CN101887590A - Method for displaying visualization organization of digital images - Google Patents

Method for displaying visualization organization of digital images Download PDF

Info

Publication number
CN101887590A
CN101887590A CN 201010209558 CN201010209558A CN101887590A CN 101887590 A CN101887590 A CN 101887590A CN 201010209558 CN201010209558 CN 201010209558 CN 201010209558 A CN201010209558 A CN 201010209558A CN 101887590 A CN101887590 A CN 101887590A
Authority
CN
China
Prior art keywords
digital image
sample
space
original digital
matrix
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
Application number
CN 201010209558
Other languages
Chinese (zh)
Other versions
CN101887590B (en
Inventor
于达仁
贺惠新
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN2010102095582A priority Critical patent/CN101887590B/en
Publication of CN101887590A publication Critical patent/CN101887590A/en
Application granted granted Critical
Publication of CN101887590B publication Critical patent/CN101887590B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention discloses a method for displaying the visualization organization of digital images, which relates to the field of information technology and solves the problem of large amount of calculation of the conventional image visualization methods. The method comprises the following steps of: extracting an image characteristic value consisting of an original image characteristic value and an extended image characteristic value; substituting the image characteristic value for an image, measuring the distance to accomplish the clustering of all the image samples and detecting remote samples in a low-dimensional space according to the clustering result; and performing a uniform organization visualization display on the images excluding the remote samples. The method is suitable for the image visualization.

Description

A kind of method of displaying visualization organization of digital images
Technical field
The present invention relates to areas of information technology, be specifically related to a kind of method of displaying visualization organization of digital images.
Background technology
Present feature extraction mode for image viewing, the method of considering details and distribution needs very big calculated amount usually, and histogram is easy to just abandon image space information, and the pixel total amount value that simple segmentation is calculated has also directly been ignored the distribution character of pixel value when calculating.
For the organizational form of image viewing, generally all based on predefined classification, each image is filed in the different classifications at present, and such classification is stiff relatively, and is very big to the threshold value dependence of kind judging.
Summary of the invention
In order to solve the big problem of existing image viewing method calculated amount, the invention provides a kind of method of displaying visualization organization of digital images.
The method of a kind of displaying visualization organization of digital images of the present invention, it may further comprise the steps:
Step 1: obtain the original digital image P that G opens un-marked k(m, n), 1≤k≤G;
Step 2: extract described every original digital image P k(m, primitive character amount FS n) k(z) and the extension feature amount FE of expanded images k(z), obtain described every original digital image P k(m, image feature amount PF n) k=F k(l), described F k(l) expression k opens original digital image P k(m, primitive character amount FS n) k(z) and extension feature amount FE k(z) merging characteristic quantity, and use the image feature amount PF that is obtained kExpression original digital image P k(m, n), wherein, z=1,2,3..., 24;
Step 3: measure b with p norm distance and open original digital image and g and open similarity distance between the original digital image
Figure BSA00000176816500011
B=1,2...G, g=1,2...G, and set up overall distance matrix , wherein, a lBe described two image feature value PF bAnd PF gThe distance calculation weights, get a lFor
Figure BSA00000176816500021
L dimension amount,
Figure BSA00000176816500022
It is the feature weight vector of 48 dimensions;
Step 4: G is opened original digital image open sample, and open from described G and to obtain the cluster centre sample that each sample belongs to the sample as G
Figure BSA00000176816500023
Wherein, (k f) is described validity matrix A to a G * GIn the k row element, (k f) is described feedback matrix R to r G * GIn the k row element;
Step 5: according to overall distance matrix D G * GWith the cluster centre sample fc that each sample belonged to that is obtained kObtain the eigenvector E (k) of each sample, and then from all samples, detect remote sample;
Step 6: G is opened sample be transformed to the little image of G hypertonic, and generate the angular deflection amount of described each downscaled images
Figure BSA00000176816500024
Rand (*) is [0,1] random value on, adjust described downscaled images with the angular deflection amount Ap (k) that is obtained, and make described downscaled images in the target spacial flex, face user's sightingpiston with eigenvector E (k) displaying, to finish displaying visualization organization of digital images.
Beneficial effect of the present invention: the invention provides a kind of based on the cluster of adapting to image fast of visual experience and then the method for realization displaying visualization organization, the present invention is when sorting out the picture tissue, digital picture is transformed into the HIS space to be handled, make image viewing more meet human vision perception characteristic, organize in the displaying process at image viewing, utilize the image feature value presentation video and then get rid of remote sample, make that the calculated amount in the image viewing process is little.
