CN104182749B - Image processing apparatus, image processing method and electronic equipment - Google Patents
Image processing apparatus, image processing method and electronic equipment Download PDFInfo
- Publication number
- CN104182749B CN104182749B CN201310187902.6A CN201310187902A CN104182749B CN 104182749 B CN104182749 B CN 104182749B CN 201310187902 A CN201310187902 A CN 201310187902A CN 104182749 B CN104182749 B CN 104182749B
- Authority
- CN
- China
- Prior art keywords
- shape
- profile
- predetermined
- fluctuation characteristic
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
Abstract
The present invention provides image processing apparatus, image processing method and electronic equipment, the problem of to overcome contours extract precision and/or the relatively low degree of accuracy present in existing outline extraction technique.Above-mentioned image processing apparatus includes:Feature extraction unit, shape facility and/or fluctuation characteristic for extracting the pending profile of destination object in image;And rejection processing unit, for refuse in the case of at least one in meeting following condition using pending profile as destination object contour detecting result:Similarity between the shape facility and predetermined shape model of pending profile is less than the second similarity threshold less than the similarity between the first similarity threshold, and the fluctuation characteristic and predetermined fluctuation characteristic model of pending profile.The above-mentioned technology of the present invention can be applied to image processing field.
Description
Technical field
The present invention relates to image processing field, more particularly to image processing apparatus, image processing method and electronic equipment.
Background technology
As the sharp increase of digital picture number is, it is necessary to research and develop effective image processing techniques.Digital picture one
As refer to for example, by the equipment such as digital camera, scanner capture image, can also be synthesized by arbitrary non-picture data
And obtain, such as by mathematical function.
In some existing image processing techniques, in order to recognize or detection image included in destination object(Such as people
Face, document, automobile etc.), the profile of the destination object can be generally extracted first, and then figure is finally determined according to its profile
Destination object as in.
However, due to the influence of some factors(Influence or the shadow of algorithm errors such as the interior perhaps ambient noise in image
Ring etc.), the precision for the profile that above-mentioned prior art is extracted and/or the degree of accuracy be often relatively low, second-rate(Such as deform serious
Deng).For example, there may exist distortion for digital picture obtained from the capture by equipment such as digital camera, scanners, inclining
Oblique or other deformations, and then cause the profile extracted by existing outline extraction technique to there is more serious distortion or change
Shape.
The content of the invention
The brief overview on the present invention is given below, to provide on the basic of certain aspects of the invention
Understand.It should be appreciated that this general introduction is not the exhaustive general introduction on the present invention.It is not intended to determine the pass of the present invention
Key or pith, nor is it intended to limit the scope of the present invention.Its purpose only provides some concepts in simplified form,
In this, as the preamble in greater detail discussed later.
In consideration of it, the invention provides image processing apparatus, image processing method and electronic equipment, it is existing at least to solve
The problem of contours extract precision present in some outline extraction techniques and/or the relatively low degree of accuracy.
According to an aspect of the invention, there is provided a kind of image processing apparatus, the image processing apparatus includes:Feature is carried
Unit is taken, it is arranged to the shape facility and/or fluctuation characteristic that extract the pending profile of destination object in image;And
Rejection processing unit, it is arranged to refuse to make pending profile in the case of at least one in meeting following condition
For the contour detecting result of destination object:Similarity between the shape facility and predetermined shape model of pending profile is less than the
Similarity between one similarity threshold, and the fluctuation characteristic and predetermined fluctuation characteristic model of pending profile is less than the second phase
Like degree threshold value;Wherein, predetermined shape model and/or predetermined fluctuation characteristic model include the training of multiple predetermined profiles by study
Sample and obtain, predetermined shape model reflection training sample shape facility the regularity of distribution, predetermined fluctuation characteristic model reflection
The regularity of distribution of the fluctuation characteristic of training sample, predetermined profile is identical with the type of pending profile.
According to another aspect of the present invention, a kind of image processing method is additionally provided, the image processing method includes:Carry
Take the shape facility and/or fluctuation characteristic of the pending profile of destination object in image;And in following condition is met extremely
Refuse in the case of few one using pending profile as destination object contour detecting result:The shape facility of pending profile
Fluctuation characteristic and pre- standing wave of the similarity less than the first similarity threshold, and pending profile between predetermined shape model
Similarity between dynamic characteristic model is less than the second similarity threshold;Wherein, predetermined shape model and/or predetermined fluctuation character modules
Type includes the training sample of multiple predetermined profiles by study and obtained, and predetermined shape model reflects the shape facility of training sample
The regularity of distribution, predetermined fluctuation characteristic model reflection training sample fluctuation characteristic the regularity of distribution, predetermined profile with it is pending
The type of profile is identical.
According to another aspect of the present invention, a kind of electronic equipment is additionally provided, the electronic equipment includes as described above
Image processing apparatus.
According to a further aspect of the invention, a kind of program production of instruction code for the machine-readable that is stored with is additionally provided
Product, said procedure product can make above-mentioned machine perform image processing method as described above upon execution.
In addition, according to other aspects of the invention, additionally provide a kind of computer-readable recording medium, be stored thereon with as
Upper described program product.
Above-mentioned image processing apparatus according to embodiments of the present invention, image processing method and electronic equipment, it is by carrying
The shape facility and/or fluctuation characteristic of the pending profile of destination object in image are taken by above-mentioned shape facility and/or fluctuation
Feature is respectively between corresponding predetermined shape model and/or predetermined fluctuation characteristic model carries out similarity-rough set, and wherein
At least one similarity results at least following benefit less than rejection pending profile in the case of corresponding predetermined threshold
One of:Judge to receive or the rejection profile by above-mentioned similarity-rough set, it is possible to increase contours extract precision and/or accurate
Degree;So that some precision present in final contour detecting result and/or the relatively low, and/or second-rate profile of the degree of accuracy
It is removed;And contribute to user recognized in the case where there are multiple pending profiles wherein precision and/or the degree of accuracy compared with
Low, and/or second-rate profile, and then precision, the degree of accuracy of these profiles can be improved by correspondingly subsequent treatment
And/or quality.
By the detailed description below in conjunction with accompanying drawing to highly preferred embodiment of the present invention, these and other of the invention is excellent
Point will be apparent from.
Brief description of the drawings
The present invention can be by reference to being better understood, wherein in institute below in association with the description given by accompanying drawing
Have and used same or analogous reference in accompanying drawing to represent same or similar part.The accompanying drawing is together with following
Describe the part for including in this manual and being formed this specification together in detail, and for this is further illustrated
The preferred embodiment of invention and the principle and advantage for explaining the present invention.In the accompanying drawings:
Fig. 1 is the frame for schematically showing a kind of exemplary construction of image processing apparatus according to an embodiment of the invention
Figure;
Fig. 2 is a kind of frame for the possible exemplary construction for schematically showing feature extraction unit 110 as shown in Figure 1
Figure;
Fig. 3 A are the schematic diagrames for a possible example for showing that pending profile is document boundaries;
Fig. 3 B are the schematic diagrames for a possible example for showing the shape point in document boundaries as shown in Figure 3A;
Fig. 3 C are to show to determine that first fluctuates each in characteristic sequence in the case of document boundaries as shown in Figure 3A
The schematic diagram of one possible example of line segment distance;
Fig. 3 D are to show to fluctuate each line segment distance in characteristic sequence for determination first in the case of not closing in profile
A possible example schematic diagram;
Fig. 4 is the frame for schematically showing another exemplary construction of image processing apparatus according to an embodiment of the invention
Figure;
Fig. 5 is the frame for schematically showing another exemplary construction of image processing apparatus according to an embodiment of the invention
Figure;
Fig. 6 A-6D are the schematic diagrames for showing to carry out predetermined profile multiple possible examples of the deformation of predefined type;
Fig. 7 is the stream for schematically showing a kind of exemplary process of image processing method according to an embodiment of the invention
Cheng Tu;And
Fig. 8 be show can be used to realize image processing apparatus according to an embodiment of the invention and image processing method,
Or a kind of possible message processing device of image processing apparatus and image processing method according to an embodiment of the invention
The structure diagram of hardware configuration.
It will be appreciated by those skilled in the art that element in accompanying drawing is just for the sake of showing for the sake of simple and clear,
And be not necessarily drawn to scale.For example, the size of some elements may be exaggerated relative to other elements in accompanying drawing, with
Just it is favorably improved the understanding to the embodiment of the present invention.
Embodiment
The one exemplary embodiment of the present invention is described hereinafter in connection with accompanying drawing.For clarity and conciseness,
All features of actual embodiment are not described in the description.It should be understood, however, that developing any this actual implementation
Many decisions specific to embodiment must be made during example, to realize the objectives of developer, for example, symbol
Those restrictive conditions related to system and business are closed, and these restrictive conditions may have with the difference of embodiment
Changed.In addition, it also should be appreciated that, although development is likely to be extremely complex and time-consuming, but to having benefited from the disclosure
For those skilled in the art of content, this development is only routine task.
Herein, in addition it is also necessary to which explanation is a bit, in order to avoid having obscured the present invention because of unnecessary details, in the accompanying drawings
It illustrate only and according to the closely related apparatus structure of the solution of the present invention and/or process step, and eliminate and the present invention
The little other details of relation.
The embodiment provides a kind of image processing apparatus, the image processing apparatus includes:Feature extraction unit,
It is arranged to the shape facility and/or fluctuation characteristic that extract the pending profile of destination object in image;And at rejection
Unit is managed, it is arranged to refuse to regard pending profile as target in the case of at least one in meeting following condition
The contour detecting result of object:Similarity between the shape facility of pending profile and predetermined shape model is similar less than first
The similarity spent between threshold value, and the fluctuation characteristic and predetermined fluctuation characteristic model of pending profile is less than the second similarity threshold
Value;Wherein, predetermined shape model and/or predetermined fluctuation characteristic model by study include multiple predetermined profiles training sample and
Obtain, the regularity of distribution of the shape facility of predetermined shape model reflection training sample, predetermined fluctuation characteristic model reflection training sample
The regularity of distribution of this fluctuation characteristic, predetermined profile is identical with the type of pending profile.
An example of image processing apparatus according to an embodiment of the invention is described in detail with reference to Fig. 1.
As shown in figure 1, image processing apparatus 100 includes feature extraction unit 110 and refused according to an embodiment of the invention
Know processing unit 120.
Feature extraction unit 110 is used to extract the shape facility of the pending profile of destination object and/or fluctuation in image
Feature.
In the specific implementation of image processing apparatus according to an embodiment of the invention, in image mentioned here
Destination object can be document(Pending profile is, for example, document boundaries in this case), face(Treat in this case
It is, for example, facial contour to handle profile), palm(Pending profile is, for example, palm profile in this case)Deng.Wherein, on
It for example can be the image captured for destination object to state image, or obtain by other means include above-mentioned target
Image of object etc..
In addition, in the specific implementation of image processing apparatus according to an embodiment of the invention, mesh in above-mentioned image
The pending profile of mark object can be for example obtained ahead of time by existing any outline extraction technique.
In the specific implementation of image processing apparatus according to an embodiment of the invention, in above-mentioned pending profile
" profile " for example can be document boundaries, that is, above-mentioned pending profile can be pending document boundaries.However, above-mentioned
" profile " is not limited to this, in other embodiments of the invention, " profile " can also be such as facial contour, palm profile with
And the profile in addition to document boundaries such as automobile profile.
It should be noted that in the specific implementation of image processing apparatus according to an embodiment of the invention, waiting to locate
Reason profile can be the pending profile that one section of continuous profile or several sections of discrete profiles are constituted;In addition, waiting to locate
Reason profile or part therein can be closed or not close.
Describe how to extract below in conjunction with Fig. 2 in image the shape facility of the pending profile of destination object and/or
Fluctuation characteristic.
Wherein, in the specific implementation of image processing apparatus according to an embodiment of the invention, feature extraction unit
110 can only include the shape spy for being used to extract the shape facility of the pending profile of destination object in image as shown in Figure 2
Levy and obtain subelement 210 and obtain son for the fluctuation characteristic for extracting the fluctuation characteristic of the pending profile of destination object in image
Any of which in unit 220, or, feature extraction unit 110 can also include above-mentioned shape facility simultaneously and obtain son list
Member 210 and fluctuation characteristic obtain subelement 220.Hereinafter, with feature extraction unit 110 simultaneously including shape as shown in Figure 2
Feature obtains subelement 210 in case of obtaining subelement 210 and fluctuation characteristic acquisition subelement 220 to describe shape facility
Processing and the function of subelement 220 are obtained with fluctuation characteristic, shape facility is only included for feature extraction unit 110 and obtains son list
The situation of one of them that member 210 and fluctuation characteristic are obtained in subelement 220 may be referred to the example to be handled, will no longer
Repeat.
As shown in Fig. 2 in one example, feature extraction unit 110 can include shape facility and obtain the He of subelement 210
Fluctuation characteristic obtains subelement 220.
Wherein, shape facility obtain subelement 210 can using multiple shape points on pending profile coordinate, according to
Second predefined procedure formation first shape characteristic sequence, is used as the shape facility of above-mentioned pending profile.
It should be noted that the shape point on some profile refer to over the outline according to certain rule selection can be anti-
Reflect some points of the contour shape.
In one implementation, on above-mentioned pending profile, can the pending profile neighboring reference point it
Between part, choose the point of predetermined quantity at predetermined intervals, and using selected point as the pending profile multiple shapes
Point.
Wherein, the reference point of pending profile can be determined according to actual conditions.For example, being hand for pending profile
The situation of profile is slapped, its corresponding reference point can be the point on each finger top and the connecting place of each two adjacent finger
Point, etc..And for example, it is the situation of some document boundaries for pending profile, its corresponding reference point can be this article flange
Multiple angle points of document corresponding to boundary, and these angle points are borderline positioned at the document.In example as shown in Figure 3A
In, the border of the document of the opening shown in Fig. 3 A(That is document boundaries, the outward flange S of document as shown in Figure 3A153And S264)'s
Reference point can for example choose the page angle point of document(Such as point P1, P2, P3 and P4), and/or document spine angle point(Such as Fig. 3 A institutes
The point P5 and P6 shown).Wherein, the file and picture shown in Fig. 3 A for example can be obtained by being scanned through by the books to opening
Image.
Assuming that in one example(Hereinafter referred to as example one), in the document boundaries S of the document shown in Fig. 3 A153And S264
(It is used as the example of pending profile)Reference point includes P1, P2, P3, P4, P5 and P6 determined by upper, according to the adjacent ginseng of each two
4 points are equably determined between examination point.As shown in Figure 3 B, for example can be equally spaced between adjacent reference point P1 and P5
4 points Pa1, Pa2, Pa3 and Pa4 are chosen, and can equally spaced choose 4 points between adjacent reference point P5 and P3
Pa5, Pa6, Pa7 and Pa8, etc., so choose 16 point Pa1~Pa16, then regard this 16 point Pa1~Pa16 as figure altogether
Document boundaries S shown in 3A153And S264Shape point.
In another example(Hereinafter referred to as example two), selected shape point can also be including reference point in itself.Such as
Shown in Fig. 3 B, between adjacent reference point P1 and P5, the shape point of selection can for example include Pa1, Pa2, Pa3 and Pa4 with
And reference point P1 and P5 totally 6 shape points in itself;And the shape point chosen between adjacent reference point P5 and P3 for example can be with
Including Pa5, Pa6, Pa7 and Pa8 and reference point P5 and P3 totally 6 shape points in itself;Etc..Wherein, in multiple selected shape
Selected same shape point is only calculated as a point during point, for example, reference point P5 is chosen during above-mentioned selection
Take twice, but be only used as a shape point.
Then, shape facility obtains subelement 210 and can utilized in document boundaries S153And S264Multiple shapes of upper selection
Point coordinate, first shape characteristic sequence is formed according to the second predefined procedure.Wherein, the second predefined procedure mentioned here can
To be determined according to actual conditions, such as it can be order from left to right(Starting point is, for example, first left point), on to
Under order(Starting point is, for example, the point of top first), or clockwise, sequence counter-clockwise(Starting point is, for example, apart from some reference
The nearest point of point), etc..It should be noted that rising in first shape characteristic sequence can be determined according to the second predefined procedure
Point.Thus, in actual applications, user can determine that above-mentioned second is predetermined suitable rule of thumb or by the method for experiment
Sequence.
For example, in example two, selected same shape point is only calculated as one during multiple selected shape point
In the case of individual point, obtained multiple shape points amount to 22, i.e. Pa1~Pa16 and P1~P6.As shown in Figure 3 B, Ke Yifen
Not in S153And S264On according to order from left to right(Wherein, S153And S264Between can be according to first S153S afterwards264Order)Come
Form above-mentioned first shape characteristic sequence, i.e. { CP1,CPa1,CPa2,CPa3,CPa4,CP5,CPa5,CPa6,CPa7,CPa8,CP3,CP2,
CPa9,CPa10,CPa11,CPa12,CP6,CPa13,CPa14,CPa15,CPa16,CP4, wherein, CP1、CPa1、CPa2... represent that its is right respectively
The shape point P1 that answers, Pa1, Pa2 ... coordinate.
It should be noted that the coordinate of shape point can be two-dimentional, can also be specific dimension more than three-dimensional or three-dimensional
Number can be determined according to actual conditions.For example, in the above-mentioned example one and two with reference to described by Fig. 3 A and Fig. 3 B, each shape
The coordinate of shape point is two-dimensional coordinate, that is to say, that CP1、CPa1、CPa2... in each is actual comprising two coordinate values.Again
Such as, the border in some medical images produced by three-dimensional body segmentation is probably three-dimensional, and the cut surface of such as certain internal organs is
Three-dimensional situation.
In addition, as shown in Fig. 2 fluctuation characteristic, which obtains subelement 220, can be primarily based on multiple shapes on pending profile
Shape point determines multiple points pair, is used as first point pair.Then, fluctuation characteristic, which obtains subelement 220, can obtain above-mentioned multiple the
The line segment distance of the point-to-point transmission of some each first point pair of centerings, utilizes these distances(I.e. above-mentioned multiple first points to corresponding
All line segment distances), the first fluctuation characteristic sequence is formed according to the first predefined procedure, be used as the fluctuation of pending profile special
Levy.
It should be noted that in actual treatment, the influence in order to avoid different document size to extraction fluctuation characteristic can
To carry out length normalization method processing to fluctuation characteristic sequence, and the detailed process of the normalized is for those skilled in the art
For can be by combining the mode of general knowledge known in this field and/or open source information to obtain, therefore I will not elaborate.
Wherein, multiple shape points on pending profile can be according to the side as described in example one above or example two
Formula is determined, is here repeated no more.In addition, the first predefined procedure mentioned here can be determined according to actual conditions, such as
It can be order from left to right(Starting point is, for example, first left line segment distance), or order starting point from top to bottom(Example
Such as it is the line segment distance of top first)Etc..It should be noted that the first fluctuation characteristic can be determined according to the first predefined procedure
Line segment distance in sequence, as starting point.Similarly, in actual applications, user can rule of thumb or pass through experiment
Method determine above-mentioned first predefined procedure.
In the specific implementation of image processing apparatus according to an embodiment of the invention, fluctuation characteristic obtains subelement
220 can determine above-mentioned first point pair according to various ways.
In one example, fluctuation characteristic acquisition subelement 220 for example can be by multiple shape points on pending profile
Matched two-by-two according to predetermined matching method, it is then possible to using as multiple points of the above-mentioned result matched two-by-two to make
For above-mentioned multiple first points pairs.
As shown in Figure 3 C, for S153And S264On shape point, can be respectively by S153Upper and S264On pair from left to right
The shape point of order is answered to be matched, such as P1 and P2, Pa1 and Pa9, Pa2 and Pa10, etc..So, as shown in Figure 3 C, respectively
With L1, L2, L3 ... come represent the distance of the line segment between P1 and P2, line segment distance, Pa2 and Pa10 between Pa1 and Pa9 it
Between line segment distance ..., then be consequently formed first fluctuation characteristic sequence can for example be expressed as { L1, L2, L3 ... ... }.
From Fig. 3 C, the first resulting fluctuation characteristic sequence can reflect the document for example shown in Fig. 3 C well in this manner
Border(Or other have symmetry characteristic or can be divided into two-part pending profile)Fluctuation characteristic.
In other examples, in the case of the curve that pending profile is not closed for example, it is single that fluctuation characteristic obtains son
Member 220 can be by a pair of shape points on the pending profile(Such as two angle points of document boundaries, or the pending profile
Beginning and end)Between line as reference line, then by its on the pending profile in addition to above-mentioned a pair of shape points
He maps shape point to the reference line respectively, obtains the mapping point of each shape point in above-mentioned other shapes point, by it is above-mentioned its
Each shape point and its mapping point be respectively as one first point pair in his shape point, and by each shape point and its mapping point it
Between distance as this first point to corresponding line segment distance, to be formed using these line segment distances according to the first predefined procedure
First fluctuation characteristic sequence.
As shown in Figure 3 D, curve S1’2’For the pending profile do not closed, it is assumed that shape point thereon be P1 ', Pa1 ',
Pa2 ', Pa3 ' and P2 ', P1 ' and P2 ' be curve S1’2’Beginning and end, straight line l1’2’For P1 ' and P2 ' between line,
Pb1 ', Pb2 ' and Pb3 ' are respectively Pa1 ', Pa2 ' and Pa3 ' in straight line l1’2’On mapping point, L1 ', L2 ' and L3 ' are represented respectively
Pa1 ' to Pb1 ' line segment distance, Pa2 ' to Pb2 ' line segment distance and Pa3 ' arrive Pb3 ' line segment distance.So, fluctuate special
The first following fluctuation characteristic sequence can be obtained according to order from left to right by levying acquisition subelement 220:{L1’,L2’,
L3’}。
It should be noted that rule of thumb or by the method for experiment determined in actual treatment according to which kind of mode come
Pairing determines above-mentioned multiple first points pairs according to which kind of mode, so as to select a kind of appropriate mode to cause
Thus obtained first fluctuation characteristic sequence can preferably reflect the fluctuation characteristic of corresponding pending profile, and then can make
The result obtained subsequently is more accurate.
So, feature extraction unit 110 obtain extract image in destination object pending profile shape facility and/
Or after fluctuation characteristic, result that rejection processing unit 120 can be obtained with feature based extraction unit 110 is judged,
Determined whether using the result according to judgement refusal by current pending profile as destination object contour detecting result(I.e.
The above-mentioned pending profile of rejection is determined whether according to the result of judgement).
It should be noted that in the specific implementation of image processing apparatus according to an embodiment of the invention, rejection
Characteristic type that the result that processing unit 120 can be obtained according to feature extraction unit 110 is included is sentenced accordingly
It is disconnected.
In one example, the result obtained in feature extraction unit 110 is while the shape comprising pending profile is special
Seek peace in the case of two kinds of features of fluctuation characteristic, rejection processing unit 120 can judge the shape facility of pending profile respectively
With the similarity between predetermined shape model(Hereinafter referred to as the first similarity)Whether it is less than the first similarity threshold and waits to locate
Manage the similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of profile(Hereinafter referred to as the second similarity)Whether less than the
Two similarity thresholds.Hereinafter, the condition of " the first similarity be less than the first similarity threshold " is referred to as condition one, and incite somebody to action " the
Two similarities be less than the second similarity threshold " condition be referred to as condition two.
If at least one in condition one and condition two is satisfied(Including three kinds of situations:First similarity is less than the first phase
Seemingly spend threshold value but the second similarity is greater than or equal to the second similarity threshold;First similarity is greater than or equal to the first similarity threshold
It is worth but the second similarity is less than the second similarity threshold;First similarity is less than the first similarity threshold and the second similarity is less than
Second similarity threshold), then rejection processing unit 120 can refuse using the pending profile as destination object contour detecting
As a result.
If condition one and condition two are not satisfied(That is, the first similarity is greater than or equal to the first similarity threshold, and
Second similarity is greater than or equal to the second similarity threshold), then rejection processing unit 120 can receive to make the pending profile
For the contour detecting result of destination object, or other follow-up places can be performed to the pending profile in actual applications
Reason, etc..
In another example, only the shape comprising pending profile is special for the result obtained in feature extraction unit 110
In the case of levying, rejection processing unit 120 can be only judged between the shape facility of pending profile and predetermined shape model
Whether the first similarity is less than the first similarity threshold, if first similarity is less than the first similarity threshold, rejection processing
Unit 120 can refuse using the pending profile as destination object contour detecting result;If judged result is "No"(That is,
First similarity is higher than the first similarity threshold), then rejection processing unit 120 can receive to regard the pending profile as target
The contour detecting result of object, or other follow-up processing, etc. can be performed in actual applications.
In another example, only the fluctuation comprising pending profile is special for the result obtained in feature extraction unit 110
In the case of levying, rejection processing unit 120 can only judge the fluctuation characteristic of pending profile and predetermined fluctuation characteristic model it
Between the second similarity whether be less than the second similarity threshold, if second similarity be less than the second similarity threshold, rejection
Processing unit 120 can refuse using the pending profile as destination object contour detecting result;If judged result is "No"
(That is, the second similarity is higher than the second similarity threshold), then rejection processing unit 120 can receive using the pending profile as
The contour detecting result of destination object, or other follow-up processing, etc. can be performed in actual applications.
It should be noted that the first similarity threshold and/or the second similarity threshold can be determined based on experience value,
It can be determined by the method for experiment, I will not elaborate.
Above-mentioned predetermined shape model and/or predetermined fluctuation characteristic model can be obtained by being learnt to training sample
, training sample includes multiple predetermined profiles.
Wherein, predetermined shape model reflects the regularity of distribution of the shape facility of training sample, and predetermined fluctuation character modules
Type then reflects the regularity of distribution of the fluctuation characteristic of training sample.
In addition, multiple predetermined profiles mentioned here are and pending types of profiles identical profile.For example, pending
In the case that the type of profile is document boundaries, multiple predetermined profiles are also the document boundaries for describing same type document,
Wherein, it is all the books opened that same type document mentioned here, which for example refers to pending profile and predetermined profile, or waits to locate
It is all single-sheet stationery, etc. to manage profile and predetermined profile.And for example, it is many in the case where the type of pending profile is facial contour
Individual predetermined profile is also facial contour.
It should be noted that in the specific implementation of image processing apparatus according to an embodiment of the invention, making a reservation for
Shape and/or predetermined fluctuation characteristic model can be stored in advance in image processing apparatus(For example it is stored therein
Some memory cell in), or or pass through some built-in or some units(In following article with reference to described by Fig. 4
Shape obtaining unit 430 and/or fluctuation characteristic model obtaining unit 440)Come what is obtained in real time.
, wherein it is desired to which explanation, is to be stored in advance in image in predetermined shape model and/or predetermined fluctuation characteristic model
In the case of in processing unit, the above-mentioned predetermined shape model and/or predetermined fluctuation characteristic model prestored can be passed through in advance
Particular procedure(As similar to the processing performed by above-mentioned " some built-in unit ")And obtain.
In an implementation of image processing apparatus according to an embodiment of the invention, rejection processing unit 120
Such as can be in the distance between the shape facility and predetermined shape model of pending profile(Such as mahalanobis distance)Higher than first away from
During from threshold value, judge that the similarity between the shape facility and predetermined shape model of pending profile is less than the first similarity threshold
Value;And/or the distance between the fluctuation characteristic and predetermined fluctuation characteristic model of pending profile(Such as mahalanobis distance)It is high
When second distance threshold value, judge similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of pending profile less than the
Two similarity thresholds.
Wherein, the first distance threshold and/or second distance threshold value can for example be set based on experience value, or, also may be used
Determine, repeat no more here in the method by experiment.
It should be noted that the first similarity described above and the second similarity are except can be based on pending profile
The fluctuation characteristic of the distance between shape facility and predetermined shape model and pending profile and predetermined fluctuation characteristic model it
Between distance come outside obtaining, can also utilize it is of the prior art other be used for the method for similarity is calculated to obtain, here
Repeat no more.
The result obtained below with feature extraction unit 110 is while the shape facility comprising pending profile and fluctuation are special
Describe how to calculate the distance between shape facility and predetermined shape model of pending profile in case of levying, and treat
Handle the distance between fluctuation characteristic and predetermined fluctuation characteristic model of profile.The knot obtained for feature extraction unit 110
The situation of one of them in fruit only shape facility and fluctuation characteristic comprising pending profile may be referred to example progress, will
Repeat no more.
In this example embodiment, these multiple predetermined profiles first can be subjected to registration process, then, rejection processing unit 120
The covariance matrix related to predetermined shape model can be calculated according to formula below one first.It should be noted that right
It is that can combine common knowledge and/or open money to the processing that multiple profiles are alignd for those skilled in the art
Expect and obtain, therefore I will not elaborate(Alignment for example for multiple document boundaries may be referred to document:T.F.Cootes,
C.J.Taylor,D.H.Cooper,and J.Graham,”Active Shape Models–Their Training and
Application ", CVIU (1995), 61 (1), pp.38~59).
Formula one:
Wherein, SshapeThe covariance matrix related to predetermined shape model is represented, i represents the sequence number of predetermined profile(i=0,
1,…,N-1), N is the sum of multiple predetermined profiles, Seq'iRepresent corresponding to i-th in multiple predetermined profiles after alignment
The second shape facility sequence(Hereinafter by the description to the example for obtaining the second shape facility sequence),Represent
The sequence of average of the corresponding second shape facility sequence of predetermined profile after multiple alignment.For example, it is assumed that multiple predetermined profiles are common
Including three, and assume that the corresponding second shape facility sequence of predetermined profile is respectively Seq after these three alignment1’、Seq2' and
Seq3', then above three align rear profile the second shape facility sequence Seq1’、Seq2' and Seq3' sequence of average be
(Seq1’+Seq2’+Seq3’)/3.
In addition, rejection processing unit 120 can calculate related to predetermined fluctuation characteristic model according to formula below two
Covariance matrix.
Formula two:
Wherein, SwaveRepresent withThe related covariance matrix of predetermined fluctuation characteristic model, Seq'i' represent after alignment
Multiple predetermined profiles in i-th corresponding to the second fluctuation characteristic sequence(Hereinafter will be to special for obtaining the second fluctuation
Levy the description of the example of sequence),Represent the average value of the corresponding second fluctuation characteristic sequence of predetermined profile after multiple alignment
Sequence.Wherein, the process of the sequence of average of each self-corresponding second fluctuation characteristic sequence of predetermined profile after multiple alignment is obtained
For example can be with the average value sequence as described above for obtaining the corresponding second shape facility sequence of predetermined profile after multiple alignment
The process of row is similar, repeats no more here.
Thus, rejection processing unit 120 can calculate pending profile respectively according to formula below three and formula four
Shape facility and predetermined shape model between mahalanobis distance, and pending profile fluctuation characteristic and predetermined fluctuation feature
Mahalanobis distance between model.
Formula three:
Formula four:
Wherein, in formula three and formula four,For SshapeInverse matrix,For SwaveInverse matrix,For
For the first shape characteristic sequence for the shape facility for describing pending profile, the sequence with the average shape in training set
Registration process is carried out;First for the fluctuation characteristic for describing pending profile fluctuates characteristic sequence, the profile
Registration process has been carried out;DistshapeRepresent geneva between the shape facility of pending profile and predetermined shape model away from
From, and DistwaveRepresent the mahalanobis distance between the fluctuation characteristic of pending profile and predetermined fluctuation characteristic model.
In one example, rejection processing unit 120 may determine that DistshapeWhether default first distance threshold is higher than
And/or DistwaveWhether default second distance threshold value is higher than.If DistshapeHigher than the first distance threshold and/or Distwave
Higher than second distance threshold value, then rejection processing unit 120 is refused to examine the profile of current pending profile as destination object
Survey result;And if DistshapeLess than the first distance threshold and DistwaveLess than second distance threshold value, then rejection processing unit
120 can receive using current pending profile as destination object contour detecting result, or to current pending wheel
Exterior feature carries out other follow-up processing.
In addition, in another example, rejection processing unit 120 can calculate pending wheel according to formula below five
The first similarity between wide shape facility and predetermined shape model, and can be calculated according to formula below six and wait to locate
Manage the second similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of profile.
Formula five:Simshape=1/Distshape
Formula six:Simwave=1/Distwave
Wherein, SimshapeRepresent above-mentioned first similarity, and SimwaveRepresent above-mentioned second similarity.In such case
Under, rejection processing unit 120 may determine that SimshapeWhether default first similarity threshold and/or Sim are less thanwaveIt is whether low
In default second similarity threshold.If SimshapeLess than the first similarity threshold and/or SimwaveLess than the second similarity threshold
Value, then rejection processing unit 120 refuse using current pending profile as destination object contour detecting result;And if
SimshapeHigher than the first similarity threshold and SimwaveHigher than the second similarity threshold, then rejection processing unit 120 can receive
Current is handled into profile as the contour detecting result of destination object, or current pending profile can be carried out follow-up
Other processing.
By above description, above-mentioned image processing apparatus according to an embodiment of the invention, it is by extracting image
The shape facility and/or fluctuation characteristic of the pending profile of middle destination object divide above-mentioned shape facility and/or fluctuation characteristic
Not between corresponding predetermined shape model and/or predetermined fluctuation characteristic model carries out similarity-rough set, and wherein at least one
Individual similarity is less than rejection pending profile in the case of corresponding predetermined threshold, i.e. refusal regard pending profile as mesh
Mark the contour detecting result of object.In actual applications, when the utilization image processing apparatus is to utilizing existing contours extract skill
The profile that art is extracted(It is used as pending profile)When being handled, rejection can be judged whether by above-mentioned similarity-rough set
The profile.So, by above-mentioned processing, some precision present in final contour detecting result can be caused and/or accurate
Degree is relatively low, and/or second-rate(Such as deform serious etc.)Profile be removed, it is possible to increase contours extract precision and/or
The degree of accuracy.
In addition, in some actual treatments, user can also be chosen whether to these by rejection according to the result of rejection
The profile fallen carries out follow-up processing, contributes to user to be taken turns the problem of being recognized wherein in the case of there are multiple pending profiles
It is wide(That is, precision and/or the degree of accuracy are relatively low, and/or second-rate(Such as deform serious etc.)Profile), with by these quilts
The profile that rejection is fallen carries out precision, the degree of accuracy and/or the quality of corresponding subsequent treatment to improve these profiles, etc..
Another example of image processing apparatus according to an embodiment of the invention is described with reference to Fig. 4.Such as Fig. 4 institutes
Show, in this example, image processing apparatus 400 is in addition to including feature extraction unit 410 and rejection processing unit 420, also
Including model obtaining unit 430.Wherein, model obtaining unit 430 can include shape acquisition subelement 432 and/or ripple
Dynamic characteristic model obtains subelement 434.
Below in conjunction with the example described by Fig. 4, son will be obtained including shape with model obtaining unit 430 simultaneously
Unit 432 and fluctuation characteristic model obtain the He of subelement 432 to provide in case of obtaining subelement 434 on shape
Fluctuation characteristic model obtains the function of subelement 434 and the description of processing.It is to be noted, however, that in other examples, mould
Type obtaining unit 430 can also only include shape and obtain in subelement 432 and fluctuation characteristic model acquisition subelement 434
Any of which, processing in this case may be referred to example described below, will not be described in great detail.
It should be noted that the feature extraction unit 410 and rejection processing in image processing apparatus 400 shown in Fig. 4 are single
Member 420 can have respectively with above in conjunction with the feature extraction unit 110 in the image processing apparatus 100 described by Fig. 1 and
The identical 26S Proteasome Structure and Function of rejection processing unit 120, and similar technique effect can be reached, repeat no more here.
As shown in figure 4, for each in the training sample after alignment, shape obtains subelement 432 can profit
With the coordinate of multiple shape points on the training sample, corresponding with the training sample second is formed according to the second predefined procedure
Shape facility sequence.So, the corresponding second shape facility sequence of each training sample can be obtained.Then, shape is obtained
The average shape characteristic sequence and shape facility of the second shape facility sequence of these training samples can be obtained by obtaining subelement 432
Covariance matrix, to describe predetermined shape model.
In specific implementation according to an embodiment of the invention, shape, which obtains subelement 432, can for example lead to
The similar processing of the processing with obtaining subelement 210 above in conjunction with the shape facility described by Fig. 2 is crossed to obtain each instruction
Practice the corresponding second shape facility sequence of sample, can also refer to and obtain each above in association with the processing in the description of formula one
The sequence of average of the corresponding second shape facility sequence of training sample(That is average shape characteristic sequence)With shape facility association side
Poor matrix, and similar effect can be reached, repeat no more here.
In addition, for each in the training sample after alignment, fluctuation characteristic model obtains subelement 434 can be at this
Multiple second points pair are selected in multiple shape points on training sample, this corresponding multiple second point centering of the training sample are obtained
Each second point centering point-to-point transmission line segment distance, and using this multiple second point to corresponding multiple line segments distance,
The corresponding second fluctuation characteristic sequence of the training sample is formed according to the first predefined procedure.So, each training can be obtained
The corresponding second fluctuation characteristic sequence of sample.Then, fluctuation characteristic model, which obtains subelement 434, can obtain these training samples
The second fluctuation characteristic sequence average fluctuation characteristic sequence and fluctuation characteristic covariance matrix, to describe predetermined fluctuation character modules
Type.
In specific implementation according to an embodiment of the invention, fluctuation characteristic model obtains subelement 434 and for example may be used
With every to obtain by the similar processing of the processing with obtaining subelement 220 above in conjunction with the fluctuation characteristic described by Fig. 2
The corresponding second fluctuation characteristic sequence of individual training sample, can also be with reference to obtaining above in association with the processing in the description of formula two
The sequence of average of the corresponding second fluctuation characteristic sequence of each training sample(Averagely fluctuate characteristic sequence)And fluctuation characteristic
Covariance matrix, and can reach it is similar can and effect, repeat no more here.
By above description, model obtaining unit 430 can obtain above-mentioned predetermined shape model in real time and/or pre-
Standing wave moves characteristic model.In addition, above-mentioned model obtaining unit 430 can also be according to handled by it pending profile come selectivity
Ground is updated to above-mentioned predetermined shape model and/or predetermined fluctuation characteristic model, for example, in pending profile not by rejection
(Received)In the case of above-mentioned predetermined shape model and/or predetermined fluctuation character modules are updated using the pending profile
Type, can more reflect that user is to be dealt with thus, it is possible to the predetermined shape model and/or predetermined fluctuation characteristic model for update
The feature of profile, and then so that the place hereafter carried out using the predetermined shape model and/or predetermined fluctuation characteristic model of renewal
Result obtained by reason is more accurate.
Another example of image processing apparatus according to an embodiment of the invention is described in detail with reference to Fig. 5.Such as
Shown in Fig. 5, in this example, image processing apparatus 500 except including feature extraction unit 510 and rejection processing unit 520 it
Outside, pretreatment unit 540 can also be included.
It should be noted that the feature extraction unit 510 and rejection processing in image processing apparatus 500 shown in Fig. 5 are single
Member 520 can have respectively with above in conjunction with the feature extraction unit 110 in the image processing apparatus 100 described by Fig. 1 and
The identical 26S Proteasome Structure and Function of rejection processing unit 120, and similar technique effect can be reached, repeat no more here.
In one implementation, pretreatment unit 540 can be by carrying out at least part in multiple predetermined profiles
The deformation of predefined type obtains extension sample, and by the extension sample obtained together with above-mentioned multiple predetermined profiles together as
Training sample(Sample will be extended as the part in training sample).
With any one predetermined profile S at least part in above-mentioned multiple predetermined profiles0Exemplified by it is single to describe pretreatment
540 couples of predetermined profile S of member0Carry out an example of the deformation of predefined type.In this example, pretreatment unit 540 can be with
By making predetermined profile S0In each shape point N-dimensional coordinate at least one-dimensional coordinate increase or reduce within a predetermined range with
Formed and update N-dimensional coordinate so that by predetermined profile S0In the shape that is constituted of renewal N-dimensional coordinate of each shape point can be used in
Description and the profile of pending profile same type.
It should be noted that N here is the integer more than or equal to 2.For example, for pending profile and predetermined profile
It is the situation of document boundaries, N can be 2.And for example, when pending profile and predetermined profile are facial contour, N
Can be 2 or 3.
In addition, in one implementation, still with predetermined profile S0Exemplified by, to predetermined profile S0The process deformed
In, pretreatment unit 540 can keep predetermined profile S0In each shape point N-dimensional coordinate at least one-dimensional coordinate it is constant,
And make predetermined profile S0In each shape point N-dimensional coordinate in except it is above-mentioned it is at least one-dimensional in addition to remaining dimension coordinate in preset range
Interior increase reduces to form predetermined profile S0In each shape point renewal N-dimensional coordinate.
For example, it is assumed that predetermined profile S0In the coordinate of each shape point be two-dimentional, use x0Represent predetermined profile S0In each
Arbitrary shape point p in shape point0Abscissa, y0Represent above-mentioned shape point p0Ordinate, then for shape point p0, Ke Yigen
Come according to formula below seven to shape point p0Abscissa deformed, or can be according to formula below eight come to shape point
p0Ordinate deformed, or, can also according to formula seven and formula eight to shape point p0Horizontal stroke, ordinate all divides
Do not deformed.
Formula seven:x0’=x0+dx
Formula eight:y0’=y0+dy
Wherein, dx is the variable quantity of abscissa(Can be positive number or negative), dy is the variable quantity of ordinate(Can be for just
Number or negative), and dx and dy can determine based on experience value or by the method for experiment, repeat no more here.In addition,
x0' represent the abscissa value after deformation, y0' represent the ordinate value after deformation.
In one example, with predetermined profile S0Exemplified by, for some shape point thereon, the horizontal seat of the shape point
Mark and ordinate be not equal proportion deformation, with cause by deformation obtain extension sample shape and deformation before shape will
Occur certain change, but the type of shape again can be essentially identical.
It should be noted that the type " identical " of shape mentioned here or " essentially identical " for example refer to, rectangular profile becomes
Still it is rectangular profile after shape(Include square contour), still it is ellipse profile after ellipse profile deformation(Include circular contour),
Etc..
In addition, in some other implementation, image processing apparatus 500 except can include feature extraction unit 510,
Outside rejection processing unit 520 and pretreatment unit 540, it is also an option that property include model obtaining unit 530.Wherein,
Model obtaining unit 530 for example can have with above in conjunction with any model obtaining unit 430 described by Fig. 4(I.e. only
Subelement 432 is obtained including shape and fluctuation characteristic model obtains the model acquisition of any of which in subelement 434
Unit 430, or the model for obtaining subelement 432 and fluctuation characteristic model acquisition subelement 434 including shape simultaneously are obtained
Obtain unit 430)Identical 26S Proteasome Structure and Function, and similar technique effect can be reached, repeat no more here.
In actual treatment, often there is a situation where the limited generation of training sample number, in such a case, it is possible to utilize
Existing training sample(That is described multiple predetermined profiles above)To expand the quantity of training sample, this is enabled to according to expansion
The model obtained by training sample after filling(That is predetermined shape model and/or predetermined fluctuation characteristic model)Generalization ability compare
By force so that the result that the processing carried out using the model is obtained is also more accurate.
It should be noted that image processing apparatus 500 only include feature extraction unit 510, rejection processing unit 520 with
And in the case of pretreatment unit 540, pretreatment unit 540 is extended to the deformation that predetermined profile carries out predefined type
The processing of sample can be used for predetermined shape model and/or predetermined fluctuation characteristic model being stored to image processing apparatus 500 in advance
In before in the processing of training sample, after the sample that is expanded, using these extension samples training sample is expanded
Exhibition, then using the training sample after extension, obtained by described above " particular procedure " above-mentioned predetermined shape model with/
Or predetermined fluctuation characteristic model, and then above-mentioned predetermined shape model and/or predetermined fluctuation characteristic model are pre-stored in image procossing
In device 500.
In addition, in image processing apparatus 500 except including feature extraction unit 510, rejection processing unit 520 and pre- place
Manage outside unit 540 in addition in the case of model obtaining unit 530, pretreatment unit 540 then can be by as described above
Extension sample is obtained to the processing of the deformation of predetermined profile progress predefined type, with spread training sample, then, model is obtained
Unit 530 can obtain above-mentioned predetermined shape model and/or predetermined fluctuation feature in real time using the training sample after extension
Model.
Below, feature extraction unit 510, rejection processing unit 520, model are included with image processing apparatus 500 and obtains single
In case of member 530 and pretreatment unit 540, example, image procossing dress are applied to describe one of pretreatment unit 540
Put 500 only including feature extraction unit 510, rejection processing unit 520 and pretreatment unit 540 situations may be referred to should answer
Handled, be will not be described in great detail with example.
Describe to carry out predetermined profile the expansion in horizontal direction by taking Fig. 6 A as an example(Show as the deformation of predefined type
Example)Exemplary process.For document boundaries S as shown in Figure 6A153(Not shown in figure)For, P1, Pa1 ..., Pa4, P5
Deng the position where the shape point represented before deformation, P1 ', Pa1 ' ..., Pa4 ', P5 ' etc. carry upper right footmark " ' " reference
Represent respectively P1, Pa1 ..., the position where the shape point after the deformation corresponding to the shape point such as Pa4, P5.
The linear equation of the coordinate by 2 points of P1 and P3 can be calculated according to formula nine first, wherein,(X1, y1)Table
Show P1 coordinate,(X3, y3)Represent P3 coordinate.
Formula nine:dA*x+dB*y+dC=0
Wherein, dA, dB and dC are the parameter of linear equation.
Assuming that the offset in default horizontal direction is dx.If it should be noted that dx>0, document shape will be carried out
Expansion, as shown in Figure 6A;If dx<0, document shape will be shunk, as shown in Figure 6B.In addition, it should also be noted that, being
For the sake of clear, not shown in Fig. 6 B and hereinafter Fig. 6 C and Fig. 6 D by description or each shape point is not entirely shown
Reference.
In the case of as shown in Figure 6A, the corresponding changes of P1 and 2 points of P3 can be calculated according to equation below ten and 11
2 points of respective abscissas of shape point P1 ' and P3 ' after shape.
Formula ten:X1'=x1-dx
Formula 11:X3'=x3+dx
Then, according to formula nine(Now known to dA, dB and dC)The respective ordinate of ' two point to obtain P1 ' and P3.
For the other shapes point between P1 and P3(Including Pa1, Pa2, Pa3, Pa4, P5, Pa5, Pa6, Pa7 and Pa8),
Each point pair in the other shapes point between above-mentioned P1 and P3 can be obtained by interpolation arithmetic according to formula below 12
The abscissa of shape point after the deformation answered.Wherein, for clarity, in formula below 12, P1 will be represented with aa
Abscissa x1, P3 abscissa x3 is represented with bb, P1 ' abscissa x1 ' is represented with cc, and P3 ' is represented with dd
Abscissa x3 '.
Formula 12:xj’=cc+(dd-cc)*(xj-aa)/(bb-aa)
Wherein, xjFor any shape point of the every other shape point between P1 and P3, j be the arbitrary shape point in P1 and
The sequence number of every other shape point between P3.For example, in example as shown in Figure 6A, j=1,2 ... ..., 9, where it is assumed that
The orders of j from small to large correspond to every other shape point between P1 and P3 according to order from left to right, then:x1Represent
Pa1 abscissa, x1' then represent Pa1 deformation after shape point Pa1 ' abscissa;x2Represent Pa2 abscissa, x2' then represent
The abscissa of shape point Pa2 ' after Pa2 deformations;……;x5Represent P5 abscissa, x5' then represent the shape point after P5 deformations
P5 ' abscissa;x6Represent Pa5 abscissa, x6' then represent Pa5 deformation after shape point Pa5 ' abscissa;……;x9Table
Show Pa8 abscissa, x9' then represent Pa8 deformation after shape point Pa8 ' abscissa.
In addition, as shown in Figure 6A, the expansion in horizontal direction is being carried out to predetermined profile(Or the level side shown in Fig. 6 B
Upward contraction)In the case of, the other shapes point between P1 and P3(Including Pa1, Pa2, Pa3, Pa4, P5, Pa5, Pa6, Pa7
And Pa8)Ordinate can keep constant, as shown in formula 13.
Formula 13:yj’=yj
Wherein, y1、y2、y3、y4、y5、y6、y7、y8And y9Respectively represent Pa1, Pa2, Pa3, Pa4, P5, Pa5, Pa6, Pa7 and
Pa8 ordinate, and y1’、y2’、y3’、y4’、y5’、y6’、y7’、y8' and y9' represent respectively Pa1 ', Pa2 ', Pa3 ', Pa4 ',
P5 ', Pa5 ', Pa6 ', Pa7 ' and Pa8 ' ordinate.
So, as shown in Figure 6A, document boundaries S can be obtained153Each shape point Pa1 ' on new border after deformation,
Pa2 ', Pa3 ', Pa4 ', P5 ', Pa5 ', Pa6 ', Pa7 ' and Pa8 ' coordinate.Similarly, document boundaries S can be obtained264Deformation
The coordinate of each shape point on new border afterwards(Not shown in figure), so as to obtain document boundaries S153And S264After deformation
Extend sample.
Similarly, Fig. 6 B show the example of the contraction carried out to predetermined profile in horizontal direction, and Fig. 6 C and Fig. 6 D then divide
Do not show to predetermined profile carry out vertical direction on expansion and contraction example, its principle with above in conjunction with Fig. 6 A institutes
The example of description is similar, no longer repeats one by one here.It should be noted that in actual applications, above-mentioned Fig. 6 A can be selected
One or more shown in~6D are deformed to deform predetermined profile, that is to say, that multiple predetermined profiles are used
Deformation is not necessarily identical, and can also carry out repeatedly different types of deformation respectively to same predetermined profile to obtain multiple expansions
Exhibit-sample sheet.In addition, in actual applications, can also be using the deformation in addition to above-mentioned Fig. 6 A~example shown in 6D to predetermined profile
It is extended, for example, can combine any of any of Fig. 6 A and Fig. 6 B and Fig. 6 C and Fig. 6 D comes pre- to some
Fixed wheel exterior feature is deformed(I.e. while deforming in the horizontal and vertical directions, but the degree of deformation is different), etc..
In addition, embodiments of the invention additionally provide a kind of image processing method, the image processing method includes:Extraction is treated
Handle the shape facility and/or fluctuation characteristic of profile;And refusal will in the case of at least one in meeting following condition
Pending profile as destination object contour detecting result:Between the shape facility and predetermined shape model of pending profile
Similarity is less than similar between the first similarity threshold and the fluctuation characteristic and predetermined fluctuation characteristic model of pending profile
Degree is less than the second similarity threshold;Wherein, predetermined shape model and/or predetermined fluctuation characteristic model are included multiple pre- by study
The wide training sample of fixed wheel and obtain, the regularity of distribution of the shape facility of predetermined shape model reflection training sample, predetermined fluctuation
The regularity of distribution of the fluctuation characteristic of characteristic model reflection training sample, predetermined profile is identical with the type of pending profile.
A kind of exemplary process of above-mentioned image processing method is described with reference to Fig. 7.
As shown in fig. 7, the handling process 700 of image processing method starts from step according to an embodiment of the invention
S710, then performs step S720.
In step S720, the shape facility and/or fluctuation characteristic of pending profile are extracted.Then step S730 is performed.
Wherein, processing performed in step S720 for example can with above in conjunction with the feature extraction unit described by Fig. 1 or Fig. 2
110 processing is identical, and can reach similar technique effect, will not be repeated here.
In step S730, refuse to regard pending profile as mesh in the case of at least one in meeting following condition
Mark the contour detecting result of object:Similarity between the shape facility of pending profile and predetermined shape model is less than the first phase
Like degree threshold value(That is described condition one above), and between the fluctuation characteristic and predetermined fluctuation characteristic model of pending profile
Similarity is less than the second similarity threshold(That is described condition two above).Wherein, predetermined shape model and/or predetermined fluctuation are special
Levying model includes the training sample of multiple predetermined profiles by study and obtains, and predetermined shape model reflects the shape of training sample
The regularity of distribution of feature, the regularity of distribution of the fluctuation characteristic of predetermined fluctuation characteristic model reflection training sample, predetermined profile is with treating
The type for handling profile is identical.Then step S740 is performed.Wherein, processing performed in step S730 for example can with above
In rejection processing unit 120 described in conjunction with Figure 1 processing it is identical, and similar technique effect can be reached, herein no longer
Repeat.
In one example, predetermined shape model can be by with obtaining above in conjunction with the shape described by Fig. 4
The processing identical of subelement 432 handles to obtain, and can realize similar function and effect, repeats no more here.
In addition, in another example, predetermined fluctuation characteristic model can by with above in conjunction with the ripple described by Fig. 4
The processing identical that dynamic characteristic model obtains subelement 434 handles to obtain, and can realize similar function and effect, this
In repeat no more.
Handling process 700 ends at step S740.
By above description, above-mentioned image processing method according to an embodiment of the invention, it waits to locate by extracting
Manage the shape facility and/or fluctuation characteristic of profile by above-mentioned shape facility and/or fluctuation characteristic respectively with corresponding preboarding
Between shape model and/or predetermined fluctuation characteristic model carry out similarity-rough set, and at least one of which similarity less than correspondence
Predetermined threshold in the case of the rejection pending profile, i.e. refuse using pending profile as destination object contour detecting
As a result.In actual applications, when using the image processing method to the profile that is extracted using existing outline extraction technique(Make
For pending profile)When being handled, it can judge to receive or the rejection profile by above-mentioned similarity-rough set.So, lead to
Above-mentioned processing is crossed, some precision present in final contour detecting result can be caused and/or the degree of accuracy is relatively low, and/or matter
Amount is poor(Such as deform serious etc.)Profile be removed, it is possible to increase contours extract precision and/or the degree of accuracy.
In addition, in some actual treatments carried out using above-mentioned image processing method, user can also be according to rejection
Result choose whether that the profile fallen to these by rejection carries out follow-up processing, this contributes to user multiple to wait to locate existing
The problem of being recognized wherein in the case of reason profile profile(That is, precision and/or the degree of accuracy are relatively low, and/or second-rate(Such as become
Shape is serious etc.)Profile), to improve precision, the degree of accuracy and/or the quality of these profiles by correspondingly subsequent treatment, etc.
Deng.
It should be noted that in the specific processing of above-mentioned image processing method according to embodiments of the present invention, its each
Step, sub-step can be respectively adopted and above in conjunction with pair in the image processing apparatus described by any of Fig. 1-Fig. 5
The processing identical processing of the unit or subelement answered, and similar function and effect can be reached, no longer go to live in the household of one's in-laws on getting married one by one here
State.
In addition, embodiments of the invention additionally provide a kind of electronic equipment, the electronic equipment includes image as described above
Processing unit.In the specific implementation of electronic equipment above-mentioned according to an embodiment of the invention, above-mentioned electronic equipment can be with
It is any one equipment in following equipment:Computer;Digital camera;Scanner;Tablet personal computer;Personal digital assistant;Many matchmakers
Body playback equipment;Mobile phone and electric paper book etc..Wherein, the feelings of image processing apparatus as described above are included in the electronic equipment
Under condition, the electronic equipment can have the various functions and technique effect of above-mentioned image processing apparatus, repeat no more here
Each component units, subelement, module in above-mentioned image processing apparatus according to an embodiment of the invention etc. can
To be configured by way of software, firmware, hardware or its any combination.In the case where being realized by software or firmware,
Can be from storage medium or network to the machine with specialized hardware structure(General-purpose machinery 800 for example shown in Fig. 8)Install and constitute
The program of the software or firmware, the machine when being provided with various programs, be able to carry out above-mentioned each component units, subelement it is each
Plant function.
Fig. 8 be show can be used to realize image processing apparatus according to an embodiment of the invention and image processing method,
Or can be used to realize a kind of possible information of image processing apparatus according to an embodiment of the invention and image processing method
The structure diagram of the hardware configuration of processing equipment.
In fig. 8, CPU (CPU) 801 is according to the program stored in read-only storage (ROM) 802 or from depositing
The program that storage part 808 is loaded into random access memory (RAM) 803 performs various processing.In RAM803, always according to needs
Store the data required when CPU801 performs various processing etc..CPU801, ROM802 and RAM803 via bus 804 each other
Connection.Input/output interface 805 is also connected to bus 804.
Components described below is also connected to input/output interface 805:Importation 806(Including keyboard, mouse etc.), output
Part 807(Including display, such as cathode-ray tube (CRT), liquid crystal display (LCD), and loudspeaker etc.), storage part
808(Including hard disk etc.), communications portion 809(Including NIC such as LAN card, modem).Communications portion 809
Communication process is performed via network such as internet.As needed, driver 810 can be connected to input/output interface 805.
Detachable media 811 such as disk, CD, magneto-optic disk, semiconductor memory etc. can be installed in driver as needed
On 810 so that the computer program read out can be installed in storage part 808 as needed.
In the case where realizing above-mentioned series of processes by software, can from network such as internet or from storage medium example
As detachable media 811 installs the program of composition software.
It will be understood by those of skill in the art that this storage medium be not limited to wherein having program stored therein shown in Fig. 8,
Separately distribute to provide a user the detachable media 811 of program with equipment.The example of detachable media 811 includes disk
(including floppy disk), CD (comprising compact disc read-only memory (CD-ROM) and digital universal disc (DVD)), magneto-optic disk(Comprising mini
Disk (MD) (registration mark)) and semiconductor memory.Or, storage medium can ROM802, storage part 808 in include
Hard disk etc., wherein computer program stored, and it is distributed to together with the equipment comprising them user.
In addition, the invention also provides a kind of program product of the instruction code for the machine-readable that is stored with.Above-mentioned instruction
When code is read and performed by machine, above-mentioned image processing method according to an embodiment of the invention can perform.Correspondingly, it is used for
The various storage mediums such as disk, CD, magneto-optic disk, semiconductor memory for carrying this program product are also included within this
In the disclosure of invention.
In description above to the specific embodiment of the invention, the feature for describing and/or showing for a kind of embodiment
It can be used in same or similar mode in one or more other embodiments, with the feature in other embodiment
It is combined, or substitute the feature in other embodiment.
In addition, the method for various embodiments of the present invention be not limited to specifications described in or shown in accompanying drawing when
Between sequentially perform, can also be according to other time sequencings, concurrently or independently perform.Therefore, described in this specification
Method execution sequence not to the present invention technical scope be construed as limiting.
It should be further understood that can also can be stored in various machines according to each operating process of the above method of the present invention
The mode of computer executable program in the storage medium of reading is realized.
Moreover, the purpose of the present invention can also be accomplished in the following manner:By the above-mentioned executable program code that is stored with
Storage medium is directly or indirectly supplied to computer or center processing in system or equipment, and the system or equipment
Unit(CPU)Read and perform said procedure code.
Now, as long as the system or equipment have the function of configuration processor, then embodiments of the present invention are not limited to
Program, and the program can also be arbitrary form, for example, program that target program, interpreter are performed or being supplied to behaviour
Make shell script of system etc..
These above-mentioned machinable mediums include but is not limited to:Various memories and memory cell, semiconductor equipment,
Disk cell such as light, magnetic and magneto-optic disk, and it is other suitable for medium of storage information etc..
In addition, client computer is by the corresponding website that is connected on internet, and by the computer according to the present invention
Program code is downloaded and is installed in computer and then performs the program, can also realize the present invention.
Finally, in addition it is also necessary to explanation, herein, such as left and right, first and second or the like relational terms be only
Only it is used for making a distinction an entity or operation with another entity or operation, and not necessarily requires or imply these realities
There is any this actual relation or order between body or operation.Moreover, term " comprising ", "comprising" or its it is any its
His variant is intended to including for nonexcludability, so that process, method, article or equipment including a series of key elements are not
Only include those key elements, but also other key elements including being not expressly set out, or also include be this process, method,
Article or the intrinsic key element of equipment.In the absence of more restrictions, by wanting that sentence "including a ..." is limited
Element, it is not excluded that also there is other identical element in the process including the key element, method, article or equipment.
To sum up, in an embodiment according to the present invention, the invention provides following scheme but not limited to this:
A kind of 1. image processing apparatus are attached, including:
Feature extraction unit, it is arranged to the shape facility and/or fluctuation characteristic that extract pending profile;And
Rejection processing unit, it is arranged to the refusal in the case of at least one in meeting following condition will be described
Pending profile as destination object contour detecting result:
Similarity between the shape facility and predetermined shape model of the pending profile is less than the first similarity threshold,
And
Similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of the pending profile is less than the second similarity
Threshold value;
Wherein, the predetermined shape model and/or the predetermined fluctuation characteristic model include multiple predetermined wheels by study
Wide training sample and obtain, the predetermined shape model reflects the regularity of distribution of the shape facility of the training sample, described
Predetermined fluctuation characteristic model reflects the regularity of distribution of the fluctuation characteristic of the training sample, the predetermined profile with it is described pending
The type of profile is identical.
Image processing apparatus of the note 2. according to note 1, wherein, the feature extraction unit is obtained including fluctuation characteristic
Subelement is obtained, the fluctuation characteristic obtains subelement and is arranged to:
Multiple first points pairs are determined based on multiple shape points on the pending profile;
Obtain the multiple first point line segment distance to respective point-to-point transmission;And
The multiple first point is fluctuated to corresponding all line segment distances according to the first predefined procedure formation first
Characteristic sequence, is used as the fluctuation characteristic of the pending profile.
Image processing apparatus of the note 3. according to note 2, wherein, the fluctuation characteristic obtains subelement and is configured to use
Multiple shape points on to the pending profile are matched two-by-two according to predetermined matching method, and will be used as many of pairing result
Individual point is to being defined as the multiple first point pair.
Image processing apparatus of the note 4. according to any one of note 1-3, in addition to:
Pretreatment unit, it is configured to carry out predefined type at least part in the multiple predetermined profile
Deformation to obtain extension sample, and regard the extension sample obtained as the part in the training sample.
Image processing apparatus of the note 5. according to note 4, wherein, the pretreatment unit is arranged to:
For each at least part in the multiple predetermined profile, at least one change of the predetermined profile is obtained
Shape, and using at least one deformation described in acquisition as the extension sample, wherein, each deformation of the predetermined profile passes through such as
Lower processing is obtained:
By making at least one-dimensional coordinate in the N-dimensional coordinate of each shape point in the predetermined profile increase within a predetermined range
Or reduce to form renewal N-dimensional coordinate so that the shape that the renewal N-dimensional coordinate of each shape point is constituted in the predetermined profile
It can be used in description and the profile of the pending profile same type;Wherein, N is the integer more than or equal to 2.
Image processing apparatus of the note 6. according to note 5, wherein, the pretreatment unit is arranged to described
During each at least part in multiple predetermined profiles is deformed, each shape point in the predetermined profile is kept
N-dimensional coordinate at least one-dimensional coordinate it is constant, make in the N-dimensional coordinate except it is described it is at least one-dimensional in addition to remaining dimension coordinate exist
Increase or reduce to form the renewal N-dimensional coordinate of each shape point in the predetermined profile in preset range.
Image processing apparatus of the note 7. according to any one of note 1-6, wherein, the feature extraction unit bag
Include:
Shape facility obtains subelement, and it is arranged to the seat using multiple shape points on the pending profile
Mark, according to the second predefined procedure formation first shape characteristic sequence, be used as the shape facility of the pending profile.
Image processing apparatus of the note 8. according to any one of note 1-7, in addition to model obtaining unit, wherein,
The model obtaining unit includes fluctuation characteristic model and obtains subelement, and the fluctuation characteristic model obtains subelement and is configured to use
In:
For each in the training sample after alignment, based in multiple shape points on the training sample come really
Fixed multiple second points pair, obtain the line segment distance of the point-to-point transmission of each second point centering, and utilize the multiple second point
To corresponding multiple line segment distances, according to the first predefined procedure formation the second fluctuation characteristic sequence corresponding with the training sample;
And
Obtain the average fluctuation characteristic sequence and ripple of the corresponding second fluctuation characteristic sequence of all training samples
Dynamic Eigen Covariance matrix, to describe the predetermined fluctuation characteristic model.
Image processing apparatus of the note 9. according to any one of note 1-8, in addition to model obtaining unit, wherein,
The model obtaining unit includes shape and obtains subelement, and the shape obtains subelement and is arranged to:
For each in the training sample after alignment, the seat of multiple shape points on the training sample is utilized
Mark, according to the second predefined procedure formation the second shape facility sequence corresponding with the training sample;And
Obtain the average shape characteristic sequence and shape of the corresponding second shape facility sequence of all training samples
Shape Eigen Covariance matrix, to describe the predetermined shape model.
Image processing apparatus of the note 10. according to any one of note 1-9, wherein, the rejection processing unit quilt
It is configured to:
Distance between the shape facility and the predetermined shape model of the pending profile is higher than first apart from threshold
During value, judge that the similarity between the shape facility and predetermined shape model of the pending profile is less than the first similarity threshold
Value;And/or
Distance between the fluctuation characteristic and the predetermined fluctuation characteristic model of the pending profile higher than second away from
During from threshold value, judge that the similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of the pending profile is less than the second phase
Like degree threshold value.
Image processing apparatus of the note 11. according to any one of note 1-10, wherein, the destination object is text
Shelves, and the profile are document boundaries.
A kind of 12. image processing methods are attached, including:
Extract the shape facility and/or fluctuation characteristic of pending profile;And
Refuse to regard the pending profile as destination object in the case of at least one in meeting following condition
Contour detecting result:
Similarity between the shape facility and predetermined shape model of the pending profile is less than the first similarity threshold,
And
Similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of the pending profile is less than the second similarity
Threshold value;
Wherein, the predetermined shape model and/or the predetermined fluctuation characteristic model include multiple predetermined wheels by study
Wide training sample and obtain, the predetermined shape model reflects the regularity of distribution of the shape facility of the training sample, described
Predetermined fluctuation characteristic model reflects the regularity of distribution of the fluctuation characteristic of the training sample, the predetermined profile with it is described pending
The type of profile is identical.
Image processing method of the note 13. according to note 12, wherein, the fluctuation characteristic of the pending profile passes through
Following manner is obtained:
Multiple first points pairs are determined based on multiple shape points on the pending profile;
Obtain the multiple first point line segment distance to respective point-to-point transmission;And
First wave is formed according to the first predefined procedure to corresponding all line segment distances using the multiple first point
Dynamic characteristic sequence, is used as the fluctuation characteristic of the pending profile.
Image processing method of the note 14. according to note 13, wherein it is determined that multiple first point pair of step includes:
Multiple shape points on the pending profile are matched two-by-two according to predetermined matching method, and the multiple of pairing result will be used as
Point is to being defined as the multiple first point pair.
Image processing method of the note 15. according to any one of note 12-14, wherein, the training sample is also wrapped
Include:
Pass through the extension sample that the deformation of predefined type is carried out at least part in the multiple predetermined profile and is obtained.
Image processing method of the note 16. according to any one of note 12-15, wherein, the pending profile
Shape facility is obtained in the following way:
Using the coordinate of multiple shape points on the pending profile, according to the second predefined procedure formation first shape spy
Sequence is levied, the shape facility of the pending profile is used as.
It is attached 17. a kind of electronic equipment, including the image processing apparatus as described in any in note 1-11.
Electronic equipments of the note 18. according to note 17, wherein, the electronic equipment is any one in following equipment
Kind:Computer;Digital camera, scanner, tablet personal computer, personal digital assistant, multimedia play equipment, mobile phone and electric paper
Book.
A kind of 19. program products of the instruction code for the machine-readable that is stored with are attached, described program product is upon execution
The machine can be made to perform according to any described image processing method in note 12-16.
A kind of 20. computer-readable recording mediums are attached, the program product according to note 19 is stored thereon with.
Claims (9)
1. a kind of image processing apparatus, including:
Feature extraction unit, it is arranged to the shape facility and/or ripple that extract the pending profile of destination object in image
Dynamic feature;And
Rejection processing unit, it is arranged to refuse to wait to locate by described in the case of at least one in meeting following condition
Profile is managed as the contour detecting result of the destination object:
Similarity between the shape facility and predetermined shape model of the pending profile is less than the first similarity threshold, and
Similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of the pending profile is less than the second similarity threshold;
Wherein, the predetermined shape model and/or the predetermined fluctuation characteristic model include multiple predetermined profiles by study
Training sample and obtain, the predetermined shape model reflects the regularity of distribution of the shape facility of the training sample, described predetermined
Fluctuation characteristic model reflects the regularity of distribution of the fluctuation characteristic of the training sample, the predetermined profile and the pending profile
Type it is identical,
Wherein, the feature extraction unit includes fluctuation characteristic acquisition subelement, and the fluctuation characteristic obtains subelement and is configured
For:
Multiple first points pairs are determined based on multiple shape points on the pending profile;
Obtain the multiple first point line segment distance to respective point-to-point transmission;And
The first fluctuation characteristic is formed according to the first predefined procedure to corresponding all line segment distances by the multiple first point
Sequence, is used as the fluctuation characteristic of the pending profile.
2. image processing apparatus according to claim 1, wherein, the fluctuation characteristic obtains subelement and is arranged to pair
Multiple shape points on the pending profile are matched two-by-two according to predetermined matching method, and will be used as multiple points of pairing result
To being defined as the multiple first point pair.
3. image processing apparatus according to claim 1 or 2, in addition to:
Pretreatment unit, it is configured to the change that predefined type is carried out at least part in the multiple predetermined profile
Shape regard the extension sample obtained as the part in the training sample to obtain extension sample.
4. image processing apparatus according to claim 1 or 2, wherein, the feature extraction unit includes:
Shape facility obtains subelement, its be arranged to coordinate using multiple shape points on the pending profile, by
According to the second predefined procedure formation first shape characteristic sequence, the shape facility of the pending profile is used as.
5. image processing apparatus according to claim 1 or 2, in addition to model obtaining unit, wherein, the model is obtained
Unit includes fluctuation characteristic model and obtains subelement, and the fluctuation characteristic model obtains subelement and is arranged to:
For each in the training sample after alignment, based on many to determine in multiple shape points on the training sample
Individual second point pair, obtains the line segment distance of the point-to-point transmission of each second point centering, and using the multiple second point to right
Multiple line segments distance for answering, according to corresponding with the training sample the second fluctuation characteristic sequence of the first predefined procedure formation;And
Average fluctuation characteristic sequence and the fluctuation for obtaining the corresponding second fluctuation characteristic sequence of all training samples are special
Covariance matrix is levied, to describe the predetermined fluctuation characteristic model.
6. image processing apparatus according to claim 1 or 2, in addition to model obtaining unit, wherein, the model is obtained
Unit includes shape and obtains subelement, and the shape obtains subelement and is arranged to:
For each in the training sample after alignment, using the coordinate of multiple shape points on the training sample, press
According to the second predefined procedure formation the second shape facility sequence corresponding with the training sample;And
The average shape characteristic sequence and shape for obtaining the corresponding second shape facility sequence of all training samples are special
Covariance matrix is levied, to describe the predetermined shape model.
7. image processing apparatus according to claim 1 or 2, wherein, the destination object is document, and the profile
For document boundaries.
8. a kind of image processing method, including:
Extract the shape facility and/or fluctuation characteristic of the pending profile of destination object in image;And
Refuse in the case of at least one in meeting following condition using the pending profile as destination object profile
Testing result:
Similarity between the shape facility and predetermined shape model of the pending profile is less than the first similarity threshold, and
Similarity between the fluctuation characteristic and predetermined fluctuation characteristic model of the pending profile is less than the second similarity threshold;
Wherein, the predetermined shape model and/or the predetermined fluctuation characteristic model include multiple predetermined profiles by study
Training sample and obtain, the predetermined shape model reflects the regularity of distribution of the shape facility of the training sample, described predetermined
Fluctuation characteristic model reflects the regularity of distribution of the fluctuation characteristic of the training sample, the predetermined profile and the pending profile
Type it is identical,
Wherein, the shape facility and/or fluctuation characteristic for extracting the pending profile of destination object in image further comprise:
Multiple first points pairs are determined based on multiple shape points on the pending profile;
Obtain the multiple first point line segment distance to respective point-to-point transmission;And
The first fluctuation characteristic is formed according to the first predefined procedure to corresponding all line segment distances by the multiple first point
Sequence, is used as the fluctuation characteristic of the pending profile.
9. a kind of electronic equipment, including the image processing apparatus as described in any in claim 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310187902.6A CN104182749B (en) | 2013-05-20 | 2013-05-20 | Image processing apparatus, image processing method and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310187902.6A CN104182749B (en) | 2013-05-20 | 2013-05-20 | Image processing apparatus, image processing method and electronic equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104182749A CN104182749A (en) | 2014-12-03 |
CN104182749B true CN104182749B (en) | 2017-08-08 |
Family
ID=51963776
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310187902.6A Expired - Fee Related CN104182749B (en) | 2013-05-20 | 2013-05-20 | Image processing apparatus, image processing method and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104182749B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10152213B2 (en) * | 2016-09-01 | 2018-12-11 | Adobe Systems Incorporated | Techniques for selecting objects in images |
WO2020194472A1 (en) * | 2019-03-25 | 2020-10-01 | オリンパス株式会社 | Movement assist system, movement assist method, and movement assist program |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102436650A (en) * | 2010-09-29 | 2012-05-02 | 奥林巴斯株式会社 | image processing apparatus and image processing method |
CN102469230A (en) * | 2010-11-11 | 2012-05-23 | 夏普株式会社 | Image processing device and method, image forming device, and image reading device |
CN102509078A (en) * | 2011-10-28 | 2012-06-20 | 北京安控科技股份有限公司 | Fire detection device based on video analysis |
CN102697446A (en) * | 2011-01-11 | 2012-10-03 | 奥林巴斯株式会社 | Image processing device and image processing method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4375448B2 (en) * | 2007-06-26 | 2009-12-02 | ソニー株式会社 | Image processing apparatus, imaging apparatus, image processing method, and program |
-
2013
- 2013-05-20 CN CN201310187902.6A patent/CN104182749B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102436650A (en) * | 2010-09-29 | 2012-05-02 | 奥林巴斯株式会社 | image processing apparatus and image processing method |
CN102469230A (en) * | 2010-11-11 | 2012-05-23 | 夏普株式会社 | Image processing device and method, image forming device, and image reading device |
CN102697446A (en) * | 2011-01-11 | 2012-10-03 | 奥林巴斯株式会社 | Image processing device and image processing method |
CN102509078A (en) * | 2011-10-28 | 2012-06-20 | 北京安控科技股份有限公司 | Fire detection device based on video analysis |
Non-Patent Citations (2)
Title |
---|
"图像检索中的特征表示模型和多信息源融合方式的研究";陆文婷;《中国博士学位论文全文数据库信息科技辑》;20130115;论文正文第17页第9段-第18页第3段、第38页第3段-第39页第8段 * |
"基于视频序列多特征的火焰检测算法的研究";刘昊;《中国优秀硕士学位论文全文数据库 信息科技辑》;20121215;论文正文第23页第1段-第29页第4段、第33页第1段-第41页第3段,附图3.1 * |
Also Published As
Publication number | Publication date |
---|---|
CN104182749A (en) | 2014-12-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106485192B (en) | Training method and device of neural network for image recognition | |
US11069047B2 (en) | Image processing method, image processing apparatus, computing device and computer-readable storage medium | |
US20230081645A1 (en) | Detecting forged facial images using frequency domain information and local correlation | |
CN104239882B (en) | Image similarity determining device and method and image feature obtaining device and method | |
CN110431564A (en) | Method and system for Fingerprint enhancement | |
JP5410355B2 (en) | Matrix generation apparatus, method, program, and information processing apparatus | |
US20150310305A1 (en) | Learning painting styles for painterly rendering | |
CN104285239A (en) | Image processing device, image processing method, program, print medium, and recording medium | |
CN109726195B (en) | Data enhancement method and device | |
CN112584062B (en) | Background audio construction method and device | |
CN112116720A (en) | Three-dimensional point cloud augmentation method and device, storage medium and computer equipment | |
JP5500400B1 (en) | Image processing apparatus, image processing method, and image processing program | |
Jung et al. | Iterative adaptive RBF methods for detection of edges in two-dimensional functions | |
CN114529728A (en) | Small sample semantic segmentation method and device | |
CN114444566A (en) | Image counterfeiting detection method and device and computer storage medium | |
CN108615006A (en) | Method and apparatus for output information | |
CN104182749B (en) | Image processing apparatus, image processing method and electronic equipment | |
CN111027577A (en) | Fabric abnormal texture type identification method and device | |
CN108241861A (en) | A kind of data visualization method and equipment | |
Lu et al. | An image enhancement method for side-scan sonar images based on multi-stage repairing image fusion | |
JP2014099027A (en) | Image feature amount calculation device, learning device, identification device, and program | |
Ling et al. | A high-performance face illumination processing method via multi-stage feature maps | |
Yue et al. | Sift-based image super-resolution | |
Amarù et al. | DALib: A Curated Repository of Libraries for Data Augmentation in Computer Vision | |
CN111882563B (en) | Semantic segmentation method based on directional full convolution network |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170808 Termination date: 20180520 |
|
CF01 | Termination of patent right due to non-payment of annual fee |