting very little；And for non-textual image, acquisition
Line segment direction relatively has diversity, and the tilt angle difference of different line segments is bigger, therefore does not meet shared by expected line segment
It is very bigger than regular meeting.
Step S150: the mean value and variance of channel S in the gray space and HSV space of original image are counted.
Step S160: according to establishing Naive Bayes Classification Model, and realize using the Naive Bayes Classification Model former
The text of beginning image and non-textual image classification.
This step may further include:
Step S161: building includes the image set of text image and non-textual image.
In the specific implementation process, image set can be divided into text image and non-textual image by the way of artificial.
Step S162: for the text image and non-textual image in image set, the number based on line of text is extracted respectively
The number and location information of the character connected domain contained with location information, each line of text, the line for not meeting the second pre-provisioning request
Ratio, the feature of mean value and variance of Duan Suozhan, and construction feature vector.
According to the number and location information of the character connected domain that the number of line of text and location information, each line of text contain
It can determine that character area area accounts for the ratio of original image.
As an example, can also be according to following information come construction feature vector: character connected domain number, character area area
Account for the ratio of original image, the number of line of text, do not meet ratio shared by the line segment of the second pre-provisioning request, image grayscale space and
The mean value and variance of channel S in HSV space.
In this way, for every picture xiAll available feature vectorsIt indicates, wherein d indicates feature dimensions
Number.
Step S163: being based on feature vector, constructs the Naive Bayes Classification Model of 2 classifications.
Wherein, the process for establishing Naive Bayes Classification Model may include:
Step S1631: training sample is established.
Such as: 1000 images, the training sample as model can be randomly selected from image data source.
Step S1632: training sample is divided into text image class sample and non-textual image class sample.
Text image can be chosen from above-mentioned 1000 picture as text image class sample, remaining is non-textual figure
As class sample.
Step S1633: text image class sample proportion and non-textual image class sample proportion are calculated.
Such as: p (ω1) indicating text image class sample proportion, then non-textual image class sample proportion is p
(ω2)=1-p (ω1).Wherein, ω1Indicate text image class sample, ω2Indicate non-textual image class sample.
Step S1634: according to text image class sample proportion and non-textual image class sample proportion, estimation is special
Levy the class conditional probability distribution of every one-dimensional characteristic in vector.
Specifically, estimation process can be accomplished by the following way: the feature vector of all training samples be calculated, to feature
Vector is all counted per one-dimensional, and the class conditional probability distribution p (x of every dimensional feature is estimated according to statistical resultj|ωk)。
Wherein, k indicates classification；J indicates dimension；xjIndicate component of the sample in jth dimension；ωkIndicate k-th of classification.
Step S1635: according to the class conditional probability distribution of every one-dimensional characteristic, the class conditional probability of each training sample is calculated
Distribution.
Specifically, the class conditional probability distribution of each training sample can be calculated according to the following formula:
Wherein, p (xj|ωk) indicate the class conditional probability distribution of every dimensional feature；p(xi|ωk) indicate each training sample
Class conditional probability distribution；K indicates classification；ωkIndicate k-th of classification；J indicates dimension.
Step S1636: Naive Bayes Classification Model is established according to the following formula:
Wherein, xiIndicate training sample；ω*Indicate classification belonging to the training sample being inferred to.
Step S164: according to Naive Bayes Classification Model, text image and non-textual image are carried out to original image
Classification.
This step, which does piece image according to established Naive Bayes Classification Model, classifies, to judge whether it belongs to
In text image.
By using above-described embodiment, the screening operation in mass image data to text image can be completed, to screening
Text image out can carry out the related works such as subsequent OCR identification again.
What needs to be explained here is that although each step is carried out in the way of above-mentioned precedence in above-described embodiment
Description, need not be by it will be recognized to those skilled in the art that effect in order to realize the present embodiment, between different steps
Order after this manner executes, and (parallel) simultaneously can execute or be executed with reverse order, these simple variations are all originally
Within the protection scope of invention.