Description of drawings
Fig. 1 is the process flow diagram of the method for a kind of displaying visualization organization of digital images of the present invention.
Embodiment
Embodiment one: specify present embodiment according to Figure of description 1, the method for the described a kind of displaying visualization organization of digital images of present embodiment, it may further comprise the steps:
Step 1: obtain the original digital image P that G opens un-marked k(m, n), 1≤k≤G;
Step 2: extract described every original digital image P k(m, primitive character amount FS n) k(z) and the extension feature amount FE of expanded images k(z), obtain described every original digital image P k(m, image feature amount PF n) k=F k(l), described F k(l) expression k opens original digital image P k(m, primitive character amount FS n) k(z) and extension feature amount FE k(z) merging characteristic quantity, and use the image feature amount PF that is obtained kExpression original digital image P k(m, n), wherein, z=1,2,3..., 24;
Step 3: measure b with p norm distance and open original digital image and g and open similarity distance between the original digital image B=1,2...G, g=1,2...G, and set up overall distance matrix
Figure BSA00000176816500032
, wherein, a lBe described two image feature value PF bAnd PF gThe distance calculation weights, get a lFor
Figure BSA00000176816500033
L dimension amount,
Figure BSA00000176816500034
It is the feature weight vector of 48 dimensions;
Step 4: G is opened original digital image open sample, and open from described G and to obtain the cluster centre sample that each sample belongs to the sample as G Wherein, (k f) is described validity matrix A to a G * GIn the k row element, (k f) is described feedback matrix R to r G * GIn the k row element;
Step 5: according to overall distance matrix D G * GWith the cluster centre sample fc that each sample belonged to that is obtained kObtain the eigenvector E (k) of each sample, and then from all samples, detect remote sample;
Step 6: G is opened sample be transformed to the little image of G hypertonic, and generate the angular deflection amount of described each downscaled images
Figure BSA00000176816500036
Rand (*) is [0,1] random value on, adjust described downscaled images with the angular deflection amount Ap (k) that is obtained, and make described downscaled images in the target spacial flex, face user's sightingpiston with eigenvector E (k) displaying, to finish displaying visualization organization of digital images.
Embodiment two: present embodiment is that embodiment one is extracted described every original digital image P in step 2 to the further specifying of embodiment one k(m, primitive character amount FS n) k(z) concrete grammar is:
Step 2 is one by one: with original digital image P k(m, n) with the RGB color coordinates as storage, i.e. P k(m, n)=(P R, P G, P B) M, n, wherein, 1≤m≤M k, 1≤n≤N k, this original digital image is of a size of M k* N k, M kAnd N kAll more than or equal to 1;
Step 2 one or two: with original digital image P k(m, n)=(P R, P G, P B) M, nAt the HIS space representation is P k(m, n)=(I M, n, H M, n, S M, n), and then calculate this original digital image P k(m, n) the pixel probability histogram in each space in HIS space
Figure BSA00000176816500041
T=1,2 ..., mn, θ=I, H, S, wherein, S θ(q t) be illustrated in original digital image P in the θ space k(m, pixel value n) are q tNumber of pixels,
Figure BSA00000176816500042
Be original digital image P in the θ space k(m, all number of pixels n);
Step 2 one or three: in each space in HIS space, according to original digital image P k(m, all pixel value q n) tThe codomain scope of forming on average is divided into 8 sections with this codomain scope, and then obtains original digital image P k(m is n) at the characteristic quantity in each space W=1,2 ..., 8, q Min=min (q t), q Max=max (q t), d Ar=(q Max-q Min)/8, and with the original digital image P that obtains k(m is n) at the characteristic quantity H in I space I(w) expression, at the characteristic quantity H in H space H(w) expression, the characteristic quantity H in the S space S(w) expression;
Step 2 one or four: according to the original digital image P that is obtained k(m, n) 8 characteristic quantities in each space in HIS space obtain original digital image P k(m is n) at 24 primitive character amount FS in HIS space k(z).
Embodiment three: present embodiment is to the further specifying of embodiment two, in the embodiment two in step 2 one or two, with original digital image P k(m, n)=(P R, P G, P B) M, nAt the HIS space representation is P k(m, n)=(I M, n, H M, n, S M, n) concrete grammar be:
At first calculate
Figure BSA00000176816500051
And then obtain H M, nAnd S M, n,
Work as P R=min (P R, P G, P B) time, then
H m , n = P B - P R 3 ( I - P R ) + 1 S m , n = 1 - P R I
Work as P G=min (P R, P G, P B) time, then
H m , n = P R - P G 3 ( I - P G ) + 1 S m , n = 1 - P G I
Work as P B=min (P R, P G, P B) time, then
H m , n = P G - P B 3 ( I - P B ) + 1 S m , n = 1 - P B I
Finally finish original digital image P k(m, n)=(P R, P G, P B) M, nAt the HIS space representation is P k(m, n)=(I M, n, H M, n, S M, n).
Embodiment four: present embodiment is to the further specifying of embodiment one, two or three, and in the embodiment one, two or three, in step 2, extracts described every original digital image P k(m, the extension feature amount FE of expanded images n) k(z) concrete grammar is:
Step 221: with original digital image P k(m, n) with the RGB color coordinates as storage, i.e. P k(m, n)=(P R, P G, P B) M, n, wherein, 1≤m≤M k, 1≤n≤N k, original digital image is of a size of M k* N k, M kAnd N kAll more than or equal to 1, and with described original digital image P k(m n) expands to expanded images Q k(u, v),
Figure BSA00000176816500061
Step 2 two or two: to described original digital image P k(m, expanded images Q n) k(u v) makes smoothing processing, obtains described expanded images Q k(u, level and smooth result v)
Y k ( m , n ) = median Q k ( m - 1 , n - 1 ) , Q k ( m - 1 , n ) , Q k ( m - 1 , n + 1 ) , Q k ( m , n - 1 ) , Q k ( m , n ) , Q k ( m , n + 1 ) , Q k ( m + 1 , m - 1 ) , Q k ( m + 1 , n ) , Q k ( m + 1 , n + 1 ) ;
Step 2 two or three: obtain described original digital image P k(m, expanded images Q n) k(u, level and smooth Y as a result v) k(m, n) the pixel probability histogram in each space in HIS space
Figure BSA00000176816500063
Wherein, S θ is level and smooth(q ' t) be illustrated in the θ space level and smooth Y as a result k(m, pixel value n) are q ' tNumber of pixels,
Figure BSA00000176816500064
For at level and smooth Y as a result described in the θ space k(m, all number of pixels n);
Step 2 two or four: in each space in HIS space, according to described level and smooth Y as a result k(m, all pixel value q ' n) tThe codomain scope of forming is with described all pixel value q ' tThe codomain scope on average be divided into 8 sections, and then obtain described level and smooth Y as a result k(m is n) at the characteristic quantity in each space
Figure BSA00000176816500065
Q ' Min=min (q ' t), q ' Max=max (q ' t), d ' Ar=(q ' Max-q ' Min)/8, and described each level and smooth Y as a result that will obtain k(m is n) at the characteristic quantity H in I space Level and smooth I(w) expression, at the characteristic quantity H in H space Level and smooth H(w) expression, the characteristic quantity H in the S space Level and smooth S(w) expression;
Step 2 two or five: the level and smooth Y as a result that is obtained according to step 2 two or four k(m, n) 8 characteristic quantities in each space in HIS space obtain original digital image P k(m is n) at 24 extension feature amount FE in HIS space k(z).
Embodiment five: present embodiment is to the further specifying of any one embodiment in the embodiment one to four, and in step 4, opens from described G and to obtain the cluster centre sample that each sample belongs to the sample in the embodiment one to four Concrete grammar be:
Step 4 one: obtain b and open original digital image and g and open similarity between the original digital image
Figure BSA00000176816500072
And set up overall similar matrix
Figure BSA00000176816500073
Wherein, (i j) is described overall similar matrix S to Sim G * GIn the element of the capable j of i row;
Step 4 two: set up feedback matrix R G * G, and make described feedback matrix R G * GAll elements r (i, initial value j) is 0;
Step 4 three: set up the validity matrix A G * G, and make described validity matrix A G * GAll elements a (i, initial value j) is 0;
Step 4 four: to feedback matrix R G * GWith the validity matrix A G * GCarry out iteration LT time, and then obtain the feedback matrix R after LT iteration upgraded G * GWith effect property matrix A G * G, the detailed process of described each iteration is:
Step 441: with the feedback matrix R that sets up G * GIn all elements r (i j) is updated to r (i.j) *=λ r Tmp(i, j)+(1-λ) r Old(i, j), wherein, λ is a update coefficients, r Old(i, j) be last iteration upgrade the r that obtains (i, j), r tmp ( i , j ) = Sim ( i , j ) - max j ′ ≠ j { a tmp ( i , j ′ ) + Sim ( i , j ′ ) } , i ≠ j Sim ( j , j ) - max j ′ ≠ j { Sim ( i , j ′ ) } , i = j ,
a TmpThe initial value of (i, j ') is 0, and execution in step 442 then;
Step 4 four or two: with the validity matrix A of setting up G * GIn all elements a (i j) is updated to a (i.j) *=λ a Tmp(i, j)+(1-λ) a Old(i, j), wherein, a Old(i, j) be last iteration upgrade a that obtains (i, j),
a tmp ( i , j ) = min { 0 , r ( i , j ) + Σ i ′ ≠ i , j max { 0 , r ( i ′ , j ) } } , i ≠ j Σ i ′ ≠ j max { 0 , r ( i ′ , j ) } , i = j ;
Step 4 five: according to described feedback matrix R after carrying out upgrading for LT time G * GWith effect property matrix A G * GObtain the cluster centre sample that each sample belongs to
Figure BSA00000176816500083
Embodiment six: present embodiment is to the further specifying of any one embodiment in the embodiment one to five, in the embodiment one to five in step 5, according to overall distance matrix D G * GWith the cluster centre sample fc that each sample belonged to that is obtained kObtain the eigenvector E (k) of each sample, and then the concrete grammar of the remote sample of detection is from all samples:
Step 5 is one by one: with described overall distance matrix D G * GIn each element Dist (PF b, PF g) be updated to
Figure BSA00000176816500084
And then obtain described overall distance matrix D G * GRenewal matrix D Ad G * G
Step May Day two: utilize multi-dimentional scale conversion MDS algorithm, from described overall distance matrix D G * GRenewal matrix D Ad G * GIn extract eigenvector E (k)=[x of each sample when three dimensions is represented k, y k, z k];
Step May Day three: the data center that obtains all samples
Figure BSA00000176816500091
Step May Day four: obtain each sample and the described E of data center CtDistance And obtain d Med(3)=median (d k(3));
The step 5 First Five-Year Plan: judge d k(3) 〉=5*d Med(3), if judge that then described sample is remote sample.
In the present embodiment, when displaying visualization organization, also can add remote sample.
Embodiment seven: present embodiment is to the further specifying of any one embodiment in the embodiment one to five, in the embodiment one to five in step 5, according to overall distance matrix D G * GWith the cluster centre sample fc that each sample belonged to that is obtained kObtain the eigenvector E (k) of each sample, and then the concrete grammar of the remote sample of detection is from all samples:
Step 521: with described overall distance matrix D G * GIn each element Dist (PF b, PF g) be updated to
Figure BSA00000176816500093
And then obtain described overall distance matrix D G * GRenewal matrix D Ad G * G
Step 5 two or two: utilize multi-dimentional scale conversion MDS algorithm, from described overall distance matrix D G * GRenewal matrix D Ad G * GIn extract eigenvector E (k)=[x of each sample when representing with two-dimensional space k, y k];
Step 5 two or three: the data center that obtains all samples
Figure BSA00000176816500094
Step 5 two or four: obtain each sample and the described E of data center CtDistance
Figure BSA00000176816500095
And obtain d Med(2)=median (d k(2));
Step 5 two or five: judge d k(2) 〉=5*d Med(2), if judge that then described sample is remote sample.
In the present embodiment, when displaying visualization organization, also can add remote sample.

Claims (7)

1. the method for a displaying visualization organization of digital images is characterized in that it may further comprise the steps:
Step 1: obtain the original digital image P that G opens un-marked k(m, n), 1≤k≤G;
Step 2: extract described every original digital image P k(m, primitive character amount FS n) k(z) and the extension feature amount FE of expanded images k(z), obtain described every original digital image P k(m, image feature amount PF n) k=F k(l), described F k(l) expression k opens original digital image P k(m, primitive character amount FS n) k(z) and extension feature amount FE k(z) merging characteristic quantity, and use the image feature amount PF that is obtained kExpression original digital image P k(m, n), wherein, z=1,2,3..., 24;
Step 3: measure b with p norm distance and open original digital image and g and open similarity distance between the original digital image B=1,2...G, g=1,2...G, and set up overall distance matrix , wherein, a lBe described two image feature value PF bAnd PF gThe distance calculation weights, get a lFor
Figure FSA00000176816400013
L dimension amount,
Figure FSA00000176816400014
It is the feature weight vector of 48 dimensions;
Step 4: G is opened original digital image open sample, and open from described G and to obtain the cluster centre sample that each sample belongs to the sample as G
Figure FSA00000176816400015
Wherein, (k f) is described validity matrix A to a G * GIn the k row element, (k f) is described feedback matrix R to r G * GIn the k row element;
Step 5: according to overall distance matrix D G * GWith the cluster centre sample fc that each sample belonged to that is obtained kObtain the eigenvector E (k) of each sample, and then from all samples, detect remote sample;
Step 6: G is opened sample be transformed to the little image of G hypertonic, and generate the angular deflection amount of described each downscaled images
Figure FSA00000176816400016
Rand (*) is [0,1] random value on, adjust described downscaled images with the angular deflection amount Ap (k) that is obtained, and make described downscaled images in the target spacial flex, face user's sightingpiston with eigenvector E (k) displaying, to finish displaying visualization organization of digital images.
2. the method for a kind of displaying visualization organization of digital images according to claim 1 is characterized in that in step 2, extracts described every original digital image P k(m, primitive character amount FS n) k(z) concrete grammar is:
Step 2 is one by one: with original digital image P k(m, n) with the RGB color coordinates as storage, i.e. P k(m, n)=(P R, P G, P B) M, n, wherein, 1≤m≤M k, 1≤n≤N k, this original digital image is of a size of M k* N k, M kAnd N kAll more than or equal to 1;
Step 2 one or two: with original digital image P k(m, n)=(P R, P G, P B) M, nAt the HIS space representation is P k(m, n)=(I M, n, H M, n, S M, n), and then calculate this original digital image P k(m, n) the pixel probability histogram in each space in HIS space
Figure FSA00000176816400021
T=1,2 ..., mn, θ=I, H, S, wherein, S θ(q t) be illustrated in original digital image P in the θ space k(m, pixel value n) are q tNumber of pixels, Be original digital image P in the θ space k(m, all number of pixels n);
Step 2 one or three: in each space in HIS space, according to original digital image P k(m, all pixel value q n) tThe codomain scope of forming on average is divided into 8 sections with this codomain scope, and then obtains original digital image P k(m is n) at the characteristic quantity in each space
Figure FSA00000176816400023
W=1,2 ..., 8, q Min=min (q t), q Max=max (q t), d Ar=(q Max-q Min)/8, and with the original digital image P that obtains k(m is n) at the characteristic quantity H in I space I(w) expression, at the characteristic quantity H in H space H(w) expression, the characteristic quantity H in the S space S(w) expression;
Step 2 one or four: according to the original digital image P that is obtained k(m, n) 8 characteristic quantities in each space in HIS space obtain original digital image P k(m is n) at 24 primitive character amount FS in HIS space k(z).
3. the method for a kind of displaying visualization organization of digital images according to claim 2 is characterized in that in step 2 one or two, with original digital image P k(m, n)=(P R, P G, P B) M, nAt the HIS space representation is P k(m, n)=(I M, n, H M, n, S M, n) concrete grammar be:
At first calculate
Figure FSA00000176816400031
And then obtain H M, nAnd S M, n,
Work as P R=min (P R, P G, P B) time, then
H m , n = P B - P R 3 ( I - P R ) + 1 S m , n = 1 - P R I
Work as P G=min (P R, P G, P B) time, then
H m , n = P R - P G 3 ( I - P G ) + 1 S m , n = 1 - P G I
Work as P B=min (P R, P G, P B) time, then
H m , n = P G - P B 3 ( I - P B ) + 1 S m , n = 1 - P B I
Finally finish original digital image P k(m, n)=(P R, P G, P B) M, nAt the HIS space representation is P k(m, n)=(I M, n, H M, n, S M, n).
4. the method for a kind of displaying visualization organization of digital images according to claim 1 is characterized in that in step 2, extracts described every original digital image P k(m, the extension feature amount FE of expanded images n) k(z) concrete grammar is:
Step 221: with original digital image P k(m, n) with the RGB color coordinates as storage, i.e. P k(m, n)=(P R, P G, P B) M, n, wherein, 1≤m≤M k, 1≤n≤N k, original digital image is of a size of M k* N k, M kAnd N kAll more than or equal to 1, and with described original digital image P k(m n) expands to expanded images Q k(u, v),
Figure FSA00000176816400041
Step 2 two or two: to described original digital image P k(m, expanded images Q n) k(u v) makes smoothing processing, obtains described expanded images Q k(u, level and smooth result v)
Y k ( m , n ) = median Q k ( m - 1 , n - 1 ) , Q k ( m - 1 , n ) , Q k ( m - 1 , n + 1 ) , Q k ( m , n - 1 ) , Q k ( m , n ) , Q k ( m , n + 1 ) , Q k ( m + 1 , m - 1 ) , Q k ( m + 1 , n ) , Q k ( m + 1 , n + 1 ) ;
Step 2 two or three: obtain described original digital image P k(m, expanded images Q n) k(u, level and smooth Y as a result v) k(m, n) the pixel probability histogram in each space in HIS space
Figure FSA00000176816400043
Wherein, S θ is level and smooth(q ' t) be illustrated in the θ space level and smooth Y as a result k(m, pixel value n) are q ' tNumber of pixels,
Figure FSA00000176816400044
For at level and smooth Y as a result described in the θ space k(m, all number of pixels n);
Step 2 two or four: in each space in HIS space, according to described level and smooth Y as a result k(m, all pixel value q ' n) tThe codomain scope of forming is with described all pixel value q ' tThe codomain scope on average be divided into 8 sections, and then obtain described level and smooth Y as a result k(m is n) at the characteristic quantity in each space
Figure FSA00000176816400045
Q ' Min=min (q ' t), q ' Max=max (q ' t), d ' Ar=(q ' Max-q ' Min)/8, and described each level and smooth Y as a result that will obtain k(m is n) at the characteristic quantity H in I space Level and smooth I(w) expression, at the characteristic quantity H in H space Level and smooth H(w) expression, the characteristic quantity H in the S space Level and smooth S(w) expression;
Step 2 two or five: the level and smooth Y as a result that is obtained according to step 2 two or four k(m, n) 8 characteristic quantities in each space in HIS space obtain original digital image P k(m is n) at 24 extension feature amount FE in HIS space k(z).
5. the method for a kind of displaying visualization organization of digital images according to claim 1 is characterized in that in step 4, opens from described G and obtains the cluster centre sample that each sample belongs to the sample
Figure FSA00000176816400051
Concrete grammar be:
Step 4 one: obtain b and open original digital image and g and open similarity between the original digital image And set up overall similar matrix
Figure FSA00000176816400053
Wherein, (i j) is described overall similar matrix S to Sim G * GIn the element of the capable j of i row;
Step 4 two: set up feedback matrix R G * G, and make described feedback matrix R G * GAll elements r (i, initial value j) is 0;
Step 4 three: set up the validity matrix A G * G, and make described validity matrix A G * GAll elements a (i, initial value j) is 0;
Step 4 four: to feedback matrix R G * GWith the validity matrix A G * GCarry out iteration LT time, and then obtain the feedback matrix R after LT iteration upgraded G * GWith effect property matrix A G * G, the detailed process of described each iteration is:
Step 441: with the feedback matrix R that sets up G * GIn all elements r (i j) is updated to r (i.j) *=λ r Tmp(i, j)+(1-λ) r Old(i, j), wherein, λ is a update coefficients, r Old(i, j) be last iteration upgrade the r that obtains (i, j), r tmp ( i , j ) = Sim ( i , j ) - max j ′ ≠ j { a tmp ( i , j ′ ) + Sim ( i , j ′ ) } , i ≠ j Sim ( j , j ) - max j ′ ≠ j { Sim ( i , j ′ ) } , i = j , a TmpThe initial value of (i, j ') is 0, and execution in step 442 then;
Step 4 four or two: with the validity matrix A of setting up G * GIn all elements a (i j) is updated to a (i.j) *=λ a Tmp(i, j)+(1-λ) a Old(i, j), wherein, a Old(i, j) be last iteration upgrade a that obtains (i, j),
a tmp ( i , j ) = min { 0 , r ( i , j ) + Σ i ′ ≠ i , j max { 0 , r ( i ′ , j ) } } , i ≠ j Σ i ′ ≠ j max { 0 , r ( i ′ , j ) } , i = j ;
Step 4 five: according to described feedback matrix R after carrying out upgrading for LT time G * GWith effect property matrix A G * GObtain the cluster centre sample that each sample belongs to
6. the method for a kind of displaying visualization organization of digital images according to claim 1 is characterized in that in step 5, according to overall distance matrix D G * GWith the cluster centre sