CN106295631A - A kind of image Uighur word recognition methods and device - Google Patents
A kind of image Uighur word recognition methods and device Download PDFInfo
- Publication number
- CN106295631A CN106295631A CN201610609772.4A CN201610609772A CN106295631A CN 106295631 A CN106295631 A CN 106295631A CN 201610609772 A CN201610609772 A CN 201610609772A CN 106295631 A CN106295631 A CN 106295631A
- Authority
- CN
- China
- Prior art keywords
- image
- uighur
- words
- parts
- line
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Character Input (AREA)
Abstract
The invention provides a kind of image Uighur word recognition methods and device, belong to field of optical character recognition.The method includes: obtain Uighur words image;According to the baseline territory that Uighur words image is corresponding, each disjunctor section of Uighur words image is divided into one or more parts;The characteristic information of each parts obtaining Uighur words image obtains the characteristic vector of Uighur words image;The characteristic vector of Uighur words image is contrasted with the feature lexicon preset, with the textual words that the characteristic vector of the Uighur words image acquired in acquisition is corresponding.The present invention utilizes the characteristic information construction feature vector of the parts in accessed Uighur words image as object to be identified, is effectively improved the recognizable rate of Uighur words image.Additionally, the characteristic vector construction feature dictionary directly obtaining textual words simplifies the building process of feature lexicon effectively.
Description
Technical field
The present invention relates to field of optical character recognition, in particular to a kind of image Uighur word recognition methods
And device.
Background technology
The research of block letter Uighur identification to the research of Uygur culture and the preservation of Uighur paper material and
Digitized important in inhibiting.Existing block letter Uighur recognition methods is all based on the thinking of " first cutting, identify again ".
Uighur words in scanogram is divided into letter, then extracts the characteristic information of letter as characteristic vector to be identified,
The characteristic vector that characteristic vector to be identified and previous sample training obtain in alphabetic feature vector storehouse is contrasted, thus to institute
The letter being syncopated as is identified.Wherein, alphabetic feature vector storehouse includes the feature of the alphabetical various fonts of Uygur 32
Vector.But, either block letter or handwritten form Uighur have write the two or more syllables of a word together and the most wide feature so that the border of letter
It is difficult to determine, causes character segmentation inaccurate, bring difficulty for Letter identification.Additionally, space between the disjunctor section of Uighur
With the space between word is difficult to differentiate, makes discrimination be restricted and have influence on post processing effect.
Summary of the invention
In consideration of it, it is an object of the invention to provide a kind of image Uighur word recognition methods and device, will be tieed up me
You are divided into multiple parts at literary composition word image, build described Uighur words figure by obtaining the characteristic information of each parts
The characteristic vector of picture, by identifying that described characteristic vector obtains the textual words that Uighur words image is corresponding.
To achieve these goals, the technical solution used in the present invention is as follows:
First aspect, embodiments provides a kind of image Uighur word recognition methods, and described method includes:
Obtaining Uighur words image, described Uighur words image includes one or more disjunctor section;According to described Uygur
Each disjunctor section of described Uighur words image is divided into one or more portion by baseline territory corresponding to literary composition word image
Part;The characteristic information of each described parts obtaining described Uighur words image obtains described Uighur words image
Characteristic vector;The characteristic vector of described Uighur words image is contrasted with the feature lexicon preset, to be obtained
The textual words that the characteristic vector of the described Uighur words image taken is corresponding, wherein, described feature lexicon includes according to literary composition
The described characteristic vector of this word acquisition and the corresponding relation of described textual words.
Second aspect, the embodiment of the present invention additionally provides a kind of image Uighur Word identifier, including: word graph
As acquisition module, parts segmentation module, characteristic vector acquisition module and identification module.Word image acquisition module is used for obtaining dimension
The civilian word image of my that, described Uighur words image includes one or more disjunctor section.Parts segmentation module is for basis
Each disjunctor section of described Uighur words image is divided into one by the baseline territory that described Uighur words image is corresponding
Individual or multiple parts.Characteristic vector acquisition module is for obtaining the feature of each described parts of described Uighur words image
Information obtains the characteristic vector of described Uighur words image.Identification module is for by the spy of described Uighur words image
Levy vector to contrast with the feature lexicon preset, with the characteristic vector pair of the described Uighur words image acquired in acquisition
The textual words answered, wherein, described feature lexicon includes the described characteristic vector according to textual words acquisition and described text list
The corresponding relation of word.
Compared to existing recognition methods, the image Uighur word recognition methods of embodiment of the present invention offer and device
It is to be one or more parts by Uighur words image cutting, it is not necessary to be syncopated as exactly in Uighur words image
Each letter, reduces the cutting difficulty of Uighur words image.Further, the Uighur words figure accessed by utilization
The characteristic information of the parts in Xiang builds the characteristic vector of this Uighur words image as object to be identified, is effectively improved
The recognizable rate of Uighur words image.
Accompanying drawing explanation
In order to be illustrated more clearly that the technical scheme of the embodiment of the present invention, below by embodiment required use attached
Figure is briefly described, it will be appreciated that the following drawings illustrate only certain embodiments of the present invention, and it is right to be therefore not construed as
The restriction of scope, for those of ordinary skill in the art, on the premise of not paying creative work, it is also possible to according to this
A little accompanying drawings obtain other relevant accompanying drawings.
Fig. 1 shows the block diagram of the computer that the embodiment of the present invention provides;
Fig. 2 shows the flow chart of a kind of image Uighur word recognition methods that the embodiment of the present invention provides;
Fig. 3 shows the flow chart of the another kind of image Uighur word recognition methods that the embodiment of the present invention provides;
Fig. 4 shows the example text image that the embodiment of the present invention provides;
Fig. 5 shows that obtained of splitting the text image shown in Fig. 4 that the embodiment of the present invention provides ties up me
That literary composition word image;
Fig. 6 shows disjunctor section and the schematic diagram in baseline territory of the Uighur words image shown in Fig. 5;
Fig. 7 shows the method flow diagram in the baseline territory obtaining each line of text image in step S204;
Fig. 8 shows that the one of a kind of image Uighur word recognition methods that the embodiment of the present invention provides is embodied as
The flow chart of mode;
Fig. 9 shows the parts segmentation schematic diagram of the Uighur words image shown in Fig. 5;
Figure 10 shows to be trizonal signal by the Uighur words image division shown in Fig. 5 along the longitudinal direction
Figure;
Figure 11 shows the structured flowchart of a kind of image Uighur Word identifier that the embodiment of the present invention provides;
Figure 12 shows the structured flowchart of the another kind of image Uighur Word identifier that the embodiment of the present invention provides;
Figure 13 shows that a kind of image Uighur Word identifier one that the embodiment of the present invention provides is embodied as
The structured flowchart of mode.
Detailed description of the invention
Below in conjunction with accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Ground describes, it is clear that described embodiment is only a part of embodiment of the present invention rather than whole embodiments.Generally exist
Can arrange and design with various different configurations with the assembly of the embodiment of the present invention that illustrates described in accompanying drawing herein.Cause
This, be not intended to limit claimed invention to the detailed description of the embodiments of the invention provided in the accompanying drawings below
Scope, but it is merely representative of the selected embodiment of the present invention.Based on embodiments of the invention, those skilled in the art are not doing
The every other embodiment obtained on the premise of going out creative work, broadly falls into the scope of protection of the invention.
It should also be noted that similar label and letter represent similar terms, therefore, the most a certain Xiang Yi in following accompanying drawing
Individual accompanying drawing is defined, then need not it be defined further and explains in accompanying drawing subsequently.
As it is shown in figure 1, be the block diagram of the computer 100 that preferred embodiment of the present invention provides.Described computer
100 include image Uighur Word identifier 200, memorizer 101, storage control 102, processor 103, Peripheral Interface
104, input-output unit 105.
Described memorizer 101, storage control 102, processor 103, Peripheral Interface 104, each yuan of input-output unit 105
Part is electrically connected with the most directly or indirectly, to realize the transmission of data or mutual.Such as, these elements each other may be used
Realize being electrically connected with by one or more communication bus or holding wire.Described image Uighur Word identifier 200 wraps
Include at least one and can be stored in the software function module in described memorizer 101 with the form of software or firmware (firmware).
Described processor 103 is for performing the executable module of storage in memorizer 101, and such as, described image Uighur words is known
Software function module that other device 200 includes or computer 100 program.
Wherein, memorizer 101 may be, but not limited to, random access memory (Random Access Memory,
RAM), read only memory (Read Only Memory, ROM), programmable read only memory (Programmable Read-Only
Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only Memory, EPROM),
Electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only Memory, EEPROM) etc..
Wherein, memorizer 101 is used for storing program, and described processor 103, after receiving execution instruction, performs described program, aforementioned
The method performed by the server flowing through Cheng Dingyi that embodiment of the present invention any embodiment discloses can apply to processor 103
In, or realized by processor 103.
Processor 103 is probably a kind of IC chip, has the disposal ability of signal.Above-mentioned processor 103 can
To be general processor, including central processing unit (Central Processing Unit is called for short CPU), network processing unit
(Network Processor is called for short NP) etc.;Can also is that digital signal processor (DSP), special IC (ASIC),
Ready-made programmable gate array (FPGA) or other PLDs, discrete gate or transistor logic, discrete firmly
Part assembly.Can realize or perform disclosed each method, step and the logic diagram in the embodiment of the present invention.General processor
Can be microprocessor or this processor 103 can also be the processor etc. of any routine.
Various input/output devices are coupled to processor 103 and memorizer 101 by described Peripheral Interface 104.At some
In embodiment, Peripheral Interface 104, processor 103 and storage control 102 can realize in one single chip.Other one
In a little examples, they can be realized by independent chip respectively.
Input-output unit 105 realizes the mutual of user and described computer for being supplied to user input data.Described
Input-output unit 105 may be, but not limited to, mouse and keyboard etc..
Fig. 2 shows the flow chart of the image Uighur word recognition methods that the embodiment of the present invention provides.Refer to figure
2, the method at least includes that step S210 is to step S240.
Step S210: obtain Uighur words image.
In the present embodiment, the concrete mode obtaining Uighur words image can be: is divided by the text image got
It is segmented into multiple line of text image, each line of text image is divided into multiple Uighur words image, thus gets dimension
The civilian word image of my that.Wherein, having multiple line of text in text image, each line of text includes multiple Uighur words.
Described text image is the image of block letter Uighur text information.For example, it is possible to utilize optical electronic equipment, such as scanning
Instrument, camera etc., import in computer 100 after block letter Uighur text information is converted into the image of correspondence and process.
Therefore, as it is shown on figure 3, before obtaining Uighur words image, the image Uygur that the embodiment of the present invention provides
Literary composition word recognition methods also includes step S201, step S202, step S203 and step S204.
Step S201: obtain text image.
Wherein, acquired text image can be to prestore block letter Uighur text information in a computer
Image, it is also possible to be to be inputted in real time by input-output unit 105.Such as, input-output unit 105 can be scanner.
Text image includes multiple line of text image, and each line of text image includes multiple Uighur words image.
It should be noted that after the text image got, in addition it is also necessary to first the text image got is carried out pretreatment.
Preprocessing process can include the Slant Rectify etc. of the binary conversion treatment of text image, denoising, image.
Concrete, the binary conversion treatment process of text image is: by gray threshold set in advance by text image
Gray level image is converted into the bianry image that gray value is 0 or 255.Wherein, gray threshold can be a fixed value, described fixing
Value can be arranged according to the empirical value of test of many times.It is of course also possible to obtain the ash of image adaptively according to Binarization methods
Degree threshold value.Such as, described Binarization methods can be big law, Two-peak method, iterative method etc..
After text image is carried out binary conversion treatment, further text image can be carried out denoising, to reject figure
The noise produced in the acquisition of picture or transmitting procedure, the follow-up process to text image.The denoising method generally used,
I.e. filtering method has medium filtering, mean filter, mathematical morphology filter etc..
It addition, during acquisition text image, the input such as placement reason or scanner being likely to be due to text information sets
Standby reason causes the text image got to there is certain inclination, and then affects follow-up image segmentation.And hence it is also possible to
Text image after binaryzation, denoising is carried out Slant Rectify.In the present embodiment, it is possible to use conventional Slant Rectify is calculated
Method obtains incline direction and the angle of inclination of text image, thus further text image is carried out Slant Rectify.At present, conventional
Text Slant Correction Algorithm include: projection properties method, Hough transform method, cross-correlation technique, fourier transform method, nearest neighbour method
Deng.
Step S202: be multiple line of text images by Document Segmentation.
Uighur is to write by running hand from right to left.In the present embodiment, text image have brighter between adjacent line of text
Aobvious blank.Therefore, it can the horizontal projective histogram by two dimensional image and realize the row cutting of text image.
Concrete, floor projection can be carried out to completing pretreated text image, between adjacent line of text
Blank and all line of text floor projection value obtains the upper and lower border of each line of text in text image, according to acquired
To the upper and lower border of each line of text can be i.e. multiple line of text images by Document Segmentation.Wherein, floor projection
It is that the pixel distribution according to image forms horizontal projective histogram to the summation of every a line pixel value respectively.
Step S203: each line of text image is divided into multiple Uighur words image.
Between adjacent Uighur words in each line of text image, also there is obvious blank.Therefore, it can
The cutting of Uighur words image in line of text image is realized by the vertical projective histogram of two dimensional image.
Concrete, on the basis of the line of text cutting in realizing text image, can be further to each line of text
Image carries out the segmentation of words.For example, it is possible to respectively each line of text image is carried out upright projection.According to current text row image
In the upright projection value of blank between adjacent Uighur words and current text row can obtain in current text row image
The left and right border of each Uighur words.According to each Uygur in each accessed line of text image
Each line of text image i.e. can be divided into multiple Uighur words image by the left and right border of literary composition word.Wherein, vertically
Projection is that the pixel distribution according to image forms vertical projective histogram to the summation of every string pixel value respectively.
Such as, according to said method, the text image shown in Fig. 4 is carried out segmentation and can obtain Uygur as shown in Figure 5
Literary composition word image.
Step S204: obtain the baseline territory of each line of text image, using the baseline territory of current text row image as working as
The baseline territory of the multiple Uighur words images included by front line of text image.
Either in block letter Uighur or handwritten form Uighur, letter is connected along a certain axis, should
Axis is referred to as baseline.Under normal circumstances, affected by image resolution ratio, the baseline of disjunctor section in Uighur words image
Width is more than a pixel, and now, baseline is also referred to as baseline territory.Such as, as shown in Figure 6, the straight line a in figure represents this Uygur
The coboundary in the baseline territory of the Uighur words in literary composition word image, straight line b represents the baseline territory of this Uighur words
Lower boundary, the region between straight line a and straight line b is baseline territory.
Owing to Uighur words image is to be formed by corresponding line of text image segmentation, it is to be understood that each
The baseline territory that Uighur words image is corresponding is the baseline territory of this line of text image belonging to Uighur words image.
In the present embodiment, as it is shown in fig. 7, in step S204, the method in the baseline territory obtaining each line of text image can
To include step S301, step S302, step S303 and S304.
Step S301: obtain the profile of current text row image.
Concrete, it is possible to use edge detection algorithm carries out rim detection and extracts current text row current text row image
The profile of image.Such as, conventional edge detection operator has Sobel operator, Canny operator, Roberts operator, Prewitt to calculate
Son etc..
Step S302: according to the first preset rules the profile of current text row image carried out straight-line detection obtain a plurality of directly
Line.
In the present embodiment, the first preset rules can be Hough line detection algorithm, it is of course also possible to use other permissible
For detecting the algorithm of image outline cathetus.
Step S303: search length in described a plurality of straight line and be more than or equal to the straight line of pre-set length threshold, according to being looked into
The coordinate position of all straight lines found obtains datum line.
Wherein, pre-set length threshold can determine according to test of many times, can prestore in a computer, it is also possible to real
Time by input-output unit, such as input through keyboard.Concrete, the length of the every straight line that step S302 can be obtained
Compare with pre-set length threshold, obtain the length straight line more than or equal to pre-set length threshold, build the first straight line collection.According to
One straight line concentrate pixel coordinate included by every straight line can obtain being positioned at the top on longitudinal direction the first straight line and
Second straight line of bottom.Can obtain according to the pixel coordinate included by the first straight line and the pixel coordinate included by the second straight line
Obtain the first straight line and the centrage of the second straight line, i.e. datum line.
Step S304: obtain in the straight line found according to the second preset rules and be positioned at the longest one above datum line
Bar straight line, as the coboundary in baseline territory, obtains the longest straight line work being positioned in the straight line found below datum line
Lower boundary for baseline territory.
Search the first straight line to be centrally located at all straight lines above datum line and build the second straight line collection, compare the second straight line collection
In the length of all straight lines obtain the length the longest straight line coboundary as baseline territory.Search the second straight line and be centrally located at benchmark
All straight lines below line build the 3rd straight line collection, compare the 3rd straight line and concentrate the length of all straight lines to obtain the longest straight of length
Line is as the lower boundary in baseline territory.Obtained coboundary and the direct region of lower boundary are baseline territory.Certainly, except above-mentioned
Outside mode, it would however also be possible to employ other modes obtain coboundary and the lower boundary in baseline territory.
It addition, according to the writing characteristics of Uighur, in line of text image, the pixel major part of character zone is concentrated
It is distributed in baseline territory.Therefore, the method in the baseline territory obtaining each line of text image can also be: first obtains line of text figure
The profile of picture;Profile to line of text image carries out floor projection to obtain the first pixel column and the second pixel column the most again.Its
In, the region between the first pixel column and the second pixel column is baseline territory.
Concrete, the concrete mode obtaining the first pixel column and the second pixel column can be:
The profile of line of text image is carried out floor projection, obtains the projection value that in image, every one-row pixels is corresponding.According to
The projection value of all pixel columns arranges second preset value so that the only i-th row pixel is to the i-th+m row pixel and the i-th+n row
Pixel is more than or equal to this second preset value to the projection value of the i-th+p row pixel.Wherein, i, m, n, p are positive integer, and m < n <
p.Hereafter, search i-th and walk to the pixel column that in the i-th+m row, projection value is maximum, as the first pixel column, search the i-th+n and walk to i-th
The pixel column that in+p row, projection value is maximum is as the second pixel column.Region between first pixel column and the second pixel column is this
The baseline territory of line of text image.Wherein, the coboundary in the first pixel behavior baseline territory, the second pixel behavior baseline territory following
Boundary.
Respectively by each the line of text image current text row image the most obtained by step S202, perform step S301
The baseline territory of each line of text image i.e. can be obtained to step S304.
Step S220: according to each by described Uighur words image of baseline territory corresponding to Uighur words image
Individual disjunctor section is divided into one or more parts.
The word of Uighur is made up of one or more letters, and Uighur has the feature of write the two or more syllables of a word together, these words
Mother front and back may be connected to form one or more connected character field, i.e. disjunctor section.Each Uighur words by one or
Multiple disjunctor sections are constituted, and have gap between adjacent disjunctor section.Such as, as shown in Figure 6, the Uighur words in Fig. 6 has three
Individual disjunctor section, the character in each dotted rectangle is a disjunctor section, as can be seen from Figure, has between two disjunctor sections
There is gap.Concrete, the method for the disjunctor section obtaining Uighur words image can be:
Uygur's word image is binary image, including character zone and background area.Assume the pixel of character zone
Point is black, and background area pixels point is white, and now, Uighur words image appearance is white gravoply, with black engraved characters.By character zone
Pixel be expressed as 1, the pixel of background area is expressed as 0.Described Uighur words image is carried out upright projection,
According to upright projection, described Uighur words image is carried out segmentation and obtain the disjunctor section in Uighur words image.
Concrete, in Uighur words image, the upright projection value of gap location pixel between adjacent disjunctor section is less than
First preset value.Wherein, the theoretical value 0 of described first preset value, due to noise that may be present in Uygur's word image,
The empirical value that the first concrete preset value can be obtained by test of many times.By the upright projection result of Uighur words image
Compare with the first preset value, it is possible to obtain the left and right border of each disjunctor section in this Uighur words image, according to being obtained
This Uighur words image cutting can be i.e. one or more disjunctor sections by the left and right border obtained.
Concrete, as shown in Figure 8, according to baseline territory corresponding to Uighur words image by Uighur words image
Each disjunctor section is divided into the method for one or more parts can include step S221 and step S222.
Step S221: do vertical throwing to being positioned at the pixel beyond baseline territory in the current disjunctor section of Uighur words image
Shadow obtains the projection peak value of one or more separation.
Coboundary according to baseline territory corresponding to Uighur words image and lower boundary, reject in current disjunctor section and be positioned at
Pixel in baseline territory, the value that i.e. will be located in the pixel in baseline territory sets to 0.Hereafter, then rejecting is positioned at the picture in baseline territory
Current disjunctor section after vegetarian refreshments carries out upright projection, obtains the projection peak value of one or more separation.
Step S222: according to described projection peak value, current disjunctor section is carried out segmentation and obtain one or more parts.
It is positioned at the pixel beyond described baseline territory in current disjunctor section to do upright projection and obtain the projection of multiple separation
During peak value, obtain the point of contact as current disjunctor section, the midpoint between adjacent two projection peak values separated.Further, according to institute
Described current disjunctor section is divided into multiple parts by the point of contact got.For example, it is possible to using the pixel column at place, point of contact as cutting
Divide pixel column, when obtaining the projection peak value that two separate, it is possible to obtain a cutting pixel column, can according to this cutting pixel column
So that current disjunctor section to be divided into two parts, the right margin of i.e. current disjunctor section to the part between described cutting pixel column is
One parts, the part between the left margin of described cutting pixel column to current disjunctor section is another parts.In like manner, when obtaining
During three projection peak values separated, it is possible to obtain two cutting pixel columns, at this point it is possible to current disjunctor section is divided into three portions
Part.
It is positioned at after the pixel beyond described baseline territory does upright projection in current disjunctor section and only obtains a projection peak
During value, represent current disjunctor section without cutting, i.e. this disjunctor section without point of contact exist.Now, current disjunctor section is the independence of letter
Form.It is to say, current disjunctor section is parts, and the form of these parts is absolute version.
Respectively to each disjunctor section execution step S221 in current Uighur word image to step S222, will dimension
Each disjunctor section cutting in my your literary composition word image is one or more parts, is namely divided by Uighur words image
It is segmented into one or more parts.Such as, as it is shown in figure 9, to each the disjunctor section in the Uighur words image shown in Fig. 6
Execution step S221, to step S222, can obtain parts v1, parts v2, parts v3, parts v4, parts v5With parts v6, in Fig. 7
Each solid-line rectangle frame represent parts.
Step S230: obtain the characteristic information of each parts of Uighur words image to obtain Uighur words figure
The characteristic vector of picture.
Wherein, described characteristic information includes global characteristics information and local characteristic information.Global characteristics information includes laterally
Position feature information and lengthwise position characteristic information, local feature information includes the number of contours of each parts and hole number and every
The lengthwise position relation in the optional feature included by individual parts and baseline territory, i.e. optional feature be positioned at baseline territory top or under
Side.Wherein, the attached stroke during optional feature is Uighur letter.
Concrete, as shown in Figure 8, step S230 includes step S231, step S232 and step S233.
Step S231: obtain the parts included by Uighur words image successively according to preset order.
Have the feature write from right to left by running hand in view of Uighur, therefore, described preset order is preferably from the right side
To a left side.
Step S232: obtain the characteristic information of each parts.
Concrete, the mode of the lateral attitude characteristic information of obtaining widget is: judge that the disjunctor section belonging to current part is
No there is point of contact, in the presence of the disjunctor section belonging to current part does not has point of contact, it is determined that current part is individual components.When currently
When disjunctor section belonging to parts exists point of contact, according to quantity and the current part lateral attitude in described disjunctor section at point of contact
Judge the form of current part.When the point of contact quantity 1 of the disjunctor section belonging to current part, represent the company belonging to current part
Body section includes two parts.Now, if current part is the parts relatively kept right by rules for writing in two parts, then judge to work as
Forepiece is initial part, if current part is the parts relatively kept left by rules for writing in two parts, then judges current
Parts are ending parts.When the point of contact quantity of the disjunctor section belonging to current part is more than 1, represent the disjunctor belonging to current part
Section includes three or the parts of more than three.Now, if by rules for writing in the parts that current part is three or more than three
The parts kept right most, then judge current part as initial part, if by book in the parts that current part is three or more than three
Write the parts that rule keeps left most, then judges current part as ending parts, when above-mentioned two condition is all unsatisfactory for, then judging ought
Forepiece is intermediate member.
Such as, as it is shown in figure 9, in the Uighur words image shown in Fig. 5, parts v1With parts v3For initial part, portion
Part v2With parts v5For ending parts, parts v4For intermediate member, parts v6For individual components.
The lateral attitude characteristic information of parts is for reflecting the form of parts.For example, it is possible to by the horizontal position of individual components
Put characteristic information and be set to 0, the lateral attitude characteristic information of initial part is set to 1, the lateral attitude of intermediate member is special
Reference breath is set to 2, and the lateral attitude characteristic information of ending parts is set to 3.
The mode of the lengthwise position characteristic information of obtaining widget is:
Obtain the boundary rectangle of current part, according to the boundary rectangle of current part in Uighur words image longitudinal direction side
The lengthwise position characteristic information of position acquisition current part upwards.
Concrete, Uighur words image can be divided into three regions along the longitudinal direction, be followed successively by from top to bottom
Top area, zone line and lower region.Such as, the concrete division methods of top area, zone line and lower region can
Think: the zone leveling between coboundary and the lower boundary of Uighur words image is divided into trisection, wherein, Uygur
The coboundary of literary composition word image is the first row pixel of this Uighur words image, the lower boundary of Uighur words image
It is last column pixel of this Uighur words image.
Now, when the boundary rectangle of current part is only located at zone line, can be special by the lengthwise position of current part
Reference breath is set to 0;When the boundary rectangle of current part is only located at top area and zone line, can be by current part
Lengthwise position characteristic information is set to 1;When the boundary rectangle of current part is only located at zone line and lower region, can be by
The lengthwise position characteristic information of current part is set to 2;When the boundary rectangle of current part be positioned at top area, zone line and
During lower region, the lengthwise position characteristic information of current part can be set to 3.
Such as, as shown in Figure 10, by the region between coboundary and the lower boundary of the Uighur words image shown in Fig. 5
Average mark is segmented into trisection, and wherein dotted line c represents that the coboundary of this Uighur words image, dotted line f represent this Uighur
The lower boundary of word image, dotted line d and dotted line d represent cut-off rule.Therefore, parts v1Lengthwise position characteristic information be 2;Parts
v2Lengthwise position characteristic information be 3;Parts v3Lengthwise position characteristic information be 1;Parts v4Lengthwise position characteristic information be
1;Parts v5Lengthwise position characteristic information be 3;Parts v6Lengthwise position characteristic information be 2.
Further, the mode of the local feature information of obtaining widget is: obtain the outlines of current part;Obtain and work as
The pore quantity of forepiece;Search the optional feature included by current part, and judge found optional feature and baseline
The lengthwise position relation in territory;Number of contours according to accessed current part, hole number and additional included by current part
The lengthwise position relation in parts and baseline territory determines the local feature information of current part.
Concrete, the concrete mode of the outlines obtaining current part can be: to the connection included by current part
Region is marked, the quantity of acquisition connected region included by current part.The number of the connected region included by current part
Amount is the outlines of current part.And specifically can using of the pore quantity of current part penetrates time counting method acquisition,
Can also be obtained by the connected region number obtaining current part.The pixel quantity included due to usual optional feature is relatively
Few, therefore, the embodiment searching the optional feature included by current part in the present embodiment can be by: is wrapped current part
After the connected region included is marked, the pixel number of each connected region is compared with the 3rd preset value, when existing even
When the pixel number in logical region is less than or equal to three preset values, then judge this connected region as optional feature, when all companies
When the pixel number in logical region is all higher than three preset values, then judge that current part does not has optional feature.Such as, such as Figure 10 institute
Show, parts v2Including an accessory components, and this optional feature is above baseline territory.
Step S233: according to the characteristic information of acquired each parts build the feature of Uighur words image to
Amount.
Based on said method, after getting the characteristic information of each parts in Uighur words image, Ke Yigen
According to the characteristic information of all parts included by this Uighur words image build the feature of this Uighur words image to
Amount.
Such as, T represents the lateral attitude characteristic information of parts, and P represents the lengthwise position characteristic information of parts, C expressed portion
The number of contours of part, H represents the hole number of parts, A represent the optional feature included by parts above baseline territory, B expressed portion
Optional feature included by part is in the lower section in baseline territory.Wherein, the local feature information of C, H, A and B coexpress parts.T's
Value can be 0,1,2 or 3, the T=0 when parts are individual components, and when parts are initial part, T=1, in parts are
Between parts time, T=2, when parts for ending parts time, T=3.The value of P can be 0,1,2 or 3, when the boundary rectangle of parts
When being only located at zone line, P=0;When the boundary rectangle of parts is only located at top area and zone line, P=1;Work as parts
Boundary rectangle when being only located at zone line and lower region, P=2;When the boundary rectangle of parts is positioned at top area, mesozone
When territory and lower region, P=3.When the optional feature included by parts is above baseline territory, A=1, B=0;When parts institute
Including optional feature when the lower section in baseline territory, A=0, B=1, when parts do not have optional feature, A=0, B=0.
Therefore, as the Uighur words image V={v got in step S231k| 1≤k≤M, k are positive integer }, its
In, vkRepresenting the kth parts in this Uighur words image, M is component count included in Uighur words image
Amount.Any one parts vkCharacteristic information be expressed as (TPCHAB)k, then the feature of this Uighur words image can be obtained
Vector S=((TPCHAB)1, (TPCHAB)2..., (TPCHAB)M)。
Step S240: the characteristic vector of Uighur words image contrasted with the feature lexicon preset, to obtain
The textual words that the characteristic vector of acquired Uighur words image is corresponding.
Wherein, the building mode of the feature lexicon preset is: obtain textual words, substitutes Uighur list by textual words
Word image, obtains the characteristic vector of textual words according to step S220 in the present embodiment to step S230.Wherein, textual words is
Computer 100 can the Uighur words of Direct Recognition character information.It can be the input-output unit by computer 100
The textual words of 105 such as input through keyboard, it is also possible to be the textual words prestored in computer 100.Therefore, it can basis
The characteristic vector construction feature dictionary of the most different textual words.Wherein, described feature lexicon include described characteristic vector with
The corresponding relation of described textual words.For example, it is possible to Uighur words inputs different more than will arrange 2.8 ten thousand or
Form textual words storehouse in a computer by the storage of other means, obtain each literary composition in textual words storehouse according to said method
The characteristic vector of this word, further according to the characteristic vector construction feature dictionary of obtained textual words, and feature lexicon
In the corresponding textual words of each characteristic vector.
The embodiment of the present invention is directly according to the characteristic vector construction feature dictionary of textual words, compared to existing by sweeping
Retouch image pattern and carry out the sample training method with acquisition characteristic vector with the corresponding relation of textual words, eliminate cost a large amount of
Man power and material is scanned the gatherer process of image pattern, effectively simplifies the building process of feature lexicon.
Further, using the characteristic vector of Uighur words image obtained in step S230 as object to be identified.
Object to be identified is contrasted with the characteristic vector in feature lexicon, a certain feature in object to be identified with feature lexicon
When the comparing result of vector meets pre-conditioned, it is right that the textual words corresponding to Uighur words image is this feature vector
The textual words answered, i.e. achieves the identification of Uighur words in above-mentioned Uighur words image.
It should be noted that in object to be identified and feature lexicon in the comparison process of characteristic vector, by characteristic vector
Included global characteristics information as invariant feature information, using local feature information included in characteristic vector as non-surely
Determine characteristic information.The above-mentioned pre-conditioned matching degree including invariant feature information and the matching degree of astable characteristic information.
Such as, above-mentioned pre-conditioned can be: in characteristic vector, the matching degree of the global characteristics information of each parts is 100%,
In local feature information, the difference of outlines is less than or equal to 1, and the difference of pore quantity is less than or equal to 1, the longitudinal direction of optional feature
The matching degree of position is 100%.
Preferably, in order to simplify the amount of calculation of comparison process, in the present embodiment, can be previously according to each in feature lexicon
Feature lexicon is divided into multiple subclass by the disjunctor hop count amount included by the textual words that individual characteristic vector is corresponding, and sets up corresponding rope
Draw table.When Uighur words image is identified, according to the disjunctor hop count amount included by Uighur words image, pass through
Index finds corresponding subclass, is contrasted with the characteristic vector in corresponding subclass by object to be identified.
It addition, the embodiment of the present invention additionally provides a kind of image Uighur Word identifier, as shown in figure 11, described
Image Uighur Word identifier 200 includes that word image acquisition module 210, parts segmentation module 220, characteristic vector obtain
Delivery block 230 and identification module 240.
Wherein, word image acquisition module 210 is used for obtaining Uighur words image, described Uighur words image
Including one or more disjunctor sections.Parts segmentation module 220 is for the baseline territory corresponding according to described Uighur words image
Each disjunctor section of described Uighur words image is divided into one or more parts.Characteristic vector acquisition module 230
Described Uighur words image is obtained for obtaining the characteristic information of each described parts of described Uighur words image
Characteristic vector.Identification module 240 is for entering the characteristic vector of described Uighur words image with the feature lexicon preset
Row contrast, with the textual words that the characteristic vector of the described Uighur words image acquired in acquisition is corresponding, wherein, described spy
Levy dictionary and include the corresponding relation of described characteristic vector and the described textual words obtained according to textual words.
Concrete, as shown in figure 12, the embodiment of the present invention additionally provides a kind of image Uighur Word identifier also
Including: text image acquisition module 201, Document Segmentation module 202, line of text image segmentation module 203 and baseline territory are obtained
Delivery block 204.
Text image acquisition module 201 is used for obtaining text image, and described text image includes multiple line of text image, often
One line of text image includes multiple Uighur words image.Document Segmentation module 202 is for by described text image
It is divided into multiple line of text image.Line of text image segmentation module 203 is for being divided into multiple dimension by each line of text image
The civilian word image of my that.Baseline territory acquisition module 204 is for obtaining the baseline territory of each line of text image, by current text row
The baseline territory of image is as the baseline territory of the multiple Uighur words images corresponding to described current text row image.
Concrete, as shown in figure 13, parts segmentation module 220 includes: projection peak value acquiring unit 221 and cutting unit
222.Wherein, projection peak value acquiring unit 221 is for described to being positioned in the current disjunctor section of described Uighur words image
Pixel beyond baseline territory is done upright projection and is obtained the projection peak value of one or more separation.Cutting unit 222 is for according to institute
State projection peak value described current disjunctor section is carried out segmentation to obtain one or more parts.
Concrete, as Figure 13 shows, described characteristic vector acquisition module 230 includes that component retrieval unit 231, characteristic information obtain
Take unit 232 and characteristic vector construction unit 233.Wherein, component retrieval unit 231 for obtaining institute successively according to preset order
State the parts included by Uighur words image.Characteristic acquisition unit 232 is for obtaining the feature of each described parts
Information.Characteristic vector construction unit 233 builds described Uygur for the characteristic information according to acquired each described parts
The characteristic vector of literary composition word image.
In sum, the image Uighur word recognition methods of embodiment of the present invention offer and device, by tieing up me
Each disjunctor section in your literary composition word image is divided into one or more parts, and obtains the characteristic information of each parts, thus
Characteristic information according to parts each in Uighur words image builds the characteristic vector of above-mentioned Uighur words image, enters
The characteristic vector of Uighur words image as object to be identified and the feature lexicon contrast preset, is obtained above-mentioned dimension by one step
The textual words that my your literary composition word image is corresponding.In existing recognition methods, the image Uighur that the embodiment of the present invention provides
Word recognition methods and device are to be one or more parts by Uighur words image cutting, it is not necessary to be syncopated as dimension exactly
Each letter in my your literary composition word image, reduces the cutting difficulty of Uighur words image.Further, acquired in utilizing
To Uighur words image in the characteristic information of parts build the characteristic vector of this Uighur words image as treating
Identify object, be effectively improved the recognizable rate of Uighur words image.
It addition, the image Uighur word recognition methods of embodiment of the present invention offer and device are directly according to the spy of word
Levy vector construction feature dictionary, carry out sample training to obtain characteristic vector and literary composition compared to existing by scanogram sample
The method of the corresponding relation of this word, eliminates the gatherer process spending a large amount of man power and materials to be scanned image pattern, has
Simplify to effect the building process of feature lexicon.
In embodiment provided herein, it should be understood that disclosed apparatus and method, it is also possible to by other
Mode realize.Device embodiment described above is only that schematically such as, flow chart and block diagram in accompanying drawing show
The device of multiple embodiments according to the present invention, the architectural framework in the cards of method and computer program product, function
And operation.In this, each square frame in flow chart or block diagram can represent of a module, program segment or code
Point, a part for described module, program segment or code comprises performing of one or more logic function for realizing regulation
Instruction.It should also be noted that at some as in the implementation replaced, the function marked in square frame can also be attached to be different from
The order marked in figure occurs.Such as, two continuous print square frames can essentially perform substantially in parallel, and they the most also may be used
To perform in the opposite order, this is depending on involved function.It is also noted that each in block diagram and/or flow chart
The combination of the square frame in square frame and block diagram and/or flow chart, can with perform the function of regulation or the special of action based on
The system of hardware realizes, or can realize with the combination of specialized hardware with computer instruction.
It addition, each functional module in each embodiment of the present invention can integrate one independent portion of formation
Point, it is also possible to it is modules individualism, it is also possible to two or more modules are integrated to form an independent part.
If described function is using the form realization of software function module and as independent production marketing or use, permissible
It is stored in a computer read/write memory medium.Based on such understanding, technical scheme is the most in other words
The part contributing prior art or the part of this technical scheme can embody with the form of software product, this meter
Calculation machine software product is stored in a storage medium, including some instructions with so that a computer equipment (can be individual
People's computer, server, or the network equipment etc.) perform all or part of step of method described in each embodiment of the present invention.
And aforesaid storage medium includes: USB flash disk, portable hard drive, read only memory (ROM, Read-Only Memory), random access memory are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic disc or CD.Need
Illustrate, in this article, the relational terms of such as first and second or the like be used merely to by an entity or operation with
Another entity or operating space separate, and there is any this reality between not necessarily requiring or imply these entities or operating
The relation on border or order.And, term " includes ", " comprising " or its any other variant are intended to the bag of nonexcludability
Contain, so that include that the process of a series of key element, method, article or equipment not only include those key elements, but also include
Other key elements being not expressly set out, or also include the key element intrinsic for this process, method, article or equipment.
In the case of there is no more restriction, statement " including ... " key element limited, it is not excluded that including described key element
Process, method, article or equipment in there is also other identical element.
The foregoing is only the preferred embodiments of the present invention, be not limited to the present invention, for the skill of this area
For art personnel, the present invention can have various modifications and variations.All within the spirit and principles in the present invention, that is made any repaiies
Change, equivalent, improvement etc., should be included within the scope of the present invention.
Claims (10)
1. an image Uighur word recognition methods, it is characterised in that described method includes:
Obtaining Uighur words image, described Uighur words image includes one or more disjunctor section;
According to baseline territory corresponding to described Uighur words image by each disjunctor section of described Uighur words image
It is divided into one or more parts;
The characteristic information of each described parts obtaining described Uighur words image obtains described Uighur words image
Characteristic vector;
The characteristic vector of described Uighur words image is contrasted with the feature lexicon preset, with the institute acquired in acquisition
Stating the textual words that the characteristic vector of Uighur words image is corresponding, wherein, described feature lexicon includes according to textual words
The described characteristic vector obtained and the corresponding relation of described textual words.
Method the most according to claim 1, it is characterised in that before the step of described acquisition Uighur words image,
Also include:
Obtain text image, described text image includes multiple line of text image, each line of text image include multiple dimension I
That literary composition word image;
It is multiple line of text images by described Document Segmentation;
Each line of text image is divided into multiple Uighur words image;
Obtain the baseline territory of each line of text image, using the baseline territory of current text row image as described current text row figure
Baseline territory as corresponding multiple Uighur words images.
Method the most according to claim 2, it is characterised in that the step in the baseline territory of each line of text image of described acquisition
Suddenly, including:
Obtain the profile of current text row image;
According to the first preset rules, the profile of described current text row image is carried out straight-line detection and obtain a plurality of straight line;
Search length in described a plurality of straight line and be more than or equal to the straight line of pre-set length threshold, according to all straight lines found
Coordinate position obtain datum line;
The longest straight line work being positioned in the straight line found above described datum line is obtained according to the second preset rules
For the coboundary in baseline territory, obtain in the straight line found and be positioned at the longest straight line below described datum line as base
The lower boundary in line territory.
Method the most according to claim 1, it is characterised in that the described base corresponding according to described Uighur words image
Each disjunctor section of described Uighur words image is divided into the step of one or more parts by line territory, including:
Do upright projection obtain the current disjunctor section of described Uighur words image is positioned at the pixel beyond described baseline territory
Projection peak value to one or more separation;
According to described projection peak value, described current disjunctor section is carried out segmentation and obtain one or more parts.
Method the most according to claim 4, it is characterised in that described according to described projection peak value to described current disjunctor section
Carry out splitting the step obtaining one or more parts, including:
It is positioned at the pixel beyond described baseline territory in current disjunctor section to do upright projection and obtain the projection peak value of multiple separation
Time, obtain the point of contact as described current disjunctor section, the midpoint between adjacent two described projection peak values separated;
According to accessed point of contact, described current disjunctor section is divided into multiple parts.
Method the most according to claim 1, it is characterised in that each institute of described acquisition described Uighur words image
State the step that the characteristic information of parts obtains the characteristic vector of described Uighur words image, including:
The parts included by described Uighur words image are obtained successively according to preset order;
Obtain the characteristic information of each described parts;
Characteristic information according to acquired each described parts builds the characteristic vector of described Uighur words image.
7. an image Uighur Word identifier, it is characterised in that including:
Word image acquisition module, is used for obtaining Uighur words image, described Uighur words image include one or
Multiple disjunctor sections;
Parts segmentation module, for the baseline territory corresponding according to described Uighur words image by described Uighur words figure
Each disjunctor section of picture is divided into one or more parts;
Characteristic vector acquisition module, the characteristic information of each described parts for obtaining described Uighur words image obtains
The characteristic vector of described Uighur words image;
Identification module, for the characteristic vector of described Uighur words image is contrasted with the feature lexicon preset, with
The textual words that the characteristic vector of the described Uighur words image acquired in acquisition is corresponding, wherein, described feature lexicon bag
Include the corresponding relation of described characteristic vector and the described textual words obtained according to textual words.
Device the most according to claim 7, it is characterised in that also include:
Text image acquisition module, is used for obtaining text image, and described text image includes multiple line of text image, each literary composition
One's own profession image includes multiple Uighur words image;
Document Segmentation module, being used for described Document Segmentation is multiple line of text images;
Line of text image segmentation module, for being divided into multiple Uighur words image by each line of text image;
Baseline territory acquisition module, for obtaining the baseline territory of each line of text image, by the baseline territory of current text row image
Baseline territory as the multiple Uighur words images corresponding to described current text row image.
Device the most according to claim 7, it is characterised in that described parts segmentation module includes:
Projection peak value acquiring unit, for the current disjunctor section of described Uighur words image is positioned at described baseline territory with
Outer pixel is done upright projection and is obtained the projection peak value of one or more separation;
Cutting unit, obtains one or more parts for described current disjunctor section being carried out segmentation according to described projection peak value.
Device the most according to claim 7, it is characterised in that described characteristic vector acquisition module includes:
Component retrieval unit, for obtaining the parts included by described Uighur words image successively according to preset order;
Characteristic acquisition unit, for obtaining the characteristic information of each described parts;
Characteristic vector construction unit, builds described Uighur list for the characteristic information according to acquired each described parts
The characteristic vector of word image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610609772.4A CN106295631A (en) | 2016-07-27 | 2016-07-27 | A kind of image Uighur word recognition methods and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610609772.4A CN106295631A (en) | 2016-07-27 | 2016-07-27 | A kind of image Uighur word recognition methods and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106295631A true CN106295631A (en) | 2017-01-04 |
Family
ID=57663089
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610609772.4A Pending CN106295631A (en) | 2016-07-27 | 2016-07-27 | A kind of image Uighur word recognition methods and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106295631A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106372639A (en) * | 2016-08-19 | 2017-02-01 | 西安电子科技大学 | Morphology and integral projection-based printed Uygur document segmentation method |
CN108090489A (en) * | 2018-01-15 | 2018-05-29 | 兰州理工大学 | Offline handwriting Balakrishnan word recognition methods of the computer based according to grapheme segmentation |
CN109800756A (en) * | 2018-12-14 | 2019-05-24 | 华南理工大学 | A kind of text detection recognition methods for the intensive text of Chinese historical document |
CN110532571A (en) * | 2017-09-12 | 2019-12-03 | 腾讯科技(深圳)有限公司 | Text handling method and relevant apparatus |
CN111553336A (en) * | 2020-04-27 | 2020-08-18 | 西安电子科技大学 | Print Uyghur document image recognition system and method based on link segment |
CN111626302A (en) * | 2020-05-25 | 2020-09-04 | 西北民族大学 | Method and system for cutting adhered text lines of ancient book document images of Ujin Tibetan |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1741035A (en) * | 2005-09-23 | 2006-03-01 | 清华大学 | Blocks letter Arabic character set text dividing method |
CN101866417A (en) * | 2010-06-18 | 2010-10-20 | 西安电子科技大学 | Method for identifying handwritten Uigur characters |
CN102446275A (en) * | 2010-09-30 | 2012-05-09 | 汉王科技股份有限公司 | Identification method and device for Arabic character |
CN104899601A (en) * | 2015-05-29 | 2015-09-09 | 西安电子科技大学宁波信息技术研究院 | Identification method of handwritten Uyghur words |
-
2016
- 2016-07-27 CN CN201610609772.4A patent/CN106295631A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1741035A (en) * | 2005-09-23 | 2006-03-01 | 清华大学 | Blocks letter Arabic character set text dividing method |
CN101866417A (en) * | 2010-06-18 | 2010-10-20 | 西安电子科技大学 | Method for identifying handwritten Uigur characters |
CN102446275A (en) * | 2010-09-30 | 2012-05-09 | 汉王科技股份有限公司 | Identification method and device for Arabic character |
CN104899601A (en) * | 2015-05-29 | 2015-09-09 | 西安电子科技大学宁波信息技术研究院 | Identification method of handwritten Uyghur words |
Non-Patent Citations (1)
Title |
---|
哈力木拉提: "基于轮廓的维吾尔文切分", 《第十一届全国民族语言文字信息学术研讨会》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106372639A (en) * | 2016-08-19 | 2017-02-01 | 西安电子科技大学 | Morphology and integral projection-based printed Uygur document segmentation method |
CN106372639B (en) * | 2016-08-19 | 2019-03-08 | 西安电子科技大学 | Block letter Uighur document cutting method based on morphology and integral projection |
CN110532571A (en) * | 2017-09-12 | 2019-12-03 | 腾讯科技(深圳)有限公司 | Text handling method and relevant apparatus |
CN110532571B (en) * | 2017-09-12 | 2022-11-18 | 腾讯科技(深圳)有限公司 | Text processing method and related device |
CN108090489A (en) * | 2018-01-15 | 2018-05-29 | 兰州理工大学 | Offline handwriting Balakrishnan word recognition methods of the computer based according to grapheme segmentation |
CN108090489B (en) * | 2018-01-15 | 2021-06-29 | 兰州理工大学 | Off-line hand-written Uygur word recognition method based on grapheme segmentation based on computer |
CN109800756A (en) * | 2018-12-14 | 2019-05-24 | 华南理工大学 | A kind of text detection recognition methods for the intensive text of Chinese historical document |
CN109800756B (en) * | 2018-12-14 | 2021-02-12 | 华南理工大学 | Character detection and identification method for dense text of Chinese historical literature |
CN111553336A (en) * | 2020-04-27 | 2020-08-18 | 西安电子科技大学 | Print Uyghur document image recognition system and method based on link segment |
CN111553336B (en) * | 2020-04-27 | 2023-03-24 | 西安电子科技大学 | Print Uyghur document image recognition system and method based on link segment |
CN111626302A (en) * | 2020-05-25 | 2020-09-04 | 西北民族大学 | Method and system for cutting adhered text lines of ancient book document images of Ujin Tibetan |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106295631A (en) | A kind of image Uighur word recognition methods and device | |
Alaei et al. | A new scheme for unconstrained handwritten text-line segmentation | |
Kumar et al. | Segmentation of isolated and touching characters in offline handwritten Gurmukhi script recognition | |
Ptak et al. | Projection–based text line segmentation with a variable threshold | |
Shivakumara et al. | Accurate video text detection through classification of low and high contrast images | |
CN104484643B (en) | The intelligent identification Method and system of a kind of handwriting table | |
US10643094B2 (en) | Method for line and word segmentation for handwritten text images | |
Fabrizio et al. | Text detection in street level images | |
Elnagar et al. | Segmentation of connected handwritten numeral strings | |
CN101122953A (en) | Picture words segmentation method | |
Kesiman et al. | A new scheme for text line and character segmentation from gray scale images of palm leaf manuscript | |
CN108154132A (en) | A kind of identity card text extraction method, system and equipment and storage medium | |
Kumar et al. | Multi-script robust reading competition in ICDAR 2013 | |
Al-Dmour et al. | Segmenting Arabic handwritten documents into text lines and words | |
Cohen et al. | Using scale-space anisotropic smoothing for text line extraction in historical documents | |
CN111507351A (en) | Ancient book document digitalization method | |
Kesiman et al. | Southeast Asian palm leaf manuscript images: a review of handwritten text line segmentation methods and new challenges | |
Ayesh et al. | A robust line segmentation algorithm for Arabic printed text with diacritics | |
Xiao et al. | Knowledge-based English cursive script segmentation | |
Suleyman et al. | An adaptive threshold algorithm for offline Uyghur handwritten text line segmentation | |
CN112418199B (en) | Multi-modal information extraction method and device, electronic equipment and storage medium | |
Bag et al. | Character segmentation of handwritten Bangla text by vertex characterization of isothetic covers | |
Liu et al. | Real-time scene text detection based on stroke model | |
CN108171144B (en) | Information processing method, information processing device, electronic equipment and storage medium | |
Li et al. | A text-line segmentation method for historical Tibetan documents based on baseline detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170104 |