Based on technical concept identical with embodiment of the method, the embodiment of the present invention also provide a kind of text image with it is non-textual
The categorizing system of image.As shown in Fig. 2, the system 20 may include: that the first acquisition module 21, second obtains module 22, divides
Module 23, the first computing module 24, the second computing module 25, statistical module 26 and categorization module 27.Wherein, first module is obtained
21 for obtaining the two values matrix of original image.Second acquisition module 22 is connected with the first acquisition module 21, for counting two-value
The connected domain that character point is constituted in matrix, and filter out the character connection that length, width, length-width ratio meet the first pre-provisioning request
Domain, and obtain position and the size of the character connected domain for meeting the first pre-provisioning request.Division module 23 and second obtains module 22
It is connected, for dividing according to the position and size for the character connected domain for meeting the first pre-provisioning request to line of text, obtains text
The number and location information for the character connected domain that the number and location information and line of text of current row contain.First computing module
24 are connected with the first acquisition module 21, for carrying out Hough transformation to two values matrix, extract the line segment in original image, and calculate
The tilt angle of line segment and the median of tilt angle.Second computing module 25 is connected with the first computing module 24, for being directed to
All line segments extracted, according to the median of the tilt angle of line segment and tilt angle, calculating does not meet the second pre-provisioning request
Line segment shared by ratio.Statistical module 26 be used to count in the gray space and HSV space of original image the mean value of channel S and
Variance.Categorization module 27 is connected with division module 23, the second computing module 25 and statistical module 26 respectively, for according to line of text
Number and the number of character connected domain that contains of location information, each line of text and location information, do not meet second and predetermined want
Ratio shared by the line segment asked and mean value and variance establish Naive Bayes Classification Model, and utilize Naive Bayes Classification
The classification of the text image of model realization original image and non-textual image.
Explanation in relation to the present embodiment can be with reference to the related content in embodiment of the method, and details are not described herein.
By using above-described embodiment, the screening operation in mass image data to text image can be completed, to screening
Text image out can also carry out the related works such as subsequent OCR identification again.
In some embodiments, on the basis of embodiment shown in Fig. 2, above-mentioned first acquisition module can also be wrapped specifically
It includes: acquiring unit, filter unit, normalization unit and two-value division unit.Wherein, acquiring unit is for obtaining original image
Gray matrix.Filter unit by gray matrix of Pasteur's likeness coefficient filter to original image for being filtered to obtain
Pasteur's similarity matrix.Normalization unit is for being normalized Pasteur's similarity matrix.Processing unit will be for that will return
The numerical value in Pasteur's similarity matrix after one change generates histogram according to size.Two-value division unit is used on the histogram,
Two-value division is carried out using OTSU method, obtains two values matrix.
Explanation in relation to the present embodiment can be with reference to the related content in embodiment of the method, and details are not described herein.
In some embodiments, on the basis of embodiment shown in Fig. 2, above-mentioned first computing module can also be wrapped specifically
It includes: converter unit, screening unit and computing unit.Wherein, converter unit is used to carry out pole seat to the character point in two values matrix
Value in polar coordinate space is greater than the character point of threshold value as the point of line segment alternative in image space by mark transformation, and will be alternative
Line segment is inverted in image space.Screening unit is used to count all standby based on the alternative line segment being inverted in image space
Start-stop point position, length and the tilt angle of route selection section, and line segment is screened according to the length of alternative line segment.Computing unit is used for needle
To the line segment filtered out, the median of line segment tilt angle is calculated.
Explanation in relation to the present embodiment can be with reference to the related content in embodiment of the method, and details are not described herein.
In some embodiments, on the basis of embodiment shown in Fig. 2, above-mentioned categorization module can also be specifically included: the
One construction unit, the second construction unit, third construction unit and taxon.Wherein, the first construction unit includes for constructing
The image set of text image and non-textual image.Second construction unit is used for for the text image and non-textual figure in image set
Picture extracts number and location information based on line of text, the number for the character connected domain that each line of text contains and position letter respectively
Cease, do not meet the feature of ratio shared by the line segment of the second pre-provisioning request, mean value and variance, and construction feature vector.Third structure
Unit is built for based on feature vector, the Naive Bayes Classification Model of 2 classifications of building.Taxon is used for according to simple pattra leaves
This disaggregated model carries out the classification of text image and non-textual image to original image.
Explanation in relation to the present embodiment can be with reference to the related content in embodiment of the method, and details are not described herein.
