CN108629264A - Method and apparatus for image procossing - Google Patents
Method and apparatus for image procossing Download PDFInfo
- Publication number
- CN108629264A CN108629264A CN201710760469.9A CN201710760469A CN108629264A CN 108629264 A CN108629264 A CN 108629264A CN 201710760469 A CN201710760469 A CN 201710760469A CN 108629264 A CN108629264 A CN 108629264A
- Authority
- CN
- China
- Prior art keywords
- image
- scale
- matrix
- under
- rectangle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 76
- 239000011159 matrix material Substances 0.000 claims abstract description 108
- 239000000203 mixture Substances 0.000 claims abstract description 60
- 239000013598 vector Substances 0.000 claims abstract description 54
- 238000001914 filtration Methods 0.000 claims abstract description 52
- 238000000513 principal component analysis Methods 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 20
- 230000011218 segmentation Effects 0.000 claims abstract description 15
- 238000003860 storage Methods 0.000 claims abstract description 8
- 238000004590 computer program Methods 0.000 claims description 20
- 238000009826 distribution Methods 0.000 claims description 15
- 230000009466 transformation Effects 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 6
- 230000002093 peripheral effect Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 229910002056 binary alloy Inorganic materials 0.000 claims 1
- 238000012545 processing Methods 0.000 description 14
- 238000005516 engineering process Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 7
- 239000000758 substrate Substances 0.000 description 7
- 238000003709 image segmentation Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000005520 cutting process Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 235000007926 Craterellus fallax Nutrition 0.000 description 2
- 240000007175 Datura inoxia Species 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000013604 expression vector Substances 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000011049 filling Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000007430 reference method Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/18—Eye characteristics, e.g. of the iris
- G06V40/193—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Ophthalmology & Optometry (AREA)
- Human Computer Interaction (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
This disclosure relates to the method and apparatus for image procossing.One embodiment discloses a kind of method for image procossing, including：Obtain pending image；Segmentation described image obtains multiple images block；Each vector is converted to by described multiple images are in the block, and the vector is merged into matrix；Principal component analysis is carried out to the obtained matrix, obtains main composition matrix；The pending image is filtered from the main composition matrix extraction filtering core.The disclosure also describes corresponding device and computer system and computer readable storage medium storing program for executing.
Description
Technical field
Present invention relates in general to field of information processing, and are more particularly to used for the method and apparatus of image procossing.
Background technology
With the development of technology, biological identification technology starts to be increasingly becoming hot spot.Biological identification technology has convenient and essence
Accurate feature, is more and more applied in security fields.
Due to the unique structure feature that iris itself has, in biological identification technology, iris recognition technology it is accurate
The universal approval of industry is conveniently obtained.During iris recognition, a very important link is from iris image
Middle extraction feature.It is only extracted characteristic information accurate and abundant enough, could preferably support subsequent verification process.
Under normal circumstances, the iris of the mankind is distributed on the annulus of a diameter of 12mm or so.It will be from so small region
Extraction feature accurate and abundant enough is an extremely complex and difficult task to identify different targets.
Currently, industry generally use Gabor characteristic expresses the texture of iris image.Gabor characteristic in order to obtain needs
It is filtered using Gabor wavelet verification iris image.However, the design of Gabor wavelet core is independently of iris image, namely
The determination of Gabor wavelet core is not rely on iris image itself, therefore is difficult fully to extract distinctive, tool in iris image
There are expression and identification feature.
Invention content
Generally, the embodiment of the present invention proposes a kind of technical solution for image procossing.
In one aspect, the embodiment of the present invention provides a kind of method for image procossing.The method includes：It obtains
Pending image；Segmentation described image obtains multiple images block；Each vector is converted to by described multiple images are in the block, and
The vector is merged into matrix；Principal component analysis is carried out to the obtained matrix, obtains main composition matrix；And from described
Main composition matrix extraction filtering core is filtered the pending image.
On the other hand, the embodiment of the present invention provides a kind of device for image procossing.Described device includes：It obtains
Module is configured as obtaining pending image；Divide module, is configured as segmentation described image and obtains multiple images block；Turn
Block is changed the mold, is configured as each being converted to vector by described multiple images are in the block, and the vector is merged into matrix；Analysis
Module is configured as carrying out principal component analysis to the obtained matrix, obtains main composition matrix；And filter module, by with
It is set to from the main composition matrix extraction filtering core and the pending image is filtered.
It yet still another aspect, the embodiment of the present invention provides a kind of computer system for image procossing, which includes：
One or more processors；One or more computer-readable mediums；What be may be stored on the computer-readable medium is used for by one
Or the computer program instructions that at least one of multiple processors execute, the computer program instructions include：For obtaining
The computer program instructions of pending image；The computer program instructions of multiple images block are obtained for dividing described image；
For being each converted to vector by described multiple images are in the block, and the computer program that the vector is merged into matrix refers to
It enables；For carrying out principal component analysis to the obtained matrix, the computer program instructions of main composition matrix are obtained；And it is used for
The computer program instructions that the pending image is filtered from the main composition matrix extraction filtering core.
In another aspect, a kind of computer-readable storage medium of the embodiment of the present invention offer, described computer-readable
At least one executable computer program instructions are stored on storage medium, the computer program instructions include for executing
The computer program instructions of each step in embodiment of the method.
According to an embodiment of the invention, the filtering core for being more suitable for image can be obtained, to improve extracted feature
Quality.
Description of the drawings
Exemplary embodiment of the invention is described in more detail in conjunction with the accompanying drawings, it is of the invention above-mentioned and its
Its purpose, feature and advantage will be apparent, wherein：
Fig. 1 shows the environment suitable for being used for realizing the embodiment of the present invention；
Fig. 2 shows another environment for being suitable for being used for realizing the embodiment of the present invention；
Fig. 3 shows a kind of schematic flow chart of method for image procossing according to the ... of the embodiment of the present invention；
Fig. 4 shows a kind of schematic flow of method that annular image is expanded into rectangle according to the ... of the embodiment of the present invention
Figure；
Fig. 5 shows that annular image is expanded into the schematic stream of the method for rectangle by another kind according to the ... of the embodiment of the present invention
Cheng Tu；
Fig. 6 shows a kind of schematic flow chart of method for image procossing according to the ... of the embodiment of the present invention；
Fig. 7 shows a kind of schematic flow chart for obtaining the method for histogram according to the ... of the embodiment of the present invention；
Fig. 8 shows a kind of schematic flow chart of method for image procossing according to the ... of the embodiment of the present invention；And
Fig. 9 shows a kind of schematical structure diagram of device for image procossing according to the ... of the embodiment of the present invention.
In the accompanying drawings, same or analogous label is used to represent same or analogous element.
Specific implementation mode
The preferred embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Preferred embodiment, however, it is to be appreciated that the present invention can also with other various forms realize without should be limited in below retouch
The specific embodiment stated.These specific embodiments are provided herein it are to keep the disclosure more thorough and complete, and
And the scope of the present disclosure can be completely communicated to those skilled in the art.
Fig. 1 shows the block diagram of the exemplary environments suitable for being used for realizing embodiment of the present invention.The environment can be
One terminal 100 with simple computation ability, can also be the node 100 with complicated calculations ability.
The environment is for example including computer-readable medium 101.These media for example can be that volatile and non-volatile is situated between
Matter can also be moveable and immovable medium, as long as can be with the node visit of computing capability.
The environment for example can also include one or more program modules 103, these program modules are commonly used in executing sheet
Invent the function and/or method in the embodiment of the description.
The environment for example can also include one or more modules 105 with computing capability.
The environment can independently execute method described in embodiment of the present invention and/or function, can also with it is outer
The communication of portion's equipment 107 completes corresponding method and/or function to cooperate.Certainly, it will be understood by those skilled in the art that the terminal
100 or calculate node 100 can be for example server or computer, can also be intelligent terminal, such as electronic lock, intelligent hand
Machine, Intelligent flat etc., the present invention is not limited thereto.
Fig. 2 shows the block diagrams for the exemplary environments for being suitable for being used for realizing embodiment of the present invention.The environment includes eventually
End 201 and calculate node 203.The environment for example can be a cloud environment, and calculate node 203 is, for example, Cloud Server at this time.
The environment for example can also be other communication systems, and calculate node 203 is, for example, the mobile terminal with computing capability, example at this time
Such as smart mobile phone, Intelligent flat, personal computer.
It will be detailed below the mechanism and principle of the embodiment of the present invention.Unless specifically stated otherwise, below and claim
The middle term "based" used indicates " being based at least partially on ".Term " comprising " indicates that opening includes, i.e., " including but it is unlimited
In ".Term " multiple " expression " two or more ".Term " one embodiment " expression " at least one embodiment ".Term is " another
Embodiment " expression " at least one other embodiment ".The definition of other terms provides in will be described below.
Fig. 3 shows the schematic of the method 300 for image procossing of property embodiment according to an example of the present invention
Flow chart.Each step for including with reference to Fig. 3 detailed description methods 300.
Method 300 starts from step 301, obtains pending image.It will be understood by those skilled in the art that the acquisition can be with
It is to be obtained by harvester (being, for example, camera), can also be to be obtained from other processing modules inside processing equipment, this
Invention does not limit.Also, the acquisition can be obtained from other equipment by communications conduit, can be from this equipment
Internal module obtains, and the present invention is not limited thereto.Acquisition in various embodiments of the present invention for example may include a series of place
Reason operation, to obtain pending image.It should be understood that the pending image is not necessarily original image, but it is directed to
It is pending for method 300.
Then, method 300 enters step 303, is multiple images block by the image segmentation.Those skilled in the art can be with
Step 303 is realized using the technology of existing any image segmentation, the image Segmentation Technology that future development can also be used to obtain
Step 303 is realized, as long as multiple images block can be divided the image into.It is not intended to limit and divides in the embodiment of the present invention
The size of the image block arrived, such as 3 × 3 size or 5 × 5 size may be used.The embodiment of the present invention is also not intended to limit
Divide the obtained shape of image block, only for facilitating the purpose of description, is retouched below with being divided into for rectangular image block
It states.In various embodiments of the present invention, rectangle for example can be rectangle, or square.
In an alternative embodiment of the invention, the multiple images block divided for example can be partly overlapping image block.
It is divided into partly overlapping image block and so that there are redundancies between image block, to ensure that fault-tolerant energy certain in subsequent processing
Power simplifies the design of follow-up system.
Return to method 300, in step 305, by multiple image it is in the block it is each be converted to vector, and by vector merging
At matrix.It will be understood by those skilled in the art that each vector is converted to by multiple images are in the block, either column vector
It can be row vector.Below for convenience, it is illustrated by taking column vector as an example.Any switch technology may be used by image
Block is converted to column vector.Hereinafter, being illustrated for being column vector by image block uniaxial direct tensile.It is with 3 × 3 image block
Example, such as can indicate the image block with 3 × 3 matrix, the one of the corresponding position of each element correspondence image block of matrix
A pixel value.So directly each row of this 3 × 3 matrix head and the tail are spliced, just obtain one 9 × 1 column vector.It will obtain
Multiple column vectors simply merge into matrix.For example, obtaining 89 × 1 column vectors, that can use each column vector
As a row of matrix after synthesis, one 9 × 8 matrix is obtained.
In step 307, principal component analysis is carried out to the matrix that step 305 obtains, obtains main composition matrix.Art technology
Personnel are appreciated that the principal component analysis that may be used in method realization step 307 in the prior art.For example, matrix X
For the matrix obtained in step 305.It is matrix using substrate V restructuring matrixes XAnd formula (1) is set up.
s.t.VVT=I
Wherein, I is unit matrix.It will be understood by those skilled in the art that the purpose of bound term is, for example, so that substrate is
Orthonormal basis.Formula (1) for example may be used singular value decomposition algorithm and be solved.In an example of the invention, obtain
Substrate V is for example made of the corresponding feature vector of one group of maximum characteristic value.Using obtained substrate V as main composition matrix.This
Field technology personnel are appreciated that a kind of schematical example that above are only step 307, can also be according to example using more
Principal component analysis method to obtain main composition matrix.
In step 309, obtained in the main composition matrix extraction filtering checking step 301 obtained from step 307 pending
Image be filtered.It can be obtained by the characteristic pattern of the pending image by step 309.
In an alternative embodiment of the invention, for example may include from main composition matrix extraction filtering core：By the main composition square
Battle array is deformed into the size for the image block divided in step 303；And using the deformed main composition matrix as filtering
The pending image is checked to be filtered.In the above-described embodiments, such as in step 303, it is divided into the figure of 3 × 3 sizes
As block, main composition matrix is also become to 3 × 3 size herein.In step 303, if being divided into 5 × 5 size, herein
Main composition matrix is become to 5 × 5 size.At least one filtering core can be obtained in this way.Thus obtained filtering core is clearly
It is relevant with pending image.
By using method 300 as shown in Figure 3, filtering core can be associated with pending image, to profit
Pending image is filtered with specially designed filtering core, can preferably extract the distinctive spy of pending image
Sign.It especially applies in iris recognition field, it, can be effectively by the relevant feature extraction filtering core of adaptive learning iris
Improve the precision of iris feature expression.
In another embodiment of the present invention, step 301 for example may include：It is rectangle by image spread, and the rectangle is made
For the pending image.That is, further processing for convenience, can be first rectangle by obtained image spread.This
Field technology personnel are appreciated that the prior art, which may be used, realizes that by obtained image spread be rectangle.
In another embodiment of the present invention, step 301 for example may include：Annular image is expanded into rectangle, and will be described
Rectangle is as the pending image.In the present embodiment, the image of acquisition, the image Jing Guo preliminary treatment is for example in other words
The annular image can be expanded into rectangular image in step 301 for annular image.It will be understood by those skilled in the art that
The annular image for example can be the positive annulus of standard, can also be oval annulus, can also be other close class circular charts
Picture, the present invention is not limited thereto.In an embodiment of the present invention, which for example can be the iris figure that detection obtains
Picture.It is appreciated that in the various embodiments of the invention, it can be by the circular chart of part that annular image, which is expanded into rectangle for example,
As expanding into rectangle.For example, being only rectangle by upper half or lower half of circular development.For example, it is also possible to which consider will be half left
Or right half expand into rectangle.Certainly it is not only half of annular image, such as can also be the fan ring at 60 ° of angles.
In another embodiment of the present invention, such as can be by the outer edge and inward flange of detection iris to obtain the circular chart
Picture.Those skilled in the art can realize the detection with existing detection technique.For example, round Hough (Hough) change may be used
Change the detection for realizing iris region.
As shown in figure 4, specifically providing a kind of method that annular image is expanded into rectangle in one embodiment of the invention
400.This method 400 includes the following steps.
Step 401, annular image is divided into multiple fan rings.It will be understood by those skilled in the art that by annular image
When being divided into multiple fan rings, both can sector first be respectively divided into the big round and small circular for constituting the annular image, then into
Row merges, and can also directly be divided to obtain fan ring to annular.In specific divide, either decile, or non-
Decile.Below for convenience, using etc. divide the case where be illustrated.Also, it is multiple annular image to be divided into
When fanning ring, only the annular image of part can be divided, such as upper half of annular image or left side or right side
Annular image.It is not required for centainly dividing entire annular image herein.
Step 403, each of the multiple fan ring is radially divided.It will be understood by those skilled in the art that can
It is divided with being realized using various ways.Such as decile can be radially carried out, it can also be divided according to logarithm, or according to
Other modes divide.In an alternative embodiment of the invention, on the one hand in order to more retain the texture information of iris, on the other hand
The complexity of system is also allowed for, such as partitioning technology appropriate (being, for example, that logarithm divides) may be used so that close to inward flange
Part it is intensive, close to outer peripheral part it is sparse.On the one hand it can retain more texture informations in this way, because of the line of iris
It manages in the part more horn of plenty close to inward flange, on the other hand can simplify the complexity of system, because close to outer peripheral
The number that part divides is less.
In step 405, annular image can be expanded into rectangle according to the image of division.
As described in the above-described embodiments, a variety of methods may be used, annular image is expanded into rectangle.Below
Describe annular image to be expanded into another embodiment of the present invention the specific example of the method for rectangle in conjunction with attached drawing 5.Method 500 is wrapped
Include following steps.
Step 501, the size of rectangle in rectangular coordinate system is set.Such as it is expected that the rectangle after expansion is 300 × 600, then
One 300 × 600 rectangle can be set in rectangular coordinate system.Step 503, it is closed according to the transformation of polar coordinates and rectangular co-ordinate
System, finds the point in the annular image that each pair of point is answered in the rectangle.It will be understood by those skilled in the art that annular image is set
It sets in polar coordinate system, and rectangle is arranged in rectangular coordinate system.It, can be by this way, for each point in the rectangle
The point in annular image is found according to the transformation relation of polar coordinates and rectangular co-ordinate.
Step 505, if corresponding point can be found in annular image, using the pixel value of the corresponding point as rectangle
In the point value.
Step 507, if can not find corresponding point in the annular image, the picture of the point on periphery in the annular image is utilized
Plain value is fitted the value as the point in rectangle.Those skilled in the art can utilize any fitting technique to realize step 507.
It is fitted for example, bicubic interpolation method may be used.
According to Fig. 4 and example shown in fig. 5, those skilled in the art can also obtain more expanding into annular image
The method of rectangle.
It, can also be into order to enable filtering core better adapts to the feature of pending image in another embodiment of the present invention
One step carries out deep layer principal component analysis.Here, can be using the characteristic image obtained in step 309 as the pending of step 301
Image executes step 303 to 309 again.It is hereby achieved that more adapting to the filtering core of pending image.Those skilled in the art
It is appreciated that such cycle can execute repeatedly, to obtain the filtering core after multilayer principal component analysis.
In another embodiment of the present invention, method 300 as shown in Figure 3 can also include step after step 300：It is right
The characteristic pattern obtained after filtering carries out statistical presentation.It will be appreciated by those skilled in the art that by carrying out statistical presentation to characteristic pattern
The superposition and other subsequent processing of more size characteristic figures can be facilitated.In an alternative embodiment of the invention, to being obtained after filtering
To characteristic pattern carry out statistical presentation for example may include：The characteristic pattern obtained after filtering is counted, pixel distribution is obtained
Histogram.The present embodiment by the histogram for the pixel distribution for obtaining that there is high-layer semantic information, in bottom visual signature and
High level establishes relationship between inferring.
In one embodiment of the invention, the characteristic pattern obtained after filtering is counted, obtains the histogram legend of pixel distribution
Method 700 as shown in Figure 7 such as may be used to realize.
Method 700 starts from step 701, and binarization operation is carried out to the characteristic pattern obtained after filtering.Those skilled in the art
It is appreciated that other feature coding processing can also be carried out to characteristic pattern, binarization operation is only an example.The present embodiment
In, for example, can be arranged 0 be binarization operation threshold value, it is of course possible to set a threshold to other values, the present invention to this not
It limits.
In step 703, using the value of the corresponding position in the characteristic pattern of each pixel after binarization as binary digit
One, obtain the binary value of each pixel.For example, one shares 10 characteristic patterns, then the center of each characteristic pattern
It is worth one as binary digit, the one 10 corresponding pixels of binary digit namely center can be obtained
Binary value.
In step 705, it converts the binary value of each pixel to decimal value.It will be understood by those skilled in the art that
Binary value can also be converted to the numerical value of other systems, it is one such example to be converted into decimal value.
In step 707, counted to obtain the histogram of pixel distribution according to decimal value.
In one embodiment of the invention, the method that above-described embodiment is stated can also carry out on multiple scales, to
To multiple dimensioned characteristic pattern, the diversity of extracted feature is enriched, the precision of image feature representation is further increased.
Below by taking embodiment shown in fig. 6 as an example, detailed retouch is carried out for the method 600 of multiple dimensioned hypograph processing
It states.
Method 600 starts from step 601, and pending image is divided on multiple scales, obtains multiple under each scale
Image block.For example, dividing pending image on S scale, the multiple images under each scale in the S scale are obtained
Block.In the present embodiment, S is, for example, 3.It it is M partly overlapping 3 by pending image segmentation for example, under first scale
× 3 image block；It is N number of partly overlapping 5 × 5 image block by pending image segmentation under second scale；
Under three scales, the image block for being P 6 × 6 by pending image segmentation.
In step 603, under each scale, by the multiple images under the scale it is in the block it is each be converted to vector, and will
The vector is merged into matrix.In the present embodiment, below for convenience, it is carried out by taking the specific method under the first scale as an example
Illustrate, those skilled in the art can obtain the corresponding processing side under other scales according to the specific example under the first scale
Method.Under the first scale, the image block of M partly overlapping 3 × 3 is respectively converted into 9 × 1 column vector, and by M arrange to
Amount is ranked sequentially the matrix of 9 × M of composition.Similar processing is also carried out under other scales.For example, by N number of partly overlapping 5 ×
5 image block is respectively converted into 25 × 1 column vector, and N number of column vector is ranked sequentially to the matrix of 25 × N of composition.For
The processing of three dimensions is also similar, and details are not described herein again.
In step 605, under each scale, principal component analysis is carried out to obtained matrix, obtain under the scale it is main at
Part matrix.With reference to the specific descriptions in above-described embodiment, similar mode may be used, led accordingly under each scale
Composition matrix.
In step 607, under each scale, filtering core is extracted to the pending figure from the main composition matrix under the scale
As being filtered.
In step 609, the filtered characteristic pattern obtained under each scale in multiple scales is merged.As a result, may be used
To enrich the diversity of extracted feature.
It will be understood by those skilled in the art that parallel can for example be held for the operation in multiple scales on each scale
Row, can also successively execute.Such as cutting operation, the cutting operation on multiple scales can be executed parallel, it can also be first
The cutting operation on multiple scales is executed afterwards.For different processing steps, all of the first scale can have for example been first carried out
Processing step, then execute all processing steps of the second scale, or can also be parallel execution different scale everywhere in manage step
Suddenly.The present invention does not limit the sequence of above-mentioned execution.
In an alternative embodiment of the invention, step 609 for example can be the filter to being obtained under each scale in multiple scales
Characteristic pattern after wave carries out statistical presentation, and the statistical presentation under multiple scale is merged.
In an alternative embodiment of the invention, step 609 for example can be the filter to being obtained under each scale in multiple scales
Characteristic pattern after wave is counted, and obtains the histogram of pixel distribution, and the histogram obtained under multiple scales is spliced,
So as to obtain it is multiple dimensioned under more horn of plenty and robust feature representation.
It can refer to and combine each other between the various embodiments described above of the present invention, to obtain more embodiments.For example, such as
Shown in Fig. 8, referred to each other, in conjunction with obtained embodiment for the various embodiments described above.With reference to Fig. 8, to one embodiment of the invention
The method 800 for image procossing provided is described in detail.
Method 800 starts from step 801, and detecting iris region using circle Hough transformation obtains annular iris image.
In step 803, annular iris image is arranged in polar coordinate system the rectangle being arranged in rectangular coordinate system.
In step 805, every bit is found in the rectangle in corresponding ring according to rectangular co-ordinate and polar correspondence
Corresponding points in shape iris image, 807 are thened follow the steps if finding, and 809 are thened follow the steps if not finding.
In step 807, it sets the value of the point in rectangle to the value of corresponding points in annular iris image.
In step 809, the value of the point in rectangle is fitted according to the value of peripheral point in annular iris image.
In step 811, under S scale, rectangle is divided into spatially partly overlapping multiple images block.For example,
Under first scale, rectangle is divided into 64 partly overlapping 3 × 3 image blocks.
In step 813, under each scale, multiple image block is converted into column vector.For example, in first scale
Under, 64 3 × 3 image blocks are for example converted to 9 × 1 column vector.
In step 815, under each scale, column vector is merged into sample matrix.For example, under first scale, it will
64 9 × 1 column vectors merge into 19 × 64 matrix.In step 817, under each scale, L layers are carried out to sample matrix
Principal component analysis, obtain main composition matrix.Such as the principal component analysis for first layer, for the sample matrix under scale i
Xi, utilize substrate Vi 1The sample matrix of reconstruct isThe target of principal component analysis is optimization Vi 1So that
Wherein I is unit matrix, and the purpose of bound term is so that substrate is orthonormal basis.Formula (2) uses singular value
Decomposition algorithm is solved.For example, obtained Vi 1It is made of the corresponding feature vector of one group of maximum characteristic value.By the substrate
Vi 1As main composition matrix.L layers of main composition matrix can be obtained by multiple iteration.For example, there are K main compositions, then
Obtain the main composition matrix of a 9 × K.
In step 819, under each scale, the main composition matrix under the scale is deformed into the image divided under the scale
The size of block.Such as under first scale, main composition matrix is deformed into 3 × 3 size.If main composition matrix is 9 × K
Matrix, then be deformed into the matrix of K 3 × 3 sizes.
In step 821, under each scale, using deformed main composition matrix as filtering core to pending image into
Row filtering.Such as under first scale, use the matrixes of deformed K 3 × 3 sizes as K filtering core to iris image
It is filtered, obtains K characteristic pattern.
In step 823, under each scale, obtained characteristic pattern is subjected to binarization operation, obtains the feature of binaryzation
Figure.For example, can use 0 as binarization operation threshold value.
In step 825, under each scale, value of each position in each characteristic pattern is formed to the table of the pixel of the position
Up to vector, or it is also assumed that it is binary value.For example, under first scale, by each position in K characteristic pattern
Value extracts one K dimensional vector of composition, in other words K binary values.
In step 827, under each scale, expression vector or binary value are converted to decimal value, thus by K
A characteristic pattern is converted into 1 characteristic pattern.
In step 829, under each scale, metric characteristic pattern is counted to obtain the histogram of pixel distribution.
In step 831, the histogram of each scale is spliced, obtains the histogram of final feature representation.
It can be seen that method 800 obtains iris image by circle Hough transformation first, then by by the iris image exhibition
It is split into rectangle, the texture information of iris region is sufficiently reserved while facilitating subsequent processing.Also, due on multiple scales
Handled, can obtain it is multiple dimensioned with the relevant convolution kernel of iris image, so as to preferably extract iris image
Feature enriches the diversity of extracted feature.Also, by the principal component analysis of multilayer, it can fully learn iris image
Feature is to obtain more adapting to the convolution kernel of extraction this feature.Meanwhile method 800 has also carried out feature coding to characteristic pattern, fully
The statistical information in characteristic pattern is excavated, being effectively compressed for feature is realized.Further, since histogram feature has high-rise language
Adopted information, method 800 are established between bottom visual signature and high-rise deduction and are contacted well.Therefore, by using method
800 can effectively improve the precision of iris feature expression.
Fig. 9 shows a kind of schematic block diagram of the device 900 for image procossing provided according to embodiments of the present invention.
The device 900 includes：Acquisition module 901 is configured as obtaining pending image；Divide module 903, is configured as segmentation institute
It states image and obtains multiple images block；Conversion module 905 is configured as each being converted to vector by described multiple images are in the block,
And the vector is merged into matrix；Analysis module 907 is configured as carrying out principal component analysis to the obtained matrix, obtain
To main composition matrix；And filter module 909, it is configured as extracting filtering core to described pending from the main composition matrix
Image is filtered.
In an embodiment of the present invention, acquisition module 901 for example including：Submodule is unfolded, is configured as annular image
Rectangle is expanded into, and using the rectangle as the pending image.
In an embodiment of the present invention, device 900 further comprises：Detection module is configured as the outside of detection iris
Edge and inward flange are to obtain the annular image.
In an embodiment of the present invention, expansion submodule includes：It fans ring and divides submodule, be configured as the circular chart
As being divided into multiple fan rings；It is radial to divide submodule, it is configured as radially dividing each of the multiple fan ring；Square
Submodule is unfolded in shape, is configured as being rectangle according to the image spread of division.
In an embodiment of the present invention, the radial submodule that divides is configured to：It radially divides described more
Each of a sector so that it is intensive close to the part of inward flange, it is sparse close to outer peripheral part.
In an embodiment of the present invention, expansion submodule includes：Rectangle sets submodule, is configured as setting rectangular co-ordinate
The size of rectangle described in system；Submodule is found, is configured as the transformation relation according to polar coordinates and rectangular co-ordinate, described in searching
Point in the annular image that each pair of point is answered in rectangle；First transformation submodule, if being configured as looking in the annular image
To corresponding point, then using the pixel value of the corresponding point as the value of the point in the rectangle；Second transformation submodule, is configured
If to can not find corresponding point in the annular image, the pixel value using the point on periphery in the annular image is intended
Cooperation is the value of the point in the rectangle.
In an embodiment of the present invention, filter module 909 includes：Deformation sub-module is configured as the main composition square
Battle array is deformed into the size of the image block obtained in the segmentation step；First filtering submodule, is configured as deformed institute
Main composition matrix is stated to be filtered the pending image as filtering core.
In one embodiment of the invention, device 900 further comprises replicated blocks, is configured as executing x following steps,
Described in x be natural number：Obtain the characteristic image after being filtered to the pending image；Using the characteristic image as institute
State the pending image obtained in pending image step；Segmentation described image obtains multiple images block；It will be described
Multiple images are in the block to be each converted to vector, and the vector is merged into matrix；To the obtained matrix carry out it is main at
Part analysis, obtains main composition matrix；The pending image is filtered from the main composition matrix extraction filtering core.
In one embodiment of the invention, device 900 further comprises statistical module, is configured as the feature to being obtained after filtering
Figure carries out statistical presentation.
In one embodiment of the invention, statistical module for example including：Histogram sub-module is configured as to obtaining after filtering
Characteristic pattern is counted, and the histogram of pixel distribution is obtained.
In one embodiment of the invention, histogram sub-module for example including：Binaryzation submodule is configured as under the scale
Obtained filtered characteristic pattern carries out binarization operation；Vectorization submodule is configured as each pixel after binarization
Characteristic pattern in corresponding position one as binary digit of value, obtain the binary value of each pixel；Convert submodule
Block is configured as converting the binary value of each pixel to decimal value；It is counted to obtain pixel point according to decimal value
The histogram of cloth.
In one embodiment of the invention, the multiple dimensioned scheme in reference method embodiment, device 900 can also use more rulers
Degree.Segmentation module is further configured to divide described image on multiple scales, obtains the multiple images block under each scale；
The conversion module is further configured under each scale, by the multiple images under the scale it is in the block it is each be converted to
Amount, and the vector is merged into matrix；The analysis module is further configured under each scale, described in obtaining
Matrix carries out principal component analysis, obtains the main composition matrix under the scale；The filter module is further configured to each
Under scale, the pending image is filtered from the main composition matrix extraction filtering core under the scale；The dress
Set and further comprise merging module, be configured as the filtered characteristic pattern that will be obtained under each scale in the multiple scale into
Row merges.
In one embodiment of the invention, merging module is configured to being obtained under each scale in the multiple scale
Filtered characteristic pattern carry out statistical presentation, and the statistical presentation under the multiple scale is merged.
The specific implementation mode of device 900 under the present invention is multiple dimensioned is referred to embodiment of the method, and details are not described herein again.
The specific implementation of apparatus of the present invention embodiment is referred to corresponding embodiment of the method, and details are not described herein again.
For clarity, all selectable units or subelement included by device 900 are not shown in Fig. 9.Above-mentioned side
All features and operation described in method embodiment and the embodiment by reference to that can be obtained with combination are respectively suitable for filling
900 are set, therefore details are not described herein.
It will be understood by those skilled in the art that the division of unit or subelement is not limiting but shows in device 900
Example property, be in order to more convenient it will be appreciated by those skilled in the art that logically describing its major function or operation.In device
In 900, the function of a unit can be realized by multiple units；Conversely, multiple units can also be realized by a unit.This
Invention limits not to this.
Likewise, it will be understood by those skilled in the art that various modes, which may be used, carrys out the list that realization device 900 is included
Member, including but not limited to software, hardware, firmware or its arbitrary combination, the present invention limit not to this.
The present invention can be system, method, computer-readable storage medium and/or computer program product.Computer
Readable storage medium storing program for executing for example can be the tangible device that can keep and store the instruction used by instruction execution equipment.
Computer-readable/executable program instruction can be downloaded to from computer readable storage medium each calculating/from
Equipment is managed, outer computer or External memory equipment can also be downloaded to by various communication modes.The present invention does not limit specifically
Make the specific programming language for realizing computer-readable/executable program instruction or instruction.
Referring herein to according to the method for the embodiment of the present invention, the flowchart and or block diagram of device (system) describe this hair
Bright various aspects.It should be appreciated that each box in each box and flowchart and or block diagram of flowchart and or block diagram
Combination can be realized by computer-readable/executable program instruction.
Various embodiments of the present invention are described above, it is stated that above description is exemplary in as described above,
And non-exclusive, and it is also not necessarily limited to disclosed each embodiment, each other can be referred between each embodiment and in conjunction with obtaining
More embodiments.Without departing from the scope and spirit of illustrated each embodiment, for the general of the art
Many modifications and changes will be apparent from for logical technical staff.
Claims (33)
1. a kind of method for image procossing, the method includes：
Obtain pending image；
Segmentation described image obtains multiple images block；
Each vector is converted to by described multiple images are in the block, and the vector is merged into matrix；
Principal component analysis is carried out to the obtained matrix, obtains main composition matrix；And
The pending image is filtered from the main composition matrix extraction filtering core.
2. according to the method described in claim 1, wherein, the pending image of the acquisition includes：It is rectangle by image spread,
And using the rectangle as the pending image.
3. according to the method described in claim 1, wherein, the pending image of the acquisition includes：
Annular image is expanded into rectangle, and using the rectangle as the pending image.
4. according to the method described in claim 3, wherein, it is described annular image is expanded into rectangle before, the method into
One step includes：
The outer edge and inward flange of detection iris are to obtain the annular image.
5. according to the method described in claim 3, wherein, the annular image, which is expanded into rectangle, includes：
The annular image is divided into multiple fan rings；
Radially divide each of the multiple fan ring；
Image spread according to division is rectangle.
6. described radially to divide each of the multiple sector, packet according to the method described in claim 5, wherein
It includes：Radially divide each of the multiple sector so that it is intensive close to the part of inward flange, close to outer peripheral portion
Divide sparse.
7. according to the method described in claim 3, wherein, the annular image, which is expanded into rectangle, includes：
Set the size of rectangle described in rectangular coordinate system；
According to the transformation relation of polar coordinates and rectangular co-ordinate, the point in the annular image that each pair of point is answered in the rectangle is found；
If finding corresponding point in the annular image, using the pixel value of the corresponding point as the point in the rectangle
Value；
If can not find corresponding point in the annular image, carried out using the pixel value of the point on periphery in the annular image
It is fitted the value as the point in the rectangle.
8. described to extract filtering core to described pending from the main composition matrix according to the method described in claim 1, wherein
Image be filtered, including：
The main composition matrix is deformed into the size of the image block obtained in the segmentation step；
The deformed main composition matrix is filtered the pending image as filtering core.
9. according to the method described in claim 1, wherein, the method further includes：Execute x following steps, wherein institute
It is natural number to state x：
Obtain the characteristic image after being filtered to the pending image；
Using the characteristic image as the pending image obtained in pending image step；
Segmentation described image obtains multiple images block；
Each vector is converted to by described multiple images are in the block, and the vector is merged into matrix；
Principal component analysis is carried out to the obtained matrix, obtains main composition matrix；And
The pending image is filtered from the main composition matrix extraction filtering core.
10. according to the method described in claim 1, the method further includes：It unites to the characteristic pattern obtained after filtering
Meter expression.
11. according to the method described in claim 10, wherein, the characteristic pattern obtained after described pair of filtering carries out statistical presentation, packet
It includes：The characteristic pattern obtained after filtering is counted, the histogram of pixel distribution is obtained.
12. according to the method for claim 11, wherein the characteristic pattern obtained after described pair of filtering counts, and obtains picture
The histogram of element distribution, including：
Binarization operation is carried out to the filtered characteristic pattern obtained under the scale；
Using the value of the corresponding position in the characteristic pattern of each pixel after binarization as one of binary digit, obtain each
The binary value of pixel；And
Convert the binary value of each pixel to decimal value；It is counted to obtain the histogram of pixel distribution according to decimal value
Figure.
13. according to the method described in claim 1, wherein,
The segmentation described image obtains multiple images block, including：Divide described image on multiple scales, obtains each scale
Under multiple images block；
It is described to be each converted to vector by described multiple images are in the block, and the vector is merged into matrix, including：Each
Under scale, each vector is converted to by the multiple images under the scale are in the block, and the vector is merged into matrix；
The described pair of obtained matrix carries out principal component analysis, obtains main composition matrix, including：Under each scale, to
The matrix arrived carries out principal component analysis, obtains the main composition matrix under the scale；
It is described that the pending image is filtered from the main composition matrix extraction filtering core, including：In each scale
Under, the pending image is filtered from the main composition matrix extraction filtering core under the scale；
The method further includes：The filtered characteristic pattern obtained under each scale in the multiple scale is closed
And.
14. according to the method for claim 13, wherein the filtering that will be obtained under each scale in the multiple scale
Characteristic pattern afterwards merges, including：It unites to the filtered characteristic pattern obtained under each scale in the multiple scale
Meter expression, and the statistical presentation under the multiple scale is merged.
15. according to the method for claim 14, wherein the filtering that will be obtained under each scale in the multiple scale
Characteristic pattern afterwards carries out statistical presentation, including：To the filtered characteristic pattern that is obtained under each scale in the multiple scale into
Row statistics, obtains the histogram of pixel distribution.
16. according to the method for claim 15, wherein
The filtered characteristic pattern obtained under each scale in the multiple scale is counted, the histogram of pixel distribution is obtained
Figure, including：Under each scale, binarization operation is carried out to the filtered characteristic pattern obtained under the scale；By each pixel
One as binary digit of the value of the corresponding position in characteristic pattern after binarization, obtains the binary system of each pixel
Value；Convert the binary value of each pixel to decimal value；It is counted to obtain the histogram of pixel distribution according to decimal value
Figure；
The statistical presentation by under the multiple scale merges, including：The histogram that will be obtained under the multiple scale
It is stitched together.
17. according to the method described in claim 1, wherein, described multiple images block is partly overlapping multiple images block.
18. a kind of device for image procossing, described device include：
Acquisition module is configured as obtaining pending image；
Divide module, is configured as segmentation described image and obtains multiple images block；
Conversion module is configured as each being converted to vector by described multiple images are in the block, and the vector is merged into square
Battle array；
Analysis module is configured as carrying out principal component analysis to the obtained matrix, obtains main composition matrix；And
Filter module is configured as being filtered the pending image from the main composition matrix extraction filtering core.
19. device according to claim 18, wherein the acquisition module includes：Submodule is unfolded, is configured as ring
Shape image spread is rectangle, and using the rectangle as the pending image.
20. device according to claim 19, wherein described device further comprises：Detection module is configured as detecting
The outer edge and inward flange of iris are to obtain the annular image.
21. device according to claim 19, wherein the expansion submodule includes：
It fans ring and divides submodule, be configured as the annular image being divided into multiple fan rings；
It is radial to divide submodule, it is configured as radially dividing each of the multiple fan ring；
Submodule is unfolded in rectangle, is configured as being rectangle according to the image spread of division.
22. device according to claim 21, wherein the radial submodule that divides is configured to：Radially
Direction divides each of the multiple sector so that and it is intensive close to the part of inward flange, it is sparse close to outer peripheral part.
23. device according to claim 19, wherein the expansion submodule includes：
Rectangle sets submodule, is configured as the size of rectangle described in setting rectangular coordinate system；
Submodule is found, the transformation relation according to polar coordinates and rectangular co-ordinate is configured as, finds each pair of point in the rectangle
Point in the annular image answered；
First transformation submodule, if being configured as finding corresponding point in the annular image, by the picture of the corresponding point
Value of the element value as the point in the rectangle；
Second transformation submodule utilizes the circular chart if being configured as can not find corresponding point in the annular image
The pixel value of the point on periphery is fitted the value as the point in the rectangle as in.
24. device according to claim 18, wherein the filter module, including：
Deformation sub-module is configured as the main composition matrix being deformed into the big of the image block obtained in the segmentation step
It is small；
First filtering submodule, is configured as using the deformed main composition matrix as filtering core to the pending figure
As being filtered.
25. device according to claim 18, wherein described device further comprises replicated blocks, is configured as executing x
Secondary following steps, wherein the x is natural number：
Obtain the characteristic image after being filtered to the pending image；
Using the characteristic image as the pending image obtained in pending image step；
Segmentation described image obtains multiple images block；
Each vector is converted to by described multiple images are in the block, and the vector is merged into matrix；
Principal component analysis is carried out to the obtained matrix, obtains main composition matrix；
The pending image is filtered from the main composition matrix extraction filtering core.
26. device according to claim 18, wherein described device further comprises statistical module, is configured as to filter
The characteristic pattern obtained after wave carries out statistical presentation.
27. device according to claim 26, wherein the statistical module includes：Histogram sub-module, is configured as pair
The characteristic pattern obtained after filtering is counted, and the histogram of pixel distribution is obtained.
28. device according to claim 27, wherein the histogram sub-module includes：
Binaryzation submodule is configured as carrying out binarization operation to the filtered characteristic pattern obtained under the scale；
Vectorization submodule, be configured as using the value of the corresponding position in each pixel characteristic pattern after binarization as two into
One of system number, obtains the binary value of each pixel；
Submodule is converted, is configured as converting the binary value of each pixel to decimal value；It is united according to decimal value
Meter obtains the histogram of pixel distribution.
29. device according to claim 18, wherein
The segmentation module is further configured to divide described image on multiple scales, obtains multiple figures under each scale
As block；
The conversion module is further configured under each scale, by each conversion in the block of the multiple images under the scale
For vector, and the vector is merged into matrix；
The analysis module is further configured under each scale, is carried out principal component analysis to the obtained matrix, is obtained
Main composition matrix under to the scale；
The filter module is further configured under each scale, from the main composition matrix extraction filtering under the scale
The pending image is checked to be filtered；
Described device further comprises merging module, after being configured as the filtering that will be obtained under each scale in the multiple scale
Characteristic pattern merge.
30. device according to claim 29, wherein the merging module is configured to the multiple scale
In the filtered characteristic pattern that obtains under each scale carry out statistical presentation, and the statistical presentation under the multiple scale is carried out
Merge.
31. device according to claim 18, wherein described multiple images block is partly overlapping multiple images block.
32. a kind of computer system for image procossing, including：
One or more processors；
One or more computer-readable mediums；
The computer journey for being executed by least one of one or more processors that may be stored on the computer-readable medium
Sequence instructs, and the computer program instructions include：
Computer program instructions for obtaining pending image；
The computer program instructions of multiple images block are obtained for dividing described image；
For being each converted to vector by described multiple images are in the block, and the vector is merged into the computer program of matrix
Instruction；
For carrying out principal component analysis to the obtained matrix, the computer program instructions of main composition matrix are obtained；And
Computer program instructions for being filtered from the main composition matrix extraction filtering core to the pending image.
33. a kind of computer readable storage medium for image procossing, be stored on the computer readable storage medium to
Few executable computer program instructions, the computer program instructions include appointing in requiring 1 to 17 for perform claim
The computer program instructions of each step of one method.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710163576 | 2017-03-18 | ||
| CN2017101635763 | 2017-03-18 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108629264A true CN108629264A (en) | 2018-10-09 |
| CN108629264B CN108629264B (en) | 2022-09-27 |
Family
ID=63705662
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710760469.9A Active CN108629264B (en) | 2017-03-18 | 2017-08-30 | Method and apparatus for image processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108629264B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021093499A1 (en) * | 2019-11-15 | 2021-05-20 | RealMe重庆移动通信有限公司 | Image processing method and apparatus, storage medium, and electronic device |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1776710A (en) * | 2005-11-17 | 2006-05-24 | 上海交通大学 | Body iris texture non-linear normalizing method |
| US20070014435A1 (en) * | 2005-07-13 | 2007-01-18 | Schlumberger Technology Corporation | Computer-based generation and validation of training images for multipoint geostatistical analysis |
| US20100239130A1 (en) * | 2009-03-18 | 2010-09-23 | Industrial Technology Research Institute | System and method for performing rapid facial recognition |
| CN101916363A (en) * | 2010-05-28 | 2010-12-15 | 深圳大学 | Iris characteristic designing and coding method and iris identifying system |
| CN105512684A (en) * | 2015-12-09 | 2016-04-20 | 江苏大为科技股份有限公司 | Vehicle logo automatic identification method based on principal component analysis convolutional neural network |
| CN105550661A (en) * | 2015-12-29 | 2016-05-04 | 北京无线电计量测试研究所 | Adaboost algorithm-based iris feature extraction method |
| CN105678774A (en) * | 2016-01-11 | 2016-06-15 | 浙江传媒学院 | Image noise level estimation method based on principal component analysis |
| CN105787488A (en) * | 2016-03-02 | 2016-07-20 | 浙江宇视科技有限公司 | Image feature extraction method and device realizing transmission from whole to local |
| CN105956571A (en) * | 2016-05-13 | 2016-09-21 | 华侨大学 | Age estimation method for face image |
| CN106022358A (en) * | 2016-05-11 | 2016-10-12 | 湖南大学 | Hyper-spectral image classification method and hyper-spectral image classification device |
| CN106485259A (en) * | 2015-08-26 | 2017-03-08 | 华东师范大学 | A kind of image classification method based on high constraint high dispersive principal component analysiss network |
-
2017
- 2017-08-30 CN CN201710760469.9A patent/CN108629264B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070014435A1 (en) * | 2005-07-13 | 2007-01-18 | Schlumberger Technology Corporation | Computer-based generation and validation of training images for multipoint geostatistical analysis |
| CN1776710A (en) * | 2005-11-17 | 2006-05-24 | 上海交通大学 | Body iris texture non-linear normalizing method |
| US20100239130A1 (en) * | 2009-03-18 | 2010-09-23 | Industrial Technology Research Institute | System and method for performing rapid facial recognition |
| CN101916363A (en) * | 2010-05-28 | 2010-12-15 | 深圳大学 | Iris characteristic designing and coding method and iris identifying system |
| CN106485259A (en) * | 2015-08-26 | 2017-03-08 | 华东师范大学 | A kind of image classification method based on high constraint high dispersive principal component analysiss network |
| CN105512684A (en) * | 2015-12-09 | 2016-04-20 | 江苏大为科技股份有限公司 | Vehicle logo automatic identification method based on principal component analysis convolutional neural network |
| CN105550661A (en) * | 2015-12-29 | 2016-05-04 | 北京无线电计量测试研究所 | Adaboost algorithm-based iris feature extraction method |
| CN105678774A (en) * | 2016-01-11 | 2016-06-15 | 浙江传媒学院 | Image noise level estimation method based on principal component analysis |
| CN105787488A (en) * | 2016-03-02 | 2016-07-20 | 浙江宇视科技有限公司 | Image feature extraction method and device realizing transmission from whole to local |
| CN106022358A (en) * | 2016-05-11 | 2016-10-12 | 湖南大学 | Hyper-spectral image classification method and hyper-spectral image classification device |
| CN105956571A (en) * | 2016-05-13 | 2016-09-21 | 华侨大学 | Age estimation method for face image |
Non-Patent Citations (6)
| Title |
|---|
| MEERA MARY ISAAC 等: "Image forgery detection based on Gabor Wavelets and Local Phase Quantization", 《PROCEDIA COMPUTER SCIENCE》 * |
| NICOLAS PINTO: "A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation", 《PLOS COMPUTATIONAL BIOLOGY》 * |
| 张顺利: "基于滤波特性的虹膜识别算法研究", 《万方数据知识服务平台》 * |
| 樊继聪: "基于多元统计分析的非高斯过程的故障诊断", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 纪效存: "虹膜识别算法中的关键问题研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 董宏兴: "基于自适应Gabor滤波的虹膜特征提取与识别方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021093499A1 (en) * | 2019-11-15 | 2021-05-20 | RealMe重庆移动通信有限公司 | Image processing method and apparatus, storage medium, and electronic device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108629264B (en) | 2022-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Suganuma et al. | Attention-based adaptive selection of operations for image restoration in the presence of unknown combined distortions | |
| Esteves et al. | Polar transformer networks | |
| CN109522874B (en) | Human body action recognition method and device, terminal equipment and storage medium | |
| Hadsell et al. | Dimensionality reduction by learning an invariant mapping | |
| CN105184303B (en) | A kind of image labeling method based on multi-modal deep learning | |
| Huang et al. | Vehicle logo recognition system based on convolutional neural networks with a pretraining strategy | |
| Naikal et al. | Informative feature selection for object recognition via sparse PCA | |
| Bian et al. | Multiscale fully convolutional network with application to industrial inspection | |
| Cheng et al. | Learning coarse-to-fine sparselets for efficient object detection and scene classification | |
| Wang et al. | Bag of contour fragments for robust shape classification | |
| Hossain et al. | Leaf shape identification based plant biometrics | |
| Borji et al. | Human vs. computer in scene and object recognition | |
| CN108154194A (en) | A kind of method with the convolutional network extraction high dimensional feature based on tensor | |
| US20170256068A1 (en) | Leveraging multi cues for fine-grained object classification | |
| CN115937655B (en) | Multi-order feature interaction target detection model, construction method, device and application thereof | |
| Follmann et al. | A rotationally-invariant convolution module by feature map back-rotation | |
| Yan et al. | Image classification with densely sampled image windows and generalized adaptive multiple kernel learning | |
| CN103839066A (en) | Feature extraction method from biological vision | |
| CN112150359B (en) | Unmanned aerial vehicle image fast splicing method based on machine learning and feature point identification | |
| Zhang et al. | Starting from non-parametric networks for 3d point cloud analysis | |
| CN111680595A (en) | Face recognition method and device and electronic equipment | |
| Guan et al. | MAN and CAT: mix attention to nn and concatenate attention to YOLO | |
| Satta et al. | Exploiting dissimilarity representations for person re-identification | |
| CN111382638A (en) | Image detection method, device, equipment and storage medium | |
| CN108629264A (en) | Method and apparatus for image procossing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230411 Address after: No.3 Factory Building, No. 198 Ruiyang Avenue, Xiping Street, Songyang County, Lishui City, Zhejiang Province, 323400 Patentee after: Zhejiang Jinghong Technology Co.,Ltd. Address before: No. 400 Fangchun Road, Pudong New Area Free Trade Pilot Area, Shanghai, 200120 Patentee before: SHANGHAI JINGHONG ELECTRONIC TECHNOLOGY Co.,Ltd. |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Methods and devices for image processing Effective date of registration: 20231205 Granted publication date: 20220927 Pledgee: Zhejiang Tailong Commercial Bank Co.,Ltd. Lishui Songyang Branch Pledgor: Zhejiang Jinghong Technology Co.,Ltd. Registration number: Y2023980069347 |