CN109919155A - A kind of the inclination angle antidote and terminal of text image - Google Patents
A kind of the inclination angle antidote and terminal of text image Download PDFInfo
- Publication number
- CN109919155A CN109919155A CN201910189109.7A CN201910189109A CN109919155A CN 109919155 A CN109919155 A CN 109919155A CN 201910189109 A CN201910189109 A CN 201910189109A CN 109919155 A CN109919155 A CN 109919155A
- Authority
- CN
- China
- Prior art keywords
- text image
- straight line
- connected domain
- inclination angle
- obtains
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Character Input (AREA)
Abstract
The present invention relates to the inclination angle antidotes and terminal of a kind of text image, belong to data processing field.The present invention obtains processed second text image by the inclination angle of the first text image of correction；The value range at the inclination angle is 0 degree to 45 degree；It calls OCR recognition engine to identify second text image, obtains the first character string；Second text image is rotated into 180 degree, obtains processed third text image；It calls OCR recognition engine to identify the third text image, obtains the second character string；If the high frequency vocabulary number in first character string is greater than the high frequency vocabulary number in second character string, marking second text image is final text image；Otherwise, marking the third text image is final text image.Realize the accuracy for improving correction text image inclination angle.
Description
Technical field
The present invention relates to the inclination angle antidotes and terminal of a kind of text image, belong to data processing field.
Background technique
People often need paper document being converted to electronic document in routine work, life.Common electronic document turns
The method of changing is first to shoot paper document, then by the photo upload of paper document into electronic equipment.But in shooting process
Often there is situations such as photo inclination, influences the experience that people pass through electronic equipment read electronic documents.
In the prior art, application No. is the Chinese patent literatures of 201510047692.X to provide a kind of inclination of text image
Angle correction and helps view apparatus, system at printed page analysis method.The inclination angle antidote of text image includes: to text image
Edge detection is carried out, edge image is obtained；Connected domain detection is carried out to text image, obtains the external of the connected domain of text image
The central point of rectangle；Connected domain detection is carried out to edge image, obtains the central point of the boundary rectangle of the connected domain of edge image；
In the boundary rectangle of the connected domain of central point and edge image to the boundary rectangle of the connected domain of obtained text image
The set application Hough transform method of heart point carries out detection of inclining, and obtains the inclination angle of text image；And it is greater than or waits at inclination angle
In the case where the threshold value of preset first inclination angle, inclination angle correction is carried out to text image.Text image of the invention inclines
Oblique angle antidote does not need to make Hough transformation to each pixel of text image, and operand is small, be illuminated by the light influence it is small,
Accuracy is high.
But the inclination angle antidote of the text image of above patent document offer, the tilt angle model that can be adjusted
It encloses for 0-45 degree, there is certain limitation, the inclination angle correction accuracy of the text image big for tilt angle is not high.
Summary of the invention
The technical problems to be solved by the present invention are: how to improve the accuracy of the big text image of correction tilt angle.
In order to solve the above-mentioned technical problem, the technical solution adopted by the present invention are as follows:
The present invention provides a kind of inclination angle antidote of text image, comprising:
The inclination angle for correcting the first text image obtains processed second text image；The value range at the inclination angle is 0
Degree is to 45 degree；
It calls OCR recognition engine to identify second text image, obtains the first character string；
Second text image is rotated into 180 degree, obtains processed third text image；
It calls OCR recognition engine to identify the third text image, obtains the second character string；
If the high frequency vocabulary number in first character string is greater than the high frequency vocabulary number in second character string, mark
Second text image is final text image；Otherwise, marking the third text image is final text image.
Preferably, the inclination angle for correcting the first text image, obtains processed second text image, specifically:
First text image is rotated by 90 °, processed 4th text image is obtained；
Connected domain detection is carried out to first text image, obtains first nodal point set；The first nodal point set by
The central point of connected domain in first text image forms；Connected domain detection is carried out to the 4th text image, is obtained
Second center point set；Second center point set is made of the central point of the connected domain in the 4th text image；
According to the first nodal point set fitting a straight line, first straight line set is obtained；It is quasi- according to second center point set
Straight line is closed, second straight line set is obtained；The straight line and the angle of horizontal direction are less than 45 degree；
If the element number of the first straight line set is greater than the element number of the second straight line set, according to described first
Straight line set corrects the inclination angle of first text image, obtains the second text image；Otherwise:
The inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text image.
Preferably, connected domain detection is carried out to first text image, obtains first nodal point set, specifically:
S11, connected domain detection is carried out to first text image, obtains all connected domains in first text image；
The connected domain that projection is fallen into same section is added to the same connection by the first text image described in S12, floor projection
Domain set, obtains multiple connected domain set；
S13, a connected domain set is obtained from the multiple connected domain set；
S14, a connected domain is obtained from a connected domain set, obtains the first connected domain；
S15, another connected domain adjacent with first connected domain is obtained from the connected domain set, obtain the second connection
Domain；
It is wide to obtain average character for the average value of the height of S16, the height for calculating first connected domain and second connected domain
Degree；
If the difference of the height of the height of S17, first connected domain and second connected domain is less than preset height threshold,
And the horizontal space between the boundary rectangle of first connected domain and the boundary rectangle of second connected domain is less than described put down
Equal character width then adds the central point of first connected domain and the central point of second connected domain into preset third
Heart point set；
S18, the step S14 to the step S17 is repeated, until the connected domain set is traversed；
S19, the step S13 to the step S18 is repeated, until the multiple connected domain set is traversed, obtained more
A third center point set；The first nodal point set is made of multiple third center point sets.
Preferably, according to the first nodal point set fitting a straight line, first straight line set is obtained, specifically:
S21, according to a third center point set fitting a straight line, and add the straight line to the first straight line set；
S22, the step S21 is repeated, until all third center point sets are traversed.
Preferably, the inclination angle that first text image is corrected according to the first straight line set, obtains the second text
Image, specifically:
S1, the inclination angle that a straight line is obtained from the first straight line set, obtain the first inclination angle；
If the inclination angle of another straight line in S2, first inclination angle and the first straight line set in addition to a straight line
Difference be less than preset inclination angle threshold value, then add another straight line to preset third straight line set；
S3, the step S2 is repeated, until the first straight line set is traversed；
S4, the step S1 to S3 is repeated, until the first straight line set is traversed, obtains multiple third straight lines
Set；
S5, the most third straight line set of element number is obtained, obtains the 4th straight line set；
S6, optimal rotation angle is obtained according to the inclination angle of each straight line in the 4th straight line set；
S7, first text image is rotated into the optimal rotation angle, obtains the second text image.
The present invention also provides a kind of inclinations angle of text image to correct terminal, including one or more processors and storage
Device, the memory are stored with program, and are configured to execute following steps by one or more of processors:
The inclination angle for correcting the first text image obtains processed second text image；The value range at the inclination angle is 0
Degree is to 45 degree；
It calls OCR recognition engine to identify second text image, obtains the first character string；
Second text image is rotated into 180 degree, obtains processed third text image；
It calls OCR recognition engine to identify the third text image, obtains the second character string；
If the high frequency vocabulary number in first character string is greater than the high frequency vocabulary number in second character string, mark
Second text image is final text image；Otherwise, marking the third text image is final text image.
Preferably, the inclination angle for correcting the first text image, obtains processed second text image, specifically:
First text image is rotated by 90 °, processed 4th text image is obtained；
Connected domain detection is carried out to first text image, obtains first nodal point set；The first nodal point set by
The central point of connected domain in first text image forms；Connected domain detection is carried out to the 4th text image, is obtained
Second center point set；Second center point set is made of the central point of the connected domain in the 4th text image；
According to the first nodal point set fitting a straight line, first straight line set is obtained；It is quasi- according to second center point set
Straight line is closed, second straight line set is obtained；The straight line and the angle of horizontal direction are less than 45 degree；
If the element number of the first straight line set is greater than the element number of the second straight line set, according to described first
Straight line set corrects the inclination angle of first text image, obtains the second text image；Otherwise:
The inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text image.
Preferably, connected domain detection is carried out to first text image, obtains first nodal point set, specifically:
S11, connected domain detection is carried out to first text image, obtains all connected domains in first text image；
The connected domain that projection is fallen into same section is added to the same connection by the first text image described in S12, floor projection
Domain set, obtains multiple connected domain set；
S13, a connected domain set is obtained from the multiple connected domain set；
S14, a connected domain is obtained from a connected domain set, obtains the first connected domain；
S15, another connected domain adjacent with first connected domain is obtained from the connected domain set, obtain the second connection
Domain；
It is wide to obtain average character for the average value of the height of S16, the height for calculating first connected domain and second connected domain
Degree；
If the difference of the height of the height of S17, first connected domain and second connected domain is less than preset height threshold,
And the horizontal space between the boundary rectangle of first connected domain and the boundary rectangle of second connected domain is less than described put down
Equal character width then adds the central point of first connected domain and the central point of second connected domain into preset third
Heart point set；
S18, the step S14 to the step S17 is repeated, until the connected domain set is traversed；
S19, the step S13 to the step S18 is repeated, until the multiple connected domain set is traversed, obtained more
A third center point set；The first nodal point set is made of multiple third center point sets.
Preferably, according to the first nodal point set fitting a straight line, first straight line set is obtained, specifically:
S21, according to a third center point set fitting a straight line, and add the straight line to the first straight line set；
S22, the step S21 is repeated, until all third center point sets are traversed.
Preferably, the inclination angle that first text image is corrected according to the first straight line set, obtains the second text
Image, specifically:
S1, the inclination angle that a straight line is obtained from the first straight line set, obtain the first inclination angle；
If the inclination angle of another straight line in S2, first inclination angle and the first straight line set in addition to a straight line
Difference be less than preset inclination angle threshold value, then add another straight line to preset third straight line set；
S3, the step S2 is repeated, until the first straight line set is traversed；
S4, the step S1 to S3 is repeated, until the first straight line set is traversed, obtains multiple third straight lines
Set；
S5, the most third straight line set of element number is obtained, obtains the 4th straight line set；
S6, optimal rotation angle is obtained according to the inclination angle of each straight line in the 4th straight line set；
S7, first text image is rotated into the optimal rotation angle, obtains the second text image.
The invention has the following beneficial effects:
1, the present invention provides the inclination angle antidote and terminal of a kind of text image, is first adjusted to the character in text image
With the degree of horizontal direction parallel, at this point, correction after the second text image may be inverted.In order to improve text image
The accuracy of inclination angle correction, the present invention identify the second text image and the second text image by OCR identification technology respectively
Inverted Image (third text image), and the OCR recognition result of the two is matched with preset high frequency lexicon respectively,
More than the two medium-high frequency vocabulary is then final correction result.Divided by the OCR recognition result to text image after each correction
Analysis, and the accuracy of each correction result is verified, so that text image is as final text image, pole after choosing an optimal correction
The accuracy of the inclination angle correction of text image is improved in big degree.
2, further, the present invention carries out straight line fitting operation to the central point of the connected domain in text image, if connection
The arrangement in domain is disorderly and unsystematic, then can not successfully be fitted and be in line, just can basis only when there are part connected domain arranged regular
The central point of the connected domain of these arranged regulars is fitted to several straight lines.The present invention is first by initial text image (the first text
This image) it is rotated by 90 °, and the 4th text image after being rotated by 90 ° is compared with initial text image, root in the two
The more person's connected domains arrangements of the fitting a straight line obtained according to the central point of connected domain are more regular, then further progress 0- based on this
The fine tuning operation at 45 degree of inclinations angle.The present invention before the fine tuning for carrying out 0-45 degree inclination angle, can first automatic identification it is initial
The general alignment direction of text in text image, and decide whether for initial text image to be rotated by 90 ° according to the actual situation,
Effectively improve the accuracy of the inclination angle correction of text image.
3, further, the present invention first passes through horizontal projection method and connected domain is classified by row, then successively chooses from every a line
It selects connected domain difference in height little and infrastructure elements of the lesser connected domain of horizontal space as fitting a straight line, eliminates in Chinese character
Disconnected pen, a bit, one it is horizontal etc. can not react the disturbing factor that Chinese character overall structure but will affect the accuracy of the straight line fitted, have
The accuracy for improving to effect the inclination angle reflection text image inclined degree of fitting a straight line, to improve inclining for text image
The accuracy of oblique angle correction.
4, further, a third center point set is made of the central point for being located at the connected region with a line, according to
Each third center point set fitting a straight line realizes fitting a straight line in the horizontal direction.
5, further, the present invention identifies inclination angle by traversing the inclination angle difference between every two fitting a straight lines
The biggish exception fitting a straight line compared with the inclination angle of other straight lines, and abnormal fitting a straight line is foreclosed, according to inclination angle
Several little fitting a straight lines of gap calculate an optimal rotation angle and correct to text image, improve text diagram
The accuracy of the inclination angle correction of picture.
Detailed description of the invention
Fig. 1 is a kind of flow chart element of the specific embodiment of the inclination angle antidote of text image provided by the invention
Figure；
Fig. 2 is the first text image schematic diagram of embodiment provided by the invention；
Fig. 3 is the 4th text image schematic diagram of embodiment provided by the invention；
Fig. 4 is the fitting a straight line schematic diagram of embodiment provided by the invention；
Fig. 5 is another fitting a straight line schematic diagram of embodiment provided by the invention；
Fig. 6 is the second text image schematic diagram of embodiment provided by the invention；
Fig. 7 is the third text image schematic diagram of embodiment provided by the invention；
Fig. 8 is the structural block diagram that the specific embodiment of terminal is corrected at a kind of inclination angle of text image provided by the invention；
Fig. 9 is the 5th text image schematic diagram of embodiment provided by the invention；
Figure 10 is the 6th text image schematic diagram of embodiment provided by the invention；
Label declaration:
1, processor；2, memory.
Specific embodiment
It is next in the following with reference to the drawings and specific embodiments that the present invention will be described in detail.
Fig. 1 to Figure 10 is please referred to,
The embodiment of the present invention one are as follows:
As shown in Figure 1, the present embodiment provides a kind of inclination angle antidotes of text image, comprising:
S1, the inclination angle for correcting the first text image, obtain processed second text image；The value range at the inclination angle
It is 0 degree to 45 degree.Specifically:
S11, first text image is rotated by 90 °, obtains processed 4th text image.
For example, the first text image is as shown in Fig. 2, the character in the first text image is inverted, is tilted, OCR recognition engine
Corresponding character can not be correctly identified according to the first text image.4th text image is as shown in figure 3, the 4th text image is
The result that first text image is rotated by 90 °.
S12, connected domain detection is carried out to first text image, obtains first nodal point set；First center
Point set is made of the central point of the connected domain in first text image.Specifically:
S121, connected domain detection is carried out to first text image, obtains all connected domains in first text image.
Wherein, connected domain refers to the set that all point to communicate with each other is constituted, and the point to communicate with each other forms one
Region, and disconnected point forms different regions.Each word in text image can be tentatively recognized by connected domain detection
The position of symbol.
The connected domain that projection is fallen into same section is added to same by the first text image described in S122, floor projection
A connected domain set obtains multiple connected domain set.
Wherein, floor projection refers to projects the first text image in the horizontal direction, falls into the pixel in same section
Point is from the same as a line connected domain.By the connected domain classification being about in text image.
S123, a connected domain set is obtained from the multiple connected domain set.
S124, a connected domain is obtained from a connected domain set, obtains the first connected domain.
S125, another connected domain adjacent with first connected domain is obtained from the connected domain set, obtain the
Two-connected domain.
The average value of the height of S126, the height for calculating first connected domain and second connected domain, is averaged
Character width.
If the difference of the height of the height of S127, first connected domain and second connected domain is less than preset height
Threshold value, and the horizontal space between the boundary rectangle of first connected domain and the boundary rectangle of second connected domain is less than institute
Average character width is stated, then adds the central point of first connected domain and the central point of second connected domain to preset
Three center point sets.
S128, the step S124 to the step S127 is repeated, until the connected domain set is traversed.
For example, connected domain set storage is located at the connected domain with a line, successively traverse each in the connected domain set
A connected domain is located at little with a line difference in height, and horizontal space is small by it compared with adjacent connected domain to pick out
Connected domain, and the central point of these connected domains is stored in a third center point set.
S129, the step S123 to the step S128 is repeated, until the multiple connected domain set is traversed,
Obtain multiple third center point sets；The first nodal point set is made of multiple third center point sets.
For example, traversing each connected domain set, multiple third centers point set is obtained.One third center point set is deposited
Contain the central point of a line connected domain.
Wherein, the present embodiment first passes through horizontal projection method and connected domain is classified by row, then the successively company of selecting from every a line
The logical infrastructure elements that domain difference in height is little and the lesser connected domain of horizontal space is as fitting a straight line eliminate Chinese character interruption
Pen, a bit, one it is horizontal etc. can not react the disturbing factor that Chinese character overall structure but will affect the accuracy of the straight line fitted, effectively
Ground improves the accuracy of the inclination angle reflection text image inclined degree of fitting a straight line, to improve the inclination of text image
The accuracy of angle correction.
S13, connected domain detection is carried out to the 4th text image, obtains the second center point set；Second center
Point set is made of the central point of the connected domain in the 4th text image.
Wherein, the specific implementation of the step S13 is identical as the step S12.
S14, according to the first nodal point set fitting a straight line, obtain first straight line set；The straight line and level side
To angle less than 45 degree.Specifically:
S141, according to a third center point set fitting a straight line, and add the straight line to the first straight line set；
S142, the step S141 is repeated, until all third center point sets are traversed.
For example, as shown in Figure 4 according to the result of first nodal point set fitting a straight line.
S15, according to the second center point set fitting a straight line, obtain second straight line set；The straight line and level side
To angle less than 45 degree.
Wherein, the specific implementation of the step S14 is identical as the step S14.It is straight according to the fitting of the second central point
The result of line is as shown in Figure 5.
If the element number of S16, the first straight line set is greater than the element number of the second straight line set, basis
The first straight line set corrects the inclination angle of first text image, obtains the second text image；Otherwise:
The inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text image.
For example, straight line number is obviously smaller than straight line number in Fig. 4 in Fig. 5, therefore the should be corrected according to first straight line set
The inclination angle of one text image obtains processed second text image.
Wherein, the inclination angle that first text image is corrected according to the first straight line set, obtains the second text diagram
Picture.Specifically:
S161, the inclination angle that a straight line is obtained from the first straight line set, obtain the first inclination angle.
If S162, first inclination angle and another straight line in the first straight line set in addition to a straight line
The difference at inclination angle is less than preset inclination angle threshold value, then adds another straight line to preset third straight line set.
S163, the step S162 is repeated, until the first straight line set is traversed.
For example, the inclination angle of a straight line is 30 degree in first straight line set, the inclination angle of other straight lines is respectively 28 degree, 29
Degree, 31 degree, 32 degree and 20 degree are foreclosed the straight line that inclination angle is 20 degree by repeating step S162, other straight lines
Element as a third straight line set.
S164, the step S161 to S163 is repeated, until the first straight line set is traversed, obtains multiple institutes
State third straight line set.
S165, the most third straight line set of element number is obtained, obtains the 4th straight line set.
For example, it is most as the similar straight line number in the obtained inclination angle of object of reference using the straight line that inclination angle is 30 degree, then
Using its third straight line set as the subsequent benchmark for calculating optimal rotation angle.
S166, optimal rotation angle is obtained according to the inclination angle of each straight line in the 4th straight line set.
Wherein, include but are not limited to choose the average value at the inclination angle of each straight line, maximum in the 4th straight line set
Value, minimum value or median are as the optimal rotation angle.
S167, first text image is rotated into the optimal rotation angle, obtains the second text image.
Wherein, the present embodiment identifies inclination angle and its by traversing the inclination angle difference between every two fitting a straight lines
The inclination angle of its straight line forecloses abnormal fitting a straight line compared to biggish abnormal fitting a straight line, according to inclination angle gap
Several little fitting a straight lines calculate an optimal rotation angle and correct to text image, improve text image
The accuracy of inclination angle correction.
Wherein, the inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text diagram
The implementation of picture and the inclination angle that first text image is corrected according to the first straight line set, obtain the second text diagram
The implementation of picture is identical.
Wherein, the present embodiment carries out straight line fitting operation to the central point of the connected domain in text image, if connected domain
It arranges disorderly and unsystematic, then can not successfully be fitted and be in line, it, just can be according to these only when there are part connected domain arranged regular
The central point of the connected domain of arranged regular is fitted to several straight lines.The present embodiment is first by initial text image (the first text
Image) it is rotated by 90 °, and the 4th text image after being rotated by 90 ° is compared with initial text image, basis in the two
The fitting a straight line that the central point of connected domain obtains more person's connected domains arrangement is more regular, then further progress 0-45 based on this
Spend the fine tuning operation at inclination angle.The present embodiment before the fine tuning for carrying out 0-45 degree inclination angle, can first automatic identification it is initial
The general alignment direction of text in text image, and decide whether for initial text image to be rotated by 90 ° according to the actual situation,
Effectively improve the accuracy of the inclination angle correction of text image.
S2, it calls OCR recognition engine to identify second text image, obtains the first character string.
For example, the second text image is as shown in fig. 6, the first character string for calling OCR recognition engine to identify that Fig. 6 is obtained is
" royal cell is fine and soft red from " j celestial being is by weary " U dirt ' lucky one or two shadow i scholar scholar of bifurcation ".
S3, second text image is rotated into 180 degree, obtains processed third text image；OCR identification is called to draw
The identification third text image is held up, the second character string is obtained.
For example, third text image is as shown in fig. 7, the second character string for calling OCR recognition engine to identify that Fig. 7 is obtained is
It " pours dry just dry！Second day early in the morning, I and " Divine Land flies song " column group directress king ".
If the high frequency vocabulary number in S4, first character string is greater than the high frequency vocabulary in second character string
Number, then marking second text image is final text image；Otherwise, marking the third text image is final text diagram
Picture.
For example, not including high frequency vocabulary in the first character string, the high frequency vocabulary number in the first character string is 0.And second
It include high frequencies vocabulary such as " second days ", " early in the morning ", " column group " and " director " in character string, the high frequency words in the second character string
Number of converging is 4.Therefore, the high frequency vocabulary number in the second character string is greater than the high frequency vocabulary number in the first character string, then and the
Three text images are final text image.
It wherein, first will be in text image the present embodiment provides the inclination angle antidote and terminal of a kind of text image
Character is adjusted to the degree with horizontal direction parallel, at this point, the second text image after correction may be inverted.In order to improve
Text image inclination angle correction accuracy, the present embodiment by OCR identification technology identify respectively the second text image and
The Inverted Image (third text image) of second text image, and by the OCR recognition result of the two respectively with preset high frequency words
Remittance library is matched, and more than the two medium-high frequency vocabulary is then final correction result.Known by the OCR to text image after each correction
Other result is analyzed, and verifies the accuracy of each correction result, is used as most to choose text image after an optimal correction
Whole text image largely improves the accuracy of the inclination angle correction of text image.
For example, Fig. 2 can only be remedied to Fig. 9 by existing declining angle rectification technology, Fig. 3 is remedied to Figure 10, rectify
Positive range is 0-45 degree, but the text image after correcting may be inverted, or image could be made by needing to be rotated by 90 °
In text be in correct position.The present embodiment judges whether to need that (45 degree of the prior art will be utilized by OCR identification technology
Correction in angular region) correction result (Fig. 2 and Fig. 3) further progress 180 degree or 90 degree rotate.For example, the OCR of Fig. 2 is identified
As a result the correct characters quantity identified in is few, and high frequency vocabulary is few；And OCR identification, OCR identification are carried out after Fig. 2 is rotated 180 degree
As a result correct character quantity is more in, and high frequency vocabulary is more.The text image corrected using the prior art is judged as a result,
(Fig. 2) needs to rotate 180 degree, expand the prior art can automatic straightening text inclination angle range, improve text image
The accuracy of inclination angle correction.
The embodiment of the present invention two are as follows:
As shown in figure 8, the present embodiment provides a kind of inclinations angle of text image to correct terminal, including one or more processors 1
And memory 2, the memory 2 are stored with program, and are configured to execute following step by one or more of processors 1
It is rapid:
S1, the inclination angle for correcting the first text image, obtain processed second text image；The value range at the inclination angle
It is 0 degree to 45 degree.Specifically:
S11, first text image is rotated by 90 °, obtains processed 4th text image.
For example, the first text image is as shown in Fig. 2, the 4th text image is as shown in Figure 3.
S12, connected domain detection is carried out to first text image, obtains first nodal point set；First center
Point set is made of the central point of the connected domain in first text image.Specifically:
S121, connected domain detection is carried out to first text image, obtains all connected domains in first text image.
Wherein, connected domain refers to the set that all point to communicate with each other is constituted, and the point to communicate with each other forms one
Region, and disconnected point forms different regions.Each word in text image can be tentatively recognized by connected domain detection
The position of symbol.
The connected domain that projection is fallen into same section is added to same by the first text image described in S122, floor projection
A connected domain set obtains multiple connected domain set.
Wherein, floor projection refers to projects the first text image in the horizontal direction, falls into the pixel in same section
Point is from the same as a line connected domain.By the connected domain classification being about in text image.
S123, a connected domain set is obtained from the multiple connected domain set.
S124, a connected domain is obtained from a connected domain set, obtains the first connected domain.
S125, another connected domain adjacent with first connected domain is obtained from the connected domain set, obtain the
Two-connected domain.
The average value of the height of S126, the height for calculating first connected domain and second connected domain, is averaged
Character width.
If the difference of the height of the height of S127, first connected domain and second connected domain is less than preset height
Threshold value, and the horizontal space between the boundary rectangle of first connected domain and the boundary rectangle of second connected domain is less than institute
Average character width is stated, then adds the central point of first connected domain and the central point of second connected domain to preset
Three center point sets.
S128, the step S124 to the step S127 is repeated, until the connected domain set is traversed.
For example, connected domain set storage is located at the connected domain with a line, successively traverse each in the connected domain set
A connected domain is located at little with a line difference in height, and horizontal space is small by it compared with adjacent connected domain to pick out
Connected domain, and the central point of these connected domains is stored in a third center point set.
S129, the step S123 to the step S128 is repeated, until the multiple connected domain set is traversed,
Obtain multiple third center point sets；The first nodal point set is made of multiple third center point sets.
For example, traversing each connected domain set, multiple third centers point set is obtained.One third center point set is deposited
Contain the central point of a line connected domain.
Wherein, the present embodiment first passes through horizontal projection method and connected domain is classified by row, then the successively company of selecting from every a line
The logical infrastructure elements that domain difference in height is little and the lesser connected domain of horizontal space is as fitting a straight line eliminate Chinese character interruption
Pen, a bit, one it is horizontal etc. can not react the disturbing factor that Chinese character overall structure but will affect the accuracy of the straight line fitted, effectively
Ground improves the accuracy of the inclination angle reflection text image inclined degree of fitting a straight line, to improve the inclination of text image
The accuracy of angle correction.
S13, connected domain detection is carried out to the 4th text image, obtains the second center point set；Second center
Point set is made of the central point of the connected domain in the 4th text image.
Wherein, the specific implementation of the step S13 is identical as the step S12.
S14, according to the first nodal point set fitting a straight line, obtain first straight line set；The straight line and level side
To angle less than 45 degree.Specifically:
S141, according to a third center point set fitting a straight line, and add the straight line to the first straight line set；
S142, the step S141 is repeated, until all third center point sets are traversed.
For example, as shown in Figure 4 according to the result of first nodal point set fitting a straight line.
S15, according to the second center point set fitting a straight line, obtain second straight line set；The straight line and level side
To angle less than 45 degree.
Wherein, the specific implementation of the step S14 is identical as the step S14.It is straight according to the fitting of the second central point
The result of line is as shown in Figure 5.
If the element number of S16, the first straight line set is greater than the element number of the second straight line set, basis
The first straight line set corrects the inclination angle of first text image, obtains the second text image；Otherwise:
The inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text image.
For example, straight line number is obviously smaller than straight line number in Fig. 4 in Fig. 5, therefore the should be corrected according to first straight line set
The inclination angle of one text image obtains processed second text image.
Wherein, the inclination angle that first text image is corrected according to the first straight line set, obtains the second text diagram
Picture.Specifically:
S161, the inclination angle that a straight line is obtained from the first straight line set, obtain the first inclination angle.
If S162, first inclination angle and another straight line in the first straight line set in addition to a straight line
The difference at inclination angle is less than preset inclination angle threshold value, then adds another straight line to preset third straight line set.
S163, the step S162 is repeated, until the first straight line set is traversed.
For example, the inclination angle of a straight line is 30 degree in first straight line set, the inclination angle of other straight lines is respectively 28 degree, 29
Degree, 31 degree, 32 degree and 20 degree are foreclosed the straight line that inclination angle is 20 degree by repeating step S162, other straight lines
Element as a third straight line set.
S164, the step S161 to S163 is repeated, until the first straight line set is traversed, obtains multiple institutes
State third straight line set.
S165, the most third straight line set of element number is obtained, obtains the 4th straight line set.
For example, it is most as the similar straight line number in the obtained inclination angle of object of reference using the straight line that inclination angle is 30 degree, then
Using its third straight line set as the subsequent benchmark for calculating optimal rotation angle.
S166, optimal rotation angle is obtained according to the inclination angle of each straight line in the 4th straight line set.
Wherein, include but are not limited to choose the average value at the inclination angle of each straight line, maximum in the 4th straight line set
Value, minimum value or median are as the optimal rotation angle.
S167, first text image is rotated into the optimal rotation angle, obtains the second text image.
Wherein, the present embodiment identifies inclination angle and its by traversing the inclination angle difference between every two fitting a straight lines
The inclination angle of its straight line forecloses abnormal fitting a straight line compared to biggish abnormal fitting a straight line, according to inclination angle gap
Several little fitting a straight lines calculate an optimal rotation angle and correct to text image, improve text image
The accuracy of inclination angle correction.
Wherein, the inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text diagram
The implementation of picture and the inclination angle that first text image is corrected according to the first straight line set, obtain the second text diagram
The implementation of picture is identical.
Wherein, the present embodiment carries out straight line fitting operation to the central point of the connected domain in text image, if connected domain
It arranges disorderly and unsystematic, then can not successfully be fitted and be in line, it, just can be according to these only when there are part connected domain arranged regular
The central point of the connected domain of arranged regular is fitted to several straight lines.The present embodiment is first by initial text image (the first text
Image) it is rotated by 90 °, and the 4th text image after being rotated by 90 ° is compared with initial text image, basis in the two
The fitting a straight line that the central point of connected domain obtains more person's connected domains arrangement is more regular, then further progress 0-45 based on this
Spend the fine tuning operation at inclination angle.The present embodiment before the fine tuning for carrying out 0-45 degree inclination angle, can first automatic identification it is initial
The general alignment direction of text in text image, and decide whether for initial text image to be rotated by 90 ° according to the actual situation,
Effectively improve the accuracy of the inclination angle correction of text image.
S2, it calls OCR recognition engine to identify second text image, obtains the first character string.
For example, the second text image is as shown in fig. 6, the first character string for calling OCR recognition engine to identify that Fig. 6 is obtained is
" royal cell is fine and soft red from " j celestial being is by weary " U dirt ' lucky one or two shadow i scholar scholar of bifurcation ".
S3, second text image is rotated into 180 degree, obtains processed third text image；OCR identification is called to draw
The identification third text image is held up, the second character string is obtained.
For example, third text image is as shown in fig. 7, the second character string for calling OCR recognition engine to identify that Fig. 7 is obtained is
It " pours dry just dry！Second day early in the morning, I and " Divine Land flies song " column group directress king ".
If the high frequency vocabulary number in S4, first character string is greater than the high frequency vocabulary in second character string
Number, then marking second text image is final text image；Otherwise, marking the third text image is final text diagram
Picture.
For example, in the first character string do not include high frequency vocabulary, and in the second character string comprising " second day ", " early in the morning ",
The high frequencies vocabulary such as " column group " and " director ", then third text image is final text image.
It wherein, first will be in text image the present embodiment provides the inclination angle antidote and terminal of a kind of text image
Character is adjusted to the degree with horizontal direction parallel, at this point, the second text image after correction may be inverted.In order to improve
Text image inclination angle correction accuracy, the present embodiment by OCR identification technology identify respectively the second text image and
The Inverted Image (third text image) of second text image, and by the OCR recognition result of the two respectively with preset high frequency words
Remittance library is matched, and more than the two medium-high frequency vocabulary is then final correction result.Known by the OCR to text image after each correction
Other result is analyzed, and verifies the accuracy of each correction result, is used as most to choose text image after an optimal correction
Whole text image largely improves the accuracy of the inclination angle correction of text image.
The above description is only an embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of inclination angle antidote of text image characterized by comprising
The inclination angle for correcting the first text image obtains processed second text image；The value range at the inclination angle is 0
Degree is to 45 degree；
It calls OCR recognition engine to identify second text image, obtains the first character string；
Second text image is rotated into 180 degree, obtains processed third text image；
It calls OCR recognition engine to identify the third text image, obtains the second character string；
If the high frequency vocabulary number in first character string is greater than the high frequency vocabulary number in second character string, mark
Second text image is final text image；Otherwise, marking the third text image is final text image.
2. the inclination angle antidote of text image according to claim 1, which is characterized in that the first text image of correction
Inclination angle, obtain processed second text image, specifically:
First text image is rotated by 90 °, processed 4th text image is obtained；
Connected domain detection is carried out to first text image, obtains first nodal point set；The first nodal point set by
The central point of connected domain in first text image forms；Connected domain detection is carried out to the 4th text image, is obtained
Second center point set；Second center point set is made of the central point of the connected domain in the 4th text image；
According to the first nodal point set fitting a straight line, first straight line set is obtained；It is quasi- according to second center point set
Straight line is closed, second straight line set is obtained；The straight line and the angle of horizontal direction are less than 45 degree；
If the element number of the first straight line set is greater than the element number of the second straight line set, according to described first
Straight line set corrects the inclination angle of first text image, obtains the second text image；Otherwise:
The inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text image.
3. the inclination angle antidote of text image according to claim 2, which is characterized in that first text diagram
As progress connected domain detection, first nodal point set is obtained, specifically:
S11, connected domain detection is carried out to first text image, obtains all connected domains in first text image；
The connected domain that projection is fallen into same section is added to the same connection by the first text image described in S12, floor projection
Domain set, obtains multiple connected domain set；
S13, a connected domain set is obtained from the multiple connected domain set；
S14, a connected domain is obtained from a connected domain set, obtains the first connected domain；
S15, another connected domain adjacent with first connected domain is obtained from the connected domain set, obtain the second connection
Domain；
It is wide to obtain average character for the average value of the height of S16, the height for calculating first connected domain and second connected domain
Degree；
If the difference of the height of the height of S17, first connected domain and second connected domain is less than preset height threshold,
And the horizontal space between the boundary rectangle of first connected domain and the boundary rectangle of second connected domain is less than described put down
Equal character width then adds the central point of first connected domain and the central point of second connected domain into preset third
Heart point set；
S18, the step S14 to the step S17 is repeated, until the connected domain set is traversed；
S19, the step S13 to the step S18 is repeated, until the multiple connected domain set is traversed, obtained more
A third center point set；The first nodal point set is made of multiple third center point sets.
4. the inclination angle antidote of text image according to claim 3, which is characterized in that according to first center
Point set fitting a straight line obtains first straight line set, specifically:
S21, according to a third center point set fitting a straight line, and add the straight line to the first straight line set；
S22, the step S21 is repeated, until all third center point sets are traversed.
5. the inclination angle antidote of text image according to claim 1, which is characterized in that according to the first straight line
Set corrects the inclination angle of first text image, obtains the second text image, specifically:
S1, the inclination angle that a straight line is obtained from the first straight line set, obtain the first inclination angle；
If the inclination angle of another straight line in S2, first inclination angle and the first straight line set in addition to a straight line
Difference be less than preset inclination angle threshold value, then add another straight line to preset third straight line set；
S3, the step S2 is repeated, until the first straight line set is traversed；
S4, the step S1 to S3 is repeated, until the first straight line set is traversed, obtains multiple third straight lines
Set；
S5, the most third straight line set of element number is obtained, obtains the 4th straight line set；
S6, optimal rotation angle is obtained according to the inclination angle of each straight line in the 4th straight line set；
S7, first text image is rotated into the optimal rotation angle, obtains the second text image.
6. terminal is corrected at a kind of inclination angle of text image, which is characterized in that including one or more processors and memory, institute
It states memory and is stored with program, and be configured to execute following steps by one or more of processors:
The inclination angle for correcting the first text image obtains processed second text image；The value range at the inclination angle is 0
Degree is to 45 degree；
It calls OCR recognition engine to identify second text image, obtains the first character string；
Second text image is rotated into 180 degree, obtains processed third text image；
It calls OCR recognition engine to identify the third text image, obtains the second character string；
If the high frequency vocabulary number in first character string is greater than the high frequency vocabulary number in second character string, mark
Second text image is final text image；Otherwise, marking the third text image is final text image.
7. terminal is corrected at the inclination angle of text image according to claim 6, which is characterized in that the first text image of correction
Inclination angle, obtain processed second text image, specifically:
First text image is rotated by 90 °, processed 4th text image is obtained；
Connected domain detection is carried out to first text image, obtains first nodal point set；The first nodal point set by
The central point of connected domain in first text image forms；Connected domain detection is carried out to the 4th text image, is obtained
Second center point set；Second center point set is made of the central point of the connected domain in the 4th text image；
According to the first nodal point set fitting a straight line, first straight line set is obtained；It is quasi- according to second center point set
Straight line is closed, second straight line set is obtained；The straight line and the angle of horizontal direction are less than 45 degree；
If the element number of the first straight line set is greater than the element number of the second straight line set, according to described first
Straight line set corrects the inclination angle of first text image, obtains the second text image；Otherwise:
The inclination angle that the 4th text image is corrected according to the second straight line set, obtains the second text image.
8. terminal is corrected at the inclination angle of text image according to claim 7, which is characterized in that first text diagram
As progress connected domain detection, first nodal point set is obtained, specifically:
S11, connected domain detection is carried out to first text image, obtains all connected domains in first text image；
The connected domain that projection is fallen into same section is added to the same connection by the first text image described in S12, floor projection
Domain set, obtains multiple connected domain set；
S13, a connected domain set is obtained from the multiple connected domain set；
S14, a connected domain is obtained from a connected domain set, obtains the first connected domain；
S15, another connected domain adjacent with first connected domain is obtained from the connected domain set, obtain the second connection
Domain；
It is wide to obtain average character for the average value of the height of S16, the height for calculating first connected domain and second connected domain
Degree；
If the difference of the height of the height of S17, first connected domain and second connected domain is less than preset height threshold,
And the horizontal space between the boundary rectangle of first connected domain and the boundary rectangle of second connected domain is less than described put down
Equal character width then adds the central point of first connected domain and the central point of second connected domain into preset third
Heart point set；
S18, the step S14 to the step S17 is repeated, until the connected domain set is traversed；
S19, the step S13 to the step S18 is repeated, until the multiple connected domain set is traversed, obtained more
A third center point set；The first nodal point set is made of multiple third center point sets.
9. terminal is corrected at the inclination angle of text image according to claim 8, which is characterized in that according to first center
Point set fitting a straight line obtains first straight line set, specifically:
S21, according to a third center point set fitting a straight line, and add the straight line to the first straight line set；
S22, the step S21 is repeated, until all third center point sets are traversed.
10. terminal is corrected at the inclination angle of text image according to claim 6, which is characterized in that straight according to described first
Line set corrects the inclination angle of first text image, obtains the second text image, specifically:
S1, the inclination angle that a straight line is obtained from the first straight line set, obtain the first inclination angle；
If the inclination angle of another straight line in S2, first inclination angle and the first straight line set in addition to a straight line
Difference be less than preset inclination angle threshold value, then add another straight line to preset third straight line set；
S3, the step S2 is repeated, until the first straight line set is traversed；
S4, the step S1 to S3 is repeated, until the first straight line set is traversed, obtains multiple third straight lines
Set；
S5, the most third straight line set of element number is obtained, obtains the 4th straight line set；
S6, optimal rotation angle is obtained according to the inclination angle of each straight line in the 4th straight line set；
S7, first text image is rotated into the optimal rotation angle, obtains the second text image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910189109.7A CN109919155B (en) | 2019-03-13 | 2019-03-13 | Inclination angle correction method for text image and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910189109.7A CN109919155B (en) | 2019-03-13 | 2019-03-13 | Inclination angle correction method for text image and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109919155A true CN109919155A (en) | 2019-06-21 |
| CN109919155B CN109919155B (en) | 2021-03-12 |
Family
ID=66964599
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910189109.7A Active CN109919155B (en) | 2019-03-13 | 2019-03-13 | Inclination angle correction method for text image and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109919155B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110647882A (en) * | 2019-09-20 | 2020-01-03 | 上海眼控科技股份有限公司 | Image correction method, device, equipment and storage medium |
| CN112085014A (en) * | 2020-09-10 | 2020-12-15 | 京东方科技集团股份有限公司 | OCR character position correction method and device, storage medium and electronic equipment |
| CN113158895A (en) * | 2021-04-20 | 2021-07-23 | 北京中科江南信息技术股份有限公司 | Bill identification method and device, electronic equipment and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050281483A1 (en) * | 2004-06-08 | 2005-12-22 | Benq Corporation | Tilted angle detection for document image deskew |
| CN101655981A (en) * | 2009-09-09 | 2010-02-24 | 谭洪舟 | Method for detecting and adjusting inversion of certificate image |
| CN104463126A (en) * | 2014-12-15 | 2015-03-25 | 湖南工业大学 | Automatic slant angle detecting method for scanned document image |
| US20160275368A1 (en) * | 2013-10-31 | 2016-09-22 | Nec Corporation | Management system, list production device, method, computer readable recording medium, data structure, and printed label |
| CN105989366A (en) * | 2015-01-30 | 2016-10-05 | 深圳市思路飞扬信息技术有限责任公司 | Inclination angle correcting method of text image, page layout analysis method of text image, vision assistant device and vision assistant system |
| CN106663207A (en) * | 2014-10-29 | 2017-05-10 | 微软技术许可有限责任公司 | Whiteboard and document image detection method and system |
| CN106652673A (en) * | 2017-01-16 | 2017-05-10 | 华南理工大学 | Method for automatically identifying and reading medicine specification |
| CN107563330A (en) * | 2017-09-04 | 2018-01-09 | 南京邮电大学 | A kind of horizontal tilt car plate antidote in monitor video |
| CN108133169A (en) * | 2016-12-01 | 2018-06-08 | 北京新唐思创教育科技有限公司 | A kind of embark on journey processing method and its device for text image |
| CN108345827A (en) * | 2017-01-24 | 2018-07-31 | 富士通株式会社 | Identify method, system and the neural network in document direction |
| CN109409363A (en) * | 2018-10-13 | 2019-03-01 | 长沙芯希电子科技有限公司 | The reverse judgement of text image based on content and bearing calibration |
-
2019
- 2019-03-13 CN CN201910189109.7A patent/CN109919155B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050281483A1 (en) * | 2004-06-08 | 2005-12-22 | Benq Corporation | Tilted angle detection for document image deskew |
| CN101655981A (en) * | 2009-09-09 | 2010-02-24 | 谭洪舟 | Method for detecting and adjusting inversion of certificate image |
| US20160275368A1 (en) * | 2013-10-31 | 2016-09-22 | Nec Corporation | Management system, list production device, method, computer readable recording medium, data structure, and printed label |
| CN106663207A (en) * | 2014-10-29 | 2017-05-10 | 微软技术许可有限责任公司 | Whiteboard and document image detection method and system |
| CN104463126A (en) * | 2014-12-15 | 2015-03-25 | 湖南工业大学 | Automatic slant angle detecting method for scanned document image |
| CN105989366A (en) * | 2015-01-30 | 2016-10-05 | 深圳市思路飞扬信息技术有限责任公司 | Inclination angle correcting method of text image, page layout analysis method of text image, vision assistant device and vision assistant system |
| CN108133169A (en) * | 2016-12-01 | 2018-06-08 | 北京新唐思创教育科技有限公司 | A kind of embark on journey processing method and its device for text image |
| CN106652673A (en) * | 2017-01-16 | 2017-05-10 | 华南理工大学 | Method for automatically identifying and reading medicine specification |
| CN108345827A (en) * | 2017-01-24 | 2018-07-31 | 富士通株式会社 | Identify method, system and the neural network in document direction |
| CN107563330A (en) * | 2017-09-04 | 2018-01-09 | 南京邮电大学 | A kind of horizontal tilt car plate antidote in monitor video |
| CN109409363A (en) * | 2018-10-13 | 2019-03-01 | 长沙芯希电子科技有限公司 | The reverse judgement of text image based on content and bearing calibration |
Non-Patent Citations (3)
| Title |
|---|
| MINXIA HUO ET AL.: ""The license plate recognition system based on improved algorithm"", 《IOP CONFERENCE SERIES MATERIALS SCIENCE AND ENGINEERING》 * |
| MOHD AMIR ET AL.: ""New Fast Content Based Skew Detection Algorithm for Document Images"", 《INTERNATIONAL CONFERENCE PATTERN RECOGNITION AND INFORMATION PROCESSING》 * |
| 李清: ""基于图像识别的网上阅卷***的设计实现与优化"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110647882A (en) * | 2019-09-20 | 2020-01-03 | 上海眼控科技股份有限公司 | Image correction method, device, equipment and storage medium |
| CN112085014A (en) * | 2020-09-10 | 2020-12-15 | 京东方科技集团股份有限公司 | OCR character position correction method and device, storage medium and electronic equipment |
| CN113158895A (en) * | 2021-04-20 | 2021-07-23 | 北京中科江南信息技术股份有限公司 | Bill identification method and device, electronic equipment and storage medium |
| CN113158895B (en) * | 2021-04-20 | 2023-11-14 | 北京中科江南信息技术股份有限公司 | Bill identification method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109919155B (en) | 2021-03-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110135411B (en) | Business card recognition method and device | |
| Burie et al. | ICDAR2015 competition on smartphone document capture and OCR (SmartDoc) | |
| CN109919155A (en) | A kind of the inclination angle antidote and terminal of text image | |
| US8515208B2 (en) | Method for document to template alignment | |
| US10699109B2 (en) | Data entry from series of images of a patterned document | |
| CN106446896B (en) | Character segmentation method and device and electronic equipment | |
| US8942484B2 (en) | Text detection using image regions | |
| CN108427950B (en) | Character line detection method and device | |
| US5844991A (en) | Script identification from images using cluster-based templates | |
| US6169822B1 (en) | Method for correcting direction of document image | |
| CN110866871A (en) | Text image correction method and device, computer equipment and storage medium | |
| US5657396A (en) | Method and apparatus for pattern recognition and validation, especially for hand-written signatures | |
| CN110472208A (en) | The method, system of form analysis, storage medium and electronic equipment in PDF document | |
| US5276742A (en) | Rapid detection of page orientation | |
| CN106407979B (en) | Method and device for correcting bill characters | |
| US7929769B2 (en) | Script recognition for ink notes | |
| CN110717492B (en) | Method for correcting direction of character string in drawing based on joint features | |
| CN110647882A (en) | Image correction method, device, equipment and storage medium | |
| CN108805116A (en) | Image text detection method and its system | |
| CN106648418A (en) | Handwriting correction method and device | |
| CN113850238B (en) | Document detection method and device, electronic equipment and storage medium | |
| KR20190091101A (en) | Automatic classification apparatus and method of document type using deep learning | |
| US8494277B2 (en) | Handwritten character recognition based on frequency variations in characters | |
| JP6690089B2 (en) | Form recognition method, form recognition device, and form recognition program | |
| CN113705673A (en) | Character detection method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |