CN106296682A - Method and device for medical image Chinese version region detection - Google Patents
Method and device for medical image Chinese version region detection Download PDFInfo
- Publication number
- CN106296682A CN106296682A CN201610648984.3A CN201610648984A CN106296682A CN 106296682 A CN106296682 A CN 106296682A CN 201610648984 A CN201610648984 A CN 201610648984A CN 106296682 A CN106296682 A CN 106296682A
- Authority
- CN
- China
- Prior art keywords
- text
- region
- text filed
- polymerized
- numerical value
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Abstract
The present invention discloses a kind of method and device for medical image Chinese version region detection, it is possible to increase text filed accuracy of detection.Described method includes: S1, obtain medical image to be detected；S2, described medical image is detected, obtain a series of connected region, and obtain the two-value template in described medical image Chinese version region based on single sample image；S3, the non-textual region utilizing described two-value template to filter out in described connected region obtain text candidates region, and further filter out the non-textual region in described text candidates region based on character feature；S4, text filed it is polymerized obtain, obtains line of text.
Description
Technical field
The present invention relates to medical image detection technique field, be specifically related to a kind of for medical image Chinese version region detection
Method and device.
Background technology
Medical image image is to be arranged in a matrix constituted image by some by the black pixel to white different gray scales.
It reflects organ and the imaging contexts being organized on specific image documentation equipment, can well show the device being made up of soft tissue
Official, such as brain, spinal cord, indulges diaphragm, lung, liver, gallbladder, pancreas and pelvic part organ etc., and demonstrates disease in good anatomic image background
The image become, therefore medical image has very important using value in medical diagnosis.Original medical image is except figure
As, outside itself, also the additional datas such as patient information being preserved with metadata form according to certain standard (such as DICOM).These numbers
Preserve respectively according to image.
Medical imaging through analyzing is used for diagnosing and preserving through conversion printout.Different from raw video, this kind of use
On the medical image of output in addition to the layer images of organ, also additional information can be directly superimposed to word pattern
On medical image.These words have usually contained the information such as patient's name, detection time and Testing index, for our Exact Solutions
Read medical image and provide important value, therefore detect the position at word place in these medical images, extracted
Have great significance.Prior art generally uses the methods such as MSER, SWT and carries out medical image Chinese version region detection, but
This kind of method is based on what textural characteristics carried out detecting, it is not easy to is made a distinction with other texture by Texture features in image, makes
Become accuracy of detection relatively low.
Summary of the invention
The deficiency existed for prior art and defect, the present invention provides a kind of for medical image Chinese version region detection
Method and device.
On the one hand, the embodiment of the present invention proposes a kind of method for medical image Chinese version region detection, including:
S1, obtain medical image to be detected；
S2, described medical image is detected, obtain a series of connected region, and obtain based on single sample image
The two-value template in described medical image Chinese version region；
S3, the non-textual region utilizing described two-value template to filter out in described connected region obtain text candidates region,
And further filter out the non-textual region in described text candidates region based on character feature；
S4, text filed it is polymerized obtain, obtains line of text.
On the other hand, the embodiment of the present invention proposes a kind of device for medical image Chinese version region detection, including:
Acquiring unit, for obtaining medical image to be detected；
Computing unit, for detecting described medical image, obtains a series of connected region, and based on single sample
Illustration picture obtains the two-value template in described medical image Chinese version region；
Filter unit, obtain text for the non-textual region utilizing described two-value template to filter out in described connected region
Candidate region, and further filter out the non-textual region in described text candidates region based on character feature；
Polymerized unit, for text filed being polymerized obtain, obtains line of text.
The method and device for medical image Chinese version region detection that the embodiment of the present invention provides, utilizes to be detected
The non-textual region that the two-value template in medical image Chinese version region filters out in connected region obtains text candidates region, based on
Character feature further filters out the non-textual region in described text candidates region, and text filed is polymerized obtain,
Obtaining line of text, compared to prior art, the embodiment of the present invention needs not distinguish between textural characteristics, it is possible to increase text filed inspection
Survey precision.
Accompanying drawing explanation
Fig. 1 is the present invention schematic flow sheet for method one embodiment of medical image Chinese version region detection；
Fig. 2 is the present invention structural representation for device one embodiment of medical image Chinese version region detection.
Detailed description of the invention
For making the purpose of the embodiment of the present invention, technical scheme and advantage clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is explicitly described, it is clear that described embodiment is the present invention
A part of embodiment rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not having
Make the every other embodiment obtained under creative work premise, broadly fall into the scope of protection of the invention.
Referring to Fig. 1, the present embodiment discloses a kind of method for medical image Chinese version region detection, including:
S1, obtain medical image to be detected；
S2, described medical image is detected, obtain a series of connected region, and obtain based on single sample image
The two-value template in described medical image Chinese version region；
It should be noted that described medical image is carried out detection can use MSER algorithm, here is omitted.
Single sample image refers to comprise the medical image of text object, it is possible to fully demonstrate the text in medical image special
Levy.In a particular application, the described two-value template obtaining described medical image Chinese version region based on single sample image, can wrap
Include:
The local auto-adaptive calculating described single sample image R returns core KR, and for each connected region T, calculate
The local auto-adaptive of this connected region T returns core KT；
To described KRIt is normalized and obtains weight vector matrix WR, to described KTIt is normalized and is weighed
Value vector matrix WT；
To described WRUse PCA algorithm (Principal Component Analysis Method) to process, obtain main constituent, and retain described main constituent
Front d item constitute matrix PR, by described WRTo described PRProject, obtain characteristic vector F of described single sample image RR, will
Described WTTo described PRProject, obtain characteristic vector F of described connected region TT；
Wherein, described d is integer, specifically can carry out value as required, such as can be with value 4,5,6 etc., the present invention
This is not construed as limiting by embodiment.By described WRTo described PRThe function expression carrying out projecting isBy described WT
To described PRThe function expression carrying out projecting is
Calculate described characteristic vector FRWith FTBetween similarity, it is judged that described similarity measurement whether more than the first numerical value,
If more than described first numerical value, then the pixel value of corresponding connected region being set to 1, obtaining text filed, otherwise, then by correspondence
The pixel value of connected region be set to 0, obtain background area, using described text filed and background area as described two-value mould
Plate.
In a particular application, cosine similarity tolerance can be used to calculate it should be noted that calculate similarity, this
Place repeats no more.Different values can be had, if than medical science figure according to difference first numerical value of medical image Chinese version font
As Chinese version is the Song typeface, then the first numerical value can be 70%, naturally it is also possible to carries out left and right as required and adjusts, the present embodiment pair
This is not construed as limiting.
Further it will be understood that the process that the present embodiment calculates two-value template its essence is the single sample image of calculating
Similarity between characteristic vector and each connected region, sets up corresponding connected region size according to the size of similarity
Completely black or the whitest two-value template.
S3, the non-textual region utilizing described two-value template to filter out in described connected region obtain text candidates region,
And further filter out the non-textual region in described text candidates region based on character feature；
In the present embodiment, the non-textual region utilizing described two-value template to filter out in described connected region obtains text and waits
Favored area, specifically says and the pixel value of the background area in connected region is set to 0, obtains text candidates region, and it processed
The corresponding mathematic(al) representation of journey is Ican=Imask∩IMSER, wherein, IcanFor text candidates region, ImaskFor two-value template, IMSER
For connected region.
Specifically, described further filter out the non-textual region in described text candidates region based on character feature, permissible
Including:
For each text candidates region, calculate stroke width feature SW of text candidate region, and retain stroke
Width characteristics SW is less than the text candidates region of second value, and wherein, the computing formula of described stroke width feature SW is
In formula, std and E is stroke width standard deviation and the meansigma methods of text candidate region respectively；
Generally, generally remain consistent because of the stroke width of single character so that the stroke width in text candidates region
The ratio of degree standard deviation and meansigma methods is less, can be filtered in part non-textual region by means of this feature.Need explanation
Being that the value of second value is relevant with the stroke width of character, the stroke width of character is the biggest, then this value value increases accordingly,
Generally value can be 0.5-1,5.
The number of the non-zero pixels in calculated text candidates region, filters the number of non-zero pixels more than third value
With the text candidates region less than the 4th numerical value；
In a particular application, the value of third value and the 4th numerical value is relevant with the number of pixels in text candidates region, and one
In the case of as, value can be 0.9 times and 0.5 times of the number of pixels in text candidates region respectively.
The region area in the number of the non-zero pixels in calculated text candidates region and corresponding text candidates region
Ratio, filter ratio more than the 5th numerical value with less than the text candidates region of the 6th numerical value；
In a particular application, the value of the 5th numerical value and the 6th numerical value generally can be respectively 70% and 10%.
Calculated text candidates region length-width ratio, filter out length-width ratio more than the 7th numerical value and less than the 8th number
The text candidates region of value；
In a particular application, the value of the 7th numerical value and the 8th numerical value generally can be respectively 1.2 and 0.5.
For each the text candidates region obtained, utilize sciagraphy or connected region domain method to text candidate region
Carrying out cutting, obtain multiple fritter, and determine whether each fritter is character, calculating is the ratio shared by the fritter of character,
Filter the ratio text candidates region less than the 9th numerical value.
It should be noted that determine whether each fritter is that character can use prior art, here is omitted.The
Nine numerical value ordinary circumstances can value be 2/3.
S4, text filed it is polymerized obtain, obtains line of text.
In a particular application, described S4, may include that
For obtain text filed in each text filed A not being polymerized, choose other text not being polymerized
The text filed B that in region one is not polymerized, it is judged that whether these two text filed A and B can be polymerized, if can gather
Close, then these two text filed A and B be polymerized, obtain text filed C, then from other be not polymerized text filed
Choose a text filed D not being polymerized, it is judged that whether described text filed C and D can be polymerized, if can be polymerized, then will
These two text filed C and D are polymerized, repeat above-mentioned choose text filed, judge whether the step that can be polymerized and be polymerized
Rapid until be not polymerized text filed choose complete.
It should be noted that the process to the text filed polymerization being polymerized obtained, its essence is medical image
On condense together at the text of a piece.Certainly, because the text at a piece on medical image is to have a certain distance to close in fact
System, such as the most adjacent text, the abscissa of the rightmost pixel that front is text filed is text filed with back
The absolute value of difference of abscissa of leftmost pixel be no more than the size of 1 pixel, and front text filed with
Vertical dimension between back is text filed is no more than the size of 0.5 pixel, for another example, for neighbouring literary composition
This, the vertical dimension between top is text filed and the most text filed is no more than the size of 1 pixel, top text area
The abscissa of the leftmost pixel in territory with the absolute value of the difference of the abscissa of the most text filed leftmost pixel is
It is not more than the size of 0.5 pixel.Accordingly, two text filed judge processs whether can being polymerized can be built, with literary composition
As a example by A and B of one's respective area, it is judged that process is as follows:
S40, the vertical dimension calculated between these two text filed A and B, it is judged that whether described vertical dimension is less than the tenth
Numerical value, if less than described tenth numerical value, then performs step S41, otherwise, performs step S42；
The text filed pixel comprised that in S41, calculating text filed A and B of said two, the abscissa of pixel is bigger
The maximum of the text filed pixel comprised that minimum abscissa is less with the abscissa of pixel in text filed A and B of said two
The absolute value of the difference of abscissa, it is judged that whether described absolute value is less than the 11st numerical value, if less than described 11st numerical value, then
By less for bigger for the abscissa of the described pixel text filed abscissa being aggregated in described pixel text filed after；
S42, judge described vertical dimension whether less than the 11st numerical value, if less than described 11st numerical value, then calculating institute
State the minimum abscissa of a text filed pixel comprised in two text filed A and B text filed to comprise with another
The absolute value of the difference of the minimum abscissa of pixel, it is judged that whether described absolute value is less than the tenth numerical value, if less than the described tenth
Numerical value, then text filed be aggregated in said two text area by less for the vertical coordinate of pixel in text filed for said two A and B
In territory below bigger text filed of the vertical coordinate of pixel.
Furthermore, it is necessary to explanation, the transverse axis of the coordinate system at coordinate place involved in the embodiment of the present invention be along with
The arragement direction of character is parallel.In addition, it is necessary to explanation, for the tenth numerical value and the value of the 11st numerical value, Ke Yigen
Determining according to the typesetting of word in medical image, for general medical image, the tenth numerical value can be the big of 1 pixel with value
Little, the 11st numerical value can be with size that value is 0.5 pixel.
The method for medical image Chinese version region detection that the embodiment of the present invention provides, utilizes medical science figure to be detected
The non-textual region filtered out in connected region as the two-value template in Chinese version region obtains text candidates region, special based on character
Levy the non-textual region further filtered out in described text candidates region, and text filed be polymerized obtain, obtain literary composition
One's own profession, compared to prior art, the embodiment of the present invention needs not distinguish between textural characteristics, it is possible to increase text filed detection essence
Degree.
Referring to Fig. 2, the present embodiment discloses a kind of device for medical image Chinese version region detection, including:
Acquiring unit 1, for obtaining medical image to be detected；
Computing unit 2, for detecting described medical image, obtains a series of connected region, and based on list
Sample image obtains the two-value template in described medical image Chinese version region；
It should be noted that described medical image is carried out detection can use MSER algorithm, here is omitted.
In a particular application, described computing unit, may be used for:
The local auto-adaptive calculating described single sample image R returns core KR, and for each connected region T, calculate
The local auto-adaptive of this connected region T returns core KT；
To described KRIt is normalized and obtains weight vector matrix WR, to described KTIt is normalized and is weighed
Value vector matrix WT；
To described WRUse PCA algorithm to process, obtain main constituent, and the front d item retaining described main constituent constitutes square
Battle array PR, by described WRTo described PRProject, obtain characteristic vector F of described single sample image RR, by described WTTo described PR
Project, obtain characteristic vector F of described connected region TT, wherein, described d is integer；
Calculate described characteristic vector FRWith FTBetween similarity, it is judged that described similarity measurement whether more than the first numerical value,
If more than described first numerical value, then the pixel value of corresponding connected region being set to 1, obtaining text filed, otherwise, then by correspondence
The pixel value of connected region be set to 0, obtain background area, using described text filed and background area as described two-value mould
Plate.
Cosine similarity tolerance can be used to calculate it should be noted that calculate similarity, here is omitted.
Filter unit 3, obtain literary composition for the non-textual region utilizing described two-value template to filter out in described connected region
This candidate region, and further filter out the non-textual region in described text candidates region based on character feature；
In actual applications, described in filter unit, specifically may be used for:
For each text candidates region, calculate stroke width feature SW of text candidate region, and retain stroke
Width characteristics SW is less than the text candidates region of second value, and wherein, the computing formula of described stroke width feature SW is
In formula, std and E is stroke width standard deviation and the meansigma methods of text candidate region respectively；
The number of the non-zero pixels in calculated text candidates region, filters the number of non-zero pixels more than third value
With the text candidates region less than the 4th numerical value；
The region area in the number of the non-zero pixels in calculated text candidates region and corresponding text candidates region
Ratio, filter ratio more than the 5th numerical value with less than the text candidates region of the 6th numerical value；
Calculated text candidates region length-width ratio, filter out length-width ratio more than the 7th numerical value and less than the 8th number
The text candidates region of value；
For each the text candidates region obtained, utilize sciagraphy or connected region domain method to text candidate region
Carrying out cutting, obtain multiple fritter, and determine whether each fritter is character, calculating is the ratio shared by the fritter of character,
Filter the ratio text candidates region less than the 9th numerical value.
Polymerized unit 4, for text filed being polymerized obtain, obtains line of text.
In the present embodiment, described polymerized unit, specifically may be used for for obtain text filed in each is not gathered
The text filed A closed, choose that other is not polymerized text filed in a text filed B not being polymerized, it is judged that these are two years old
Whether individual text filed A and B can be polymerized, if can be polymerized, is then polymerized by these two text filed A and B, obtains text
Region C, then from other be not polymerized text filed choose a text filed D not being polymerized, it is judged that described text area
Whether territory C and D can be polymerized, if can be polymerized, is then polymerized by these two text filed C and D, repeats above-mentioned to choose text
Region, judge whether the step can be polymerized and be polymerized until be not polymerized text filed choose complete.
In a particular application, described polymerized unit, specifically may be used for:
Calculate the vertical dimension between these two text filed A and B, it is judged that whether described vertical dimension is less than the tenth number
Value, if less than described tenth numerical value, then calculates bigger text filed of the abscissa of pixel in text filed A and B of said two
In minimum abscissa A and B text filed with said two of the pixel comprised, less text filed of the abscissa of pixel comprises
The absolute value of difference of maximum abscissa of pixel, it is judged that whether described absolute value less than the 11st numerical value, if less than described
11st numerical value, then by text less for bigger for the abscissa of the described pixel text filed abscissa being aggregated in described pixel
After region；Or
If not less than described tenth numerical value, then judge whether described vertical dimension is less than the 11st numerical value, if less than described
11st numerical value, then calculate in text filed A and B of said two the minimum abscissa of a text filed pixel comprised with another
The absolute value of the difference of the minimum abscissa of one text filed pixel comprised, it is judged that whether described absolute value is less than the tenth number
Value, if less than described tenth numerical value, then by less for the vertical coordinate of pixel in text filed for said two A and B text filed poly-
It is combined in below bigger text filed of the vertical coordinate of the text filed middle pixel of said two.
The device for medical image Chinese version region detection that the embodiment of the present invention provides, utilizes medical science figure to be detected
The non-textual region filtered out in connected region as the two-value template in Chinese version region obtains text candidates region, special based on character
Levy the non-textual region further filtered out in described text candidates region, and text filed be polymerized obtain, obtain literary composition
One's own profession, compared to prior art, the embodiment of the present invention needs not distinguish between textural characteristics, it is possible to increase text filed detection essence
Degree.
The device for medical image Chinese version region detection of the present embodiment, may be used for performing side shown in earlier figures 1
The technical scheme of method embodiment, it is similar with technique effect that it realizes principle, and here is omitted.
Those skilled in the art are it should be appreciated that embodiments herein can be provided as method, system or computer program
Product.Therefore, the reality in terms of the application can use complete hardware embodiment, complete software implementation or combine software and hardware
Execute the form of example.And, the application can use at one or more computers wherein including computer usable program code
The upper computer program product implemented of usable storage medium (including but not limited to disk memory, CD-ROM, optical memory etc.)
The form of product.
The application is with reference to method, equipment (system) and the flow process of computer program according to the embodiment of the present application
Figure and/or block diagram describe.It should be understood that can the most first-class by computer program instructions flowchart and/or block diagram
Flow process in journey and/or square frame and flow chart and/or block diagram and/or the combination of square frame.These computer programs can be provided
Instruction arrives the processor of general purpose computer, special-purpose computer, Embedded Processor or other programmable data processing device to produce
A raw machine so that the instruction performed by the processor of computer or other programmable data processing device is produced for real
The device of the function specified in one flow process of flow chart or multiple flow process and/or one square frame of block diagram or multiple square frame now.
These computer program instructions may be alternatively stored in and computer or other programmable data processing device can be guided with spy
Determine in the computer-readable memory that mode works so that the instruction being stored in this computer-readable memory produces and includes referring to
Make the manufacture of device, this command device realize at one flow process of flow chart or multiple flow process and/or one square frame of block diagram or
The function specified in multiple square frames.
These computer program instructions also can be loaded in computer or other programmable data processing device so that at meter
Perform sequence of operations step on calculation machine or other programmable devices to produce computer implemented process, thus at computer or
The instruction performed on other programmable devices provides for realizing at one flow process of flow chart or multiple flow process and/or block diagram one
The step of the function specified in individual square frame or multiple square frame.
It should be noted that in this article, the relational terms of such as first and second or the like is used merely to a reality
Body or operation separate with another entity or operating space, and deposit between not necessarily requiring or imply these entities or operating
Relation or order in any this reality.And, term " includes ", " comprising " or its any other variant are intended to
Comprising of nonexcludability, so that include that the process of a series of key element, method, article or equipment not only include that those are wanted
Element, but also include other key elements being not expressly set out, or also include for this process, method, article or equipment
Intrinsic key element.In the case of there is no more restriction, statement " including ... " key element limited, it is not excluded that
Including process, method, article or the equipment of described key element there is also other identical element.Term " on ", D score etc. refers to
The orientation shown or position relationship, for based on orientation shown in the drawings or position relationship, are for only for ease of the description present invention and simplification
Describe rather than indicate or imply that the device of indication or element must have specific orientation, with specific azimuth configuration and behaviour
Make, be therefore not considered as limiting the invention.Unless otherwise clearly defined and limited, term " install ", " being connected ",
" connect " and should be interpreted broadly, connect for example, it may be fixing, it is also possible to be to removably connect, or be integrally connected；Can be
It is mechanically connected, it is also possible to be electrical connection；Can be to be joined directly together, it is also possible to be indirectly connected to by intermediary, can be two
The connection of element internal.For the ordinary skill in the art, can understand that above-mentioned term is at this as the case may be
Concrete meaning in invention.
In the description of the present invention, illustrate a large amount of detail.Although it is understood that, embodiments of the invention can
To put into practice in the case of there is no these details.In some instances, it is not shown specifically known method, structure and skill
Art, in order to do not obscure the understanding of this description.Similarly, it will be appreciated that disclose to simplify the present invention and help to understand respectively
One or more in individual inventive aspect, above in the description of the exemplary embodiment of the present invention, each of the present invention is special
Levy and be sometimes grouped together in single embodiment, figure or descriptions thereof.But, should be by the method solution of the disclosure
Release in reflecting an intention that i.e. the present invention for required protection requires than the feature being expressly recited in each claim more
Many features.More precisely, as the following claims reflect, inventive aspect is less than single reality disclosed above
Execute all features of example.Therefore, it then follows claims of detailed description of the invention are thus expressly incorporated in this detailed description of the invention,
The most each claim itself is as the independent embodiment of the present invention.It should be noted that in the case of not conflicting, this
Embodiment in application and the feature in embodiment can be mutually combined.The invention is not limited in any single aspect, also
It is not limited to any single embodiment, is also not limited to these aspects and/or the combination in any of embodiment and/or displacement.And
And, can be used alone each aspect of the present invention and/or embodiment or with other aspects one or more and/or its implement
Example is used in combination.
Last it is noted that various embodiments above is only in order to illustrate technical scheme, it is not intended to limit；To the greatest extent
The present invention has been described in detail by pipe with reference to foregoing embodiments, it will be understood by those within the art that: it depends on
So the technical scheme described in foregoing embodiments can be modified, or the most some or all of technical characteristic is entered
Row equivalent；And these amendments or replacement, do not make the essence of appropriate technical solution depart from various embodiments of the present invention technology
The scope of scheme, it all should be contained in the middle of the claim of the present invention and the scope of description.
Claims (10)
1. the method for medical image Chinese version region detection, it is characterised in that including:
S1, obtain medical image to be detected；
S2, described medical image is detected, obtain a series of connected region, and obtain described based on single sample image
The two-value template in medical image Chinese version region；
S3, the non-textual region utilizing described two-value template to filter out in described connected region obtain text candidates region, and base
Non-textual region in character feature further filters out described text candidates region；
S4, text filed it is polymerized obtain, obtains line of text.
Method the most according to claim 1, it is characterised in that described obtain in described medical image based on single sample image
Text filed two-value template, including:
Calculate described single sample imageRLocal auto-adaptive return core KR, and for each connected region T, calculate this connection
The local auto-adaptive of region T returns core KT；
To described KRIt is normalized and obtains weight vector matrix WR, to described KTBe normalized obtain weights to
Moment matrix WT；
To described WRUse PCA algorithm to process, obtain main constituent, and the front d item retaining described main constituent constitutes matrix PR,
By described WRTo described PRProject, obtain characteristic vector F of described single sample image RR, by described WTTo described PRThrow
Shadow, obtains characteristic vector F of described connected region TT, wherein, described d is integer；
Calculate described characteristic vector FRWith FTBetween similarity, it is judged that whether described similarity measurement more than the first numerical value, if greatly
In described first numerical value, then the pixel value of corresponding connected region is set to 1, obtains text filed, otherwise, then by corresponding company
The pixel value in logical region is set to 0, obtains background area, will described text filed with background area as described two-value template.
Method the most according to claim 1 and 2, it is characterised in that described further filter out described literary composition based on character feature
Non-textual region in this candidate region, including:
For each text candidates region, calculate stroke width feature SW of text candidate region, and retain stroke width
Feature SW is less than the text candidates region of second value, and wherein, the computing formula of described stroke width feature SW is
In formula, std and E is stroke width standard deviation and the meansigma methods of text candidate region respectively；
The number of the non-zero pixels in calculated text candidates region, filters the number of non-zero pixels more than third value and little
Text candidates region in the 4th numerical value；
The ratio of the number of the non-zero pixels in calculated text candidates region and the region area in corresponding text candidates region
Value, filters ratio and is more than the 5th numerical value and the text candidates region less than the 6th numerical value；
Calculated text candidates region length-width ratio, filter out length-width ratio more than the 7th numerical value with less than the 8th numerical value
Text candidates region；
For each the text candidates region obtained, utilize sciagraphy or connected region domain method that text candidate region is carried out
Cutting, obtains multiple fritter, and determines whether each fritter is character, and calculating is the ratio shared by the fritter of character, filters
Ratio is less than the text candidates region of the 9th numerical value.
Method the most according to claim 1, it is characterised in that described S4, including:
For obtain text filed in each text filed A not being polymerized, that chooses that other is not polymerized is text filed
In a text filed B not being polymerized, it is judged that whether these two text filed A and B can be polymerized, if can be polymerized, then
These two text filed A and B are polymerized, obtain text filed C, then from other be not polymerized text filed choose one
The individual text filed D not being polymerized, it is judged that whether described text filed C and D can be polymerized, if can be polymerized, then by these two
Text filed C and D is polymerized, repeat above-mentioned choose text filed, judge whether the step that can be polymerized and be polymerized until
Be not polymerized text filed choose complete.
Method the most according to claim 4, it is characterised in that described judge whether these two text filed A and B can gather
Close, including:
S40, the vertical dimension calculated between these two text filed A and B, it is judged that whether described vertical dimension is less than the tenth number
Value, if less than described tenth numerical value, then performs step S41, otherwise, performs step S42；
The minimum of the text filed pixel comprised that the abscissa of pixel is bigger in S41, calculating text filed A and B of said two
The horizontal seat of maximum of the text filed pixel comprised that abscissa is less with the abscissa of pixel in text filed A and B of said two
The absolute value of target difference, it is judged that whether described absolute value is less than the 11st numerical value, if less than described 11st numerical value, then by institute
State bigger text filed of the abscissa of pixel to be aggregated in after less text filed of the abscissa of described pixel；
S42, judge described vertical dimension whether less than the 11st numerical value, if less than described 11st numerical value, then calculating described two
The minimum abscissa of a text filed pixel comprised and another text filed pixel comprised in individual text filed A and B
The absolute value of difference of minimum abscissa, it is judged that whether described absolute value less than the tenth numerical value, if less than described tenth numerical value,
Then by less for the vertical coordinate of pixel in text filed for said two A and B text filed be aggregated in said two text filed in
Below bigger text filed of the vertical coordinate of pixel.
6. the device for medical image Chinese version region detection, it is characterised in that including:
Acquiring unit, for obtaining medical image to be detected；
Computing unit, for detecting described medical image, obtains a series of connected region, and based on single sample figure
As obtaining the two-value template in described medical image Chinese version region；
Filter unit, obtain text candidates for the non-textual region utilizing described two-value template to filter out in described connected region
Region, and further filter out the non-textual region in described text candidates region based on character feature；
Polymerized unit, for text filed being polymerized obtain, obtains line of text.
Device the most according to claim 6, it is characterised in that described computing unit, specifically for:
The local auto-adaptive calculating described single sample image R returns core KR, and for each connected region T, calculate this connection
The local auto-adaptive of region T returns core KT；
To described KRIt is normalized and obtains weight vector matrix WR, to described KTBe normalized obtain weights to
Moment matrix WT；
To described WRUse PCA algorithm to process, obtain main constituent, and the front d item retaining described main constituent constitutes matrix PR,
By described WRTo described PRProject, obtain characteristic vector F of described single sample image RR, by described WTTo described PRThrow
Shadow, obtains characteristic vector F of described connected region TT, wherein, described d is integer；
Calculate described characteristic vector FRWith FTBetween similarity, it is judged that whether described similarity measurement more than the first numerical value, if greatly
In described first numerical value, then the pixel value of corresponding connected region is set to 1, obtains text filed, otherwise, then by corresponding company
The pixel value in logical region is set to 0, obtains background area, will described text filed with background area as described two-value template.
8. according to the device described in claim 6 or 7, it is characterised in that described in filter unit, specifically for:
For each text candidates region, calculate stroke width feature SW of text candidate region, and retain stroke width
Feature SW is less than the text candidates region of second value, and wherein, the computing formula of described stroke width feature SW is
In formula, std and E is stroke width standard deviation and the meansigma methods of text candidate region respectively；
The number of the non-zero pixels in calculated text candidates region, filters the number of non-zero pixels more than third value and little
Text candidates region in the 4th numerical value；
The ratio of the number of the non-zero pixels in calculated text candidates region and the region area in corresponding text candidates region
Value, filters ratio and is more than the 5th numerical value and the text candidates region less than the 6th numerical value；
Calculated text candidates region length-width ratio, filter out length-width ratio more than the 7th numerical value with less than the 8th numerical value
Text candidates region；
For each the text candidates region obtained, utilize sciagraphy or connected region domain method that text candidate region is carried out
Cutting, obtains multiple fritter, and determines whether each fritter is character, and calculating is the ratio shared by the fritter of character, filters
Ratio is less than the text candidates region of the 9th numerical value.
Device the most according to claim 6, it is characterised in that described polymerized unit, specifically for the text for obtaining
Each text filed A not being polymerized in region, choose that other is not polymerized text filed in one be not polymerized
Text filed B, it is judged that whether these two text filed A and B can be polymerized, if can be polymerized, then by these two text filed A and
B is polymerized, and obtains text filed C, then from other be not polymerized text filed choose a text area not being polymerized
These two text filed C and D if can be polymerized, are then gathered by territory D, it is judged that whether described text filed C and D can be polymerized
Close, repeat above-mentioned choose text filed, judge whether that the step can be polymerized and be polymerized is until that is not polymerized is text filed
Choose complete.
Device the most according to claim 9, it is characterised in that described polymerized unit, specifically for:
Calculate the vertical dimension between these two text filed A and B, it is judged that whether described vertical dimension is less than the tenth numerical value, if
Less than described tenth numerical value, then calculate what bigger text filed of the abscissa of pixel in text filed A and B of said two comprised
The text filed pixel comprised that the minimum abscissa of pixel is less with the abscissa of pixel in text filed A and B of said two
The absolute value of difference of maximum abscissa, it is judged that whether described absolute value less than the 11st numerical value, if less than the described 11st
Numerical value, then by less for bigger for the abscissa of the described pixel text filed abscissa being aggregated in described pixel text filed after
Face；Or
If not less than described tenth numerical value, then judge whether described vertical dimension is less than the 11st numerical value, if less than the described tenth
One numerical value, then calculate the minimum abscissa of a text filed pixel comprised in text filed A and B of said two and another
The absolute value of the difference of the minimum abscissa of the text filed pixel comprised, it is judged that whether described absolute value is less than the tenth numerical value,
If less than described tenth numerical value, then text filed it is aggregated in less for the vertical coordinate of pixel in text filed for said two A and B
Below bigger text filed of the vertical coordinate of the text filed middle pixel of said two.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610648984.3A CN106296682B (en) | 2016-08-09 | 2016-08-09 | Method and device for detection text filed in medical image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610648984.3A CN106296682B (en) | 2016-08-09 | 2016-08-09 | Method and device for detection text filed in medical image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106296682A true CN106296682A (en) | 2017-01-04 |
| CN106296682B CN106296682B (en) | 2019-05-21 |
Family
ID=57667387
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610648984.3A Active CN106296682B (en) | 2016-08-09 | 2016-08-09 | Method and device for detection text filed in medical image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106296682B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107545262A (en) * | 2017-07-31 | 2018-01-05 | 华为技术有限公司 | A kind of method and device that text is detected in natural scene image |
| WO2018145470A1 (en) * | 2017-02-13 | 2018-08-16 | 广州视源电子科技股份有限公司 | Image detection method and device |
| CN113673523A (en) * | 2021-10-22 | 2021-11-19 | 北京世纪好未来教育科技有限公司 | Text detection method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102938062A (en) * | 2012-10-16 | 2013-02-20 | 山东山大鸥玛软件有限公司 | Document image slant angle estimation method based on content |
| US20130294696A1 (en) * | 2012-05-04 | 2013-11-07 | Fujitsu Limited | Image processing method and apparatus |
-
2016
- 2016-08-09 CN CN201610648984.3A patent/CN106296682B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130294696A1 (en) * | 2012-05-04 | 2013-11-07 | Fujitsu Limited | Image processing method and apparatus |
| CN102938062A (en) * | 2012-10-16 | 2013-02-20 | 山东山大鸥玛软件有限公司 | Document image slant angle estimation method based on content |
Non-Patent Citations (2)
| Title |
|---|
| 刘新瀚等: "自然场景下基于连通域检测的文字识别算法研究", 《计算机技术与发展》 * |
| 宋丽: "自然场景图像中文本信息检测方法的研究", 《万方硕士学位论文全文数据库》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018145470A1 (en) * | 2017-02-13 | 2018-08-16 | 广州视源电子科技股份有限公司 | Image detection method and device |
| CN107545262A (en) * | 2017-07-31 | 2018-01-05 | 华为技术有限公司 | A kind of method and device that text is detected in natural scene image |
| CN113673523A (en) * | 2021-10-22 | 2021-11-19 | 北京世纪好未来教育科技有限公司 | Text detection method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106296682B (en) | 2019-05-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP2095332B1 (en) | Feature-based registration of sectional images | |
| CN101742961B (en) | Diagnosis support device and system | |
| Kuhn et al. | Data pre-processing | |
| CN100566655C (en) | Be used to handle image to determine the method for picture characteristics or analysis candidate | |
| CN106170799A (en) | From image zooming-out information and information is included in clinical report | |
| US20150269314A1 (en) | Method and apparatus for unsupervised segmentation of microscopic color image of unstained specimen and digital staining of segmented histological structures | |
| CN104093354A (en) | Method and apparatus for assessment of medical images | |
| KR101645292B1 (en) | System and method for automatic planning of two-dimensional views in 3d medical images | |
| CN106296682A (en) | Method and device for medical image Chinese version region detection | |
| Corral Acero et al. | SMOD-data augmentation based on statistical models of deformation to enhance segmentation in 2D cine cardiac MRI | |
| CN110930414A (en) | Lung region shadow marking method and device of medical image, server and storage medium | |
| Peiffer et al. | A novel method for quantifying spatial correlations between patterns of atherosclerosis and hemodynamic factors | |
| Wang et al. | A novel cortical thickness estimation method based on volumetric Laplace–Beltrami operator and heat kernel | |
| Tanner et al. | Quantitative evaluation of free‐form deformation registration for dynamic contrast‐enhanced MR mammography | |
| Ginley et al. | Neural network segmentation of interstitial fibrosis, tubular atrophy, and glomerulosclerosis in renal biopsies | |
| Tward | An optical flow based left-invariant metric for natural gradient descent in affine image registration | |
| CN110517300A (en) | Elastic image registration algorithm based on partial structurtes operator | |
| Liu et al. | CAM‐Wnet: An effective solution for accurate pulmonary embolism segmentation | |
| Li et al. | Semi-automatic multiparametric MR imaging classification using novel image input sequences and 3D convolutional neural networks | |
| CN114937044A (en) | Lightweight image segmentation method and device and storage medium | |
| Mikula et al. | Finite volume schemes for the generalized subjective surface equation in image segmentation | |
| Xiong et al. | Mapping mouse brain slice sequence to a reference brain without 3D Reconstruction | |
| Raina | Energy-efficient circuits and systems for computational imaging and vision on mobile devices | |
| CN109741833A (en) | A kind of method and apparatus of data processing | |
| CN109858513A (en) | Brain cognitive ability measurement method based on the multiple dimensionality reduction of ectocinerea white matter morphological feature |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 100080 Beijing Haidian District Gaolizhang Road 18 Building 103-86 Patentee after: Beijing medical pat Intelligent Technology Co., Ltd. Address before: 100085 room 3, building 8, Chuang Chuang Road, Haidian District, Beijing (five story), room 3-7, -839. Patentee before: BEIJING HAOYUNDAO INFORMATION TECHNOLOGY CO., LTD. |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210715 Address after: 310018 22nd floor, building 1, 199 Yuancheng Road, Xiasha street, Hangzhou Economic and Technological Development Zone, Zhejiang Province Patentee after: Hangzhou Zhuojian Information Technology Co.,Ltd. Address before: 100080 Beijing Haidian District Gaolizhang Road 18 Building 103-86 Patentee before: BEIJING MEDP.AI INTELLIGENT TECHNOLOGY Co.,Ltd. |