It should be noted that the classification method and categorizing system of text image provided by the above embodiment and non-textual image
When carrying out image classification, only carried out with the division of above-mentioned each functional module or step for example, in practical applications, it can be with
Above-mentioned function distribution is completed by different functional module or step as needed, i.e., by the module in the embodiment of the present invention
Either step is decomposed or is combined again, for example, the first acquisition module of above-described embodiment and the second acquisition module can be merged into
One acquisition module, can also be further split into multiple submodule, to complete all or part of the functions described above.It is right
The title of module, step involved in the embodiment of the present invention, it is only for distinguish modules or step, be not intended as pair
Improper restriction of the invention.
It will be understood by those skilled in the art that the categorizing system of above-mentioned text image and non-textual image can also include one
A little other known features, such as processor, controller, memory and bus etc., wherein memory is including but not limited to deposited at random
Reservoir, flash memory, read-only memory, programmable read only memory, volatile memory, nonvolatile memory, serial storage,
Parallel storage or register etc., processor include but is not limited to CPLD/FPGA, DSP, arm processor, MIPS processor etc.；
Bus may include data/address bus, address bus and control bus.In order to unnecessarily obscure embodiment of the disclosure, these are public
The structure known is not shown in FIG. 2.
It should be understood that the quantity of the modules in Fig. 2 is only schematical.According to actual needs, each module can be with
With arbitrary quantity.
The above system embodiment can be used for executing above method embodiment, technical principle, it is solved the technical issues of
And the technical effect generated is similar, person of ordinary skill in the field can be understood that, for the convenience and letter of description
Clean, the specific work process of the system of foregoing description and related explanation can refer to corresponding processes in the foregoing method embodiment,
Details are not described herein.
It should be pointed out that system embodiment and embodiment of the method for the invention are described respectively above, but it is right
The details of one embodiment description can also be applied to another embodiment.
Technical solution is provided for the embodiments of the invention above to be described in detail.Although applying herein specific
A example the principle of the present invention and embodiment are expounded, still, the explanation of above-described embodiment be only applicable to help manage
Solve the principle of the embodiment of the present invention；Meanwhile to those skilled in the art, according to an embodiment of the present invention, it is being embodied
It can be made a change within mode and application range.
It should be noted that: label and text in attached drawing are intended merely to be illustrated more clearly that the present invention, are not intended as pair
The improper restriction of the scope of the present invention.
Again it should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, "
Two ", " third " etc. is to be used to distinguish similar objects, rather than be used to describe or indicate specific sequence or precedence.It answers
The data that the understanding uses in this way can be interchanged in appropriate circumstances, so that the embodiment of the present invention described herein can be with
Sequence other than those of illustrating or describing herein is implemented.
Term " includes " or any other like term are intended to cover non-exclusive inclusion, so that including a system
Process, method, article or equipment/device of column element not only includes those elements, but also including being not explicitly listed
Other elements, or further include the intrinsic element of these process, method, article or equipment/devices.
As used herein, term " module ", " unit " may refer to the software object executed on a computing system
Or routine.Disparate modules described herein can be embodied as to the object executed on a computing system or process (for example, making
It is independent thread).While it is preferred that realize system and method described herein with software, but with hardware or soft
The realization of the combination of part and hardware is also possible and can be conceived to.
Each step of the invention can be realized with general computing device, for example, they can concentrate on it is single
On computing device, such as: personal computer, server computer, handheld device or portable device, laptop device or more
Processor device can also be distributed over a network of multiple computing devices, they can be to be different from sequence herein
Shown or described step is executed, perhaps they are fabricated to each integrated circuit modules or will be more in them
A module or step are fabricated to single integrated circuit module to realize.Therefore, the present invention is not limited to any specific hardware and soft
Part or its combination.
Method provided by the invention can also be realized using programmable logic device, also may be embodied as computer program
(it includes routines performing specific tasks or implementing specific abstract data types, programs, objects, component for software or program module
Or data structure etc.), such as embodiment according to the present invention can be a kind of computer program product, run the computer journey
Sequence product executes computer for demonstrated method.The computer program product includes computer readable storage medium,
It include computer program logic or code section on the medium, for realizing the method.The computer readable storage medium
It can be the built-in medium being mounted in a computer or the removable medium (example that can be disassembled from basic computer
Such as: using the storage equipment of hot plug technology).The built-in medium includes but is not limited to rewritable nonvolatile memory,
Such as: RAM, ROM, flash memory and hard disk.The removable medium includes but is not limited to: and optical storage media (such as: CD-
ROM and DVD), magnetic-optical storage medium (such as: MO), magnetic storage medium (such as: tape or mobile hard disk), can with built-in
Rewrite the media (such as: storage card) of nonvolatile memory and the media (such as: ROM box) with built-in ROM.
It shall also be noted that language used in this specification primarily to readable and introduction purpose and select,
Rather than in order to explain or defining the subject matter of the present invention and select.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain
Lid is within protection scope of the present invention.Therefore, protection scope of the present invention should be based on the protection scope of the described claims.
Claims (10)
1. the classification method of a kind of text image and non-textual image, which is characterized in that the method includes at least:
Obtain the two values matrix of original image；
The connected domain that character point is constituted in the two values matrix is counted, and filters out length, width, length-width ratio and meets first in advance
The character connected domain of provisioning request, and obtain position and the size of the character connected domain for meeting first pre-provisioning request；
Line of text is divided according to the position of the character connected domain for meeting the first pre-provisioning request and size, is obtained described
The number and location information for the character connected domain that the number and location information and the line of text of line of text contain；
Hough transformation is carried out to the two values matrix, extracts the line segment in the original image, and calculate the inclination of the line segment
The median of angle and the tilt angle；
For all line segments extracted, according to the middle position of the tilt angle of the line segment and the tilt angle
Number, calculating do not meet ratio shared by the line segment of the second pre-provisioning request；
Count the mean value and variance of channel S in the gray space and HSV space of the original image；
The number of the character connected domain contained according to the number of the line of text and location information, each described line of text and position
Ratio and the mean value and the variance shared by confidence breath, the line segment for not meeting the second pre-provisioning request are established simple
Bayesian Classification Model, and realize using the Naive Bayes Classification Model text image and the institute of the original image
State the classification of non-textual image.
2. the method according to claim 1, wherein the two values matrix for obtaining original image specifically includes:
Obtain the gray matrix of the original image；
It is filtered to obtain the Pasteur by gray matrix of Pasteur's likeness coefficient filter to the original image similar
Property matrix；
Pasteur's similarity matrix is normalized；
Numerical value in Pasteur's similarity matrix after normalization is generated into histogram according to size；
On the histogram, two-value division is carried out using OTSU method, obtains two values matrix.
3. the method according to claim 1, wherein the character for meeting the first pre-provisioning request according to connects
The position in logical domain and size divide line of text, obtain the number and location information and the text of the line of text
The number and location information for the character connected domain that row contains, specifically include:
Position and size based on the character connected domain for meeting the first pre-provisioning request carry out following judgement:
If there are overlapping regions for the ordinate of any two character connected domain, described two character connected domains are divided into same
Otherwise described two character connected domains are divided into different line of text by line of text；
If any character connected domain and current all line of text non-overlapping region on the vertical scale, for the character
Connected domain creates a new line of text；
If there are overlapping regions for the ordinate of any character connected domain and any line of text, by any character
Connected domain is divided into this article current row；
If there are overlapping regions for the ordinate of any character connected domain and wantonly two line of text, by any character
Connected domain is divided into the big line of text of overlapping region proportion；
All character connected domains are traversed, above-mentioned judging result is based on, obtains the number and location information of the line of text, with
And the number and location information of character connected domain that the line of text contains.
4. being extracted the method according to claim 1, wherein described carry out Hough transformation to the two values matrix
Line segment in the original image, and the tilt angle of the line segment and the median of the tilt angle are calculated, it specifically includes:
Polar coordinate transform is carried out to the character point in the two values matrix, value in polar coordinate space is greater than to the character point of threshold value
As the point of line segment alternative in image space, and the alternative line segment is inverted in described image space；
Based on the alternative line segment being inverted in described image space, the start-stop point of all alternative line segments is counted
It sets, length and tilt angle, and line segment is screened according to the length of the alternative line segment；
For the line segment filtered out, the median of the line segment tilt angle is calculated.
5. the method according to claim 1, wherein the number and location information according to the line of text,
The number and location information of the character connected domain that each described line of text contains, the line segment for not meeting the second pre-provisioning request
Shared ratio and the mean value and the variance establish Naive Bayes Classification Model, and utilize the naive Bayesian
Disaggregated model realizes the text image of the original image and the classification of the non-textual image, specifically includes:
Building includes the image set of the text image and the non-textual image；
For the text image and the non-textual image that described image is concentrated, based on the line of text is extracted respectively
The number and location information of the character connected domains that several and location information, each described line of text contain described do not meet second
The feature of ratio shared by the line segment of pre-provisioning request, the mean value and the variance, and construction feature vector；
Based on described eigenvector, the Naive Bayes Classification Model of 2 classifications is constructed；
According to the Naive Bayes Classification Model, the text image is carried out to original image and is divided with the non-textual image
Class.
6. according to the method described in claim 5, it is characterized in that, described be based on described eigenvector, the simplicity of 2 classifications of building
Bayesian Classification Model specifically includes:
Establish training sample；
The training sample is divided into text image class sample and non-textual image class sample；
Calculate the text image class sample proportion and the non-textual image class sample proportion；
According to the text image class sample proportion and the non-textual image class sample proportion, the feature is estimated
The class conditional probability distribution of each dimensional feature in vector；
According to the class conditional probability distribution of each dimensional feature, the class conditional probability distribution of each training sample is calculated；
Naive Bayes Classification Model is established according to the following formula:
Wherein,P (the xj|ωk) indicate that the class condition of each dimensional feature is general
Rate distribution；P (the xi|ωk) indicate the class conditional probability distribution of each training sample；The k indicates classification；The ωkTable
Show k-th of classification；The j indicates dimension；The xiIndicate the training sample；The ω*Indicate the trained sample being inferred to
Classification belonging to this.
7. the categorizing system of a kind of text image and non-textual image, which is characterized in that the system includes at least:
First obtains module, for obtaining the two values matrix of original image；
Second obtains module, for counting the connected domain that character point is constituted in the two values matrix, and filters out length, width
Degree, length-width ratio meet the character connected domain of the first pre-provisioning request, and obtain the character company for meeting first pre-provisioning request
The position in logical domain and size；
Division module, position and size for meeting the character connected domain of the first pre-provisioning request according to carry out line of text
It divides, the number for the character connected domain that the number and location information and the line of text for obtaining the line of text contain
And location information；
First computing module extracts the line segment in the original image, and count for carrying out Hough transformation to the two values matrix
Calculate the tilt angle of the line segment and the median of the tilt angle；
Second computing module, for being directed to all line segments extracted, according to the tilt angle of the line segment and institute
The median of tilt angle is stated, calculating does not meet ratio shared by the line segment of the second pre-provisioning request；
Statistical module, the mean value and variance of channel S in the gray space and HSV space for counting the original image；
Categorization module, the character contained for number and location information, each described line of text according to the line of text connect
Ratio and the mean value shared by the number and location information, the line segment for not meeting the second pre-provisioning request in logical domain and described
Variance establishes Naive Bayes Classification Model, and the institute of the original image is realized using the Naive Bayes Classification Model
State the classification of text image Yu the non-textual image.
8. system according to claim 7, which is characterized in that the first acquisition module specifically includes:
Acquiring unit, for obtaining the gray matrix of the original image；
Filter unit, for being filtered to obtain by gray matrix of Pasteur's likeness coefficient filter to the original image
Pasteur's similarity matrix；
Normalization unit, for Pasteur's similarity matrix to be normalized；
Processing unit generates histogram according to size for the numerical value in Pasteur's similarity matrix after normalizing；
Two-value division unit, for carrying out two-value division using OTSU method, obtaining two values matrix in the histogram.
9. system according to claim 7, which is characterized in that first computing module specifically includes:
Converter unit is big by value in polar coordinate space for carrying out polar coordinate transform to the character point in the two values matrix
In point of the character point as line segment alternative in image space of threshold value, and the alternative line segment is inverted to described image space
In；
Screening unit, for counting all alternative lines based on the alternative line segment being inverted in described image space
Start-stop point position, length and the tilt angle of section, and line segment is screened according to the length of the alternative line segment；
Computing unit, for calculating the median of the line segment tilt angle for the line segment filtered out.
10. system according to claim 7, which is characterized in that the categorization module specifically includes:
First construction unit, for constructing the image set including the text image and the non-textual image；
Second construction unit, the text image and the non-textual image for concentrating for described image, is extracted respectively
The number for the character connected domain that number and location information, each described line of text based on the line of text contain and position letter
Ratio, the feature of the mean value and the variance shared by breath, the line segment for not meeting the second pre-provisioning request, and construction feature
Vector；
Third construction unit constructs the Naive Bayes Classification Model of 2 classifications for being based on described eigenvector；
Taxon, for according to the Naive Bayes Classification Model, to original image carry out the text image with it is described
The classification of non-textual image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610892308.0A CN106503732B (en) | 2016-10-13 | 2016-10-13 | The classification method and categorizing system of text image and non-textual image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610892308.0A CN106503732B (en) | 2016-10-13 | 2016-10-13 | The classification method and categorizing system of text image and non-textual image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106503732A CN106503732A (en) | 2017-03-15 |
| CN106503732B true CN106503732B (en) | 2019-07-19 |
Family
ID=58294876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610892308.0A Expired - Fee Related CN106503732B (en) | 2016-10-13 | 2016-10-13 | The classification method and categorizing system of text image and non-textual image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106503732B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108537283A (en) * | 2018-04-13 | 2018-09-14 | 厦门美图之家科技有限公司 | A kind of image classification method and convolutional neural networks generation method |
| CN109543709A (en) * | 2018-10-11 | 2019-03-29 | 成都三零凯天通信实业有限公司 | Rapid picture preprocessing method |
| CN109948549B (en) * | 2019-03-20 | 2022-11-29 | 深圳市华付信息技术有限公司 | OCR data generation method and device, computer equipment and storage medium |
| CN110135519B (en) * | 2019-05-27 | 2022-10-21 | 广东工业大学 | Image classification method and device |
| CN110321892B (en) * | 2019-06-04 | 2022-12-13 | 腾讯科技（深圳）有限公司 | Picture screening method and device and electronic equipment |
| CN110414505A (en) * | 2019-06-27 | 2019-11-05 | 深圳中兴网信科技有限公司 | Processing method, processing system and the computer readable storage medium of image |
| CN110378282B (en) * | 2019-07-18 | 2021-11-02 | 北京字节跳动网络技术有限公司 | Image processing method and device |
| CN112836696A (en) * | 2019-11-22 | 2021-05-25 | 搜狗（杭州）智能科技有限公司 | Text data detection method and device and electronic equipment |
| CN111881916B (en) * | 2020-07-17 | 2023-07-25 | 中国工商银行股份有限公司 | Text positioning method, device and equipment |
| CN112966318B (en) * | 2021-04-12 | 2022-06-10 | 青矩技术股份有限公司 | Method for displaying characters in CAD file and terminal equipment |
| CN113610068B (en) * | 2021-10-11 | 2022-07-08 | 江西风向标教育科技有限公司 | Test question disassembling method, system, storage medium and equipment based on test paper image |
| CN115410191B (en) * | 2022-11-03 | 2023-02-03 | 平安银行股份有限公司 | Text image recognition method, device, equipment and storage medium |
| CN116071652A (en) * | 2023-02-21 | 2023-05-05 | 吉林大学 | Metamorphic rock reservoir corrosion pore identification method based on imaging logging and Bayesian classification |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101266654A (en) * | 2007-03-14 | 2008-09-17 | 中国科学院自动化研究所 | Image text location method and device based on connective component and support vector machine |
| CN101840514A (en) * | 2009-03-19 | 2010-09-22 | 株式会社理光 | Image object classification device and method |
| CN102163284A (en) * | 2011-04-11 | 2011-08-24 | 西安电子科技大学 | Chinese environment-oriented complex scene text positioning method |
| CN103136523A (en) * | 2012-11-29 | 2013-06-05 | 浙江大学 | Arbitrary direction text line detection method in natural image |
| US9076056B2 (en) * | 2013-08-20 | 2015-07-07 | Adobe Systems Incorporated | Text detection in natural images |
| CN105761219A (en) * | 2016-02-03 | 2016-07-13 | 北京云江科技有限公司 | Inclination correction method and system of text image |
| CN105844275A (en) * | 2016-03-25 | 2016-08-10 | 北京云江科技有限公司 | Method for positioning text lines in text image |
-
2016
- 2016-10-13 CN CN201610892308.0A patent/CN106503732B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101266654A (en) * | 2007-03-14 | 2008-09-17 | 中国科学院自动化研究所 | Image text location method and device based on connective component and support vector machine |
| CN101840514A (en) * | 2009-03-19 | 2010-09-22 | 株式会社理光 | Image object classification device and method |
| CN102163284A (en) * | 2011-04-11 | 2011-08-24 | 西安电子科技大学 | Chinese environment-oriented complex scene text positioning method |
| CN103136523A (en) * | 2012-11-29 | 2013-06-05 | 浙江大学 | Arbitrary direction text line detection method in natural image |
| US9076056B2 (en) * | 2013-08-20 | 2015-07-07 | Adobe Systems Incorporated | Text detection in natural images |
| CN105761219A (en) * | 2016-02-03 | 2016-07-13 | 北京云江科技有限公司 | Inclination correction method and system of text image |
| CN105844275A (en) * | 2016-03-25 | 2016-08-10 | 北京云江科技有限公司 | Method for positioning text lines in text image |
Non-Patent Citations (2)
| Title |
|---|
| A Novel Mutual Nearest Neighbor based Symmetry for Text Frame Classification in Video;Palaiahnakote Shivakumara et al.;《Pattern Recognition》;20110831;第44卷(第8期);第1671-1683页 |
| 基于机器学习的文字判别算法;黄俊亮 等;《信息安全与通信保密》;20100305(第12期);第64-66页，第69页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106503732A (en) | 2017-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106503732B (en) | The classification method and categorizing system of text image and non-textual image | |
| CN105844285B (en) | Cucumber disease recognition methods and device based on image information | |
| Pan et al. | Leukocyte image segmentation by visual attention and extreme learning machine | |
| CN101877064B (en) | Image classification method and image classification device | |
| Comaniciu et al. | Bimodal system for interactive indexing and retrieval of pathology images | |
| Strokina et al. | Detection of bubbles as concentric circular arrangements | |
| Chica | Authentication of bee pollen grains in bright‐field microscopy by combining one‐class classification techniques and image processing | |
| CN110232318A (en) | Acupuncture point recognition methods, device, electronic equipment and storage medium | |
| CN109002463A (en) | A kind of Method for text detection based on depth measure model | |
| CN109710788A (en) | Image pattern mark and management method and equipment | |
| CN109446300A (en) | A kind of corpus preprocess method, the pre- mask method of corpus and electronic equipment | |
| CN105893916A (en) | New method for detection of face pretreatment, feature extraction and dimensionality reduction description | |
| Mehri et al. | A texture-based pixel labeling approach for historical books | |
| Xiao et al. | Adaptive trapezoid region intercept histogram based Otsu method for brain MR image segmentation | |
| Christlein | Handwriting analysis with focus on writer identification and writer retrieval | |
| CN108664968B (en) | Unsupervised text positioning method based on text selection model | |
| CN113378609B (en) | Agent proxy signature identification method and device | |
| CN105631451A (en) | Plant leave identification method based on android system | |
| CN116612478A (en) | Off-line handwritten Chinese character scoring method, device and storage medium | |
| Lee et al. | Learning non-homogenous textures and the unlearning problem with application to drusen detection in retinal images | |
| CN113139581B (en) | Image classification method and system based on multi-image fusion | |
| Peters | Fuzzy sets, near sets, and rough sets for your computational intelligence toolbox | |
| Marantan et al. | Image cognition using contour curvature statistics | |
| Zhou et al. | Comparative analysis of the performance of complex texture clustering driven by computational intelligence methods using multiple clustering models | |
| AU2021105815A4 (en) | A Multi-Faceted Hierarchical Image Segmentation Taxonomy (MFHIST) System for A Categorization of An Image Segmentation Algorithms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20191108 Address after: Room 501, floor 5, comprehensive building, Zhongchuang Industrial Park, New District, Huaibei Economic Development Zone, Anhui Province Patentee after: Huaibei Avanti Education Technology Co.,Ltd. Address before: 100083 Beijing city Haidian District Wudaokou Rossmax building A, room 1801 Patentee before: BEIJING LEJENT TECHNOLOGY Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190719 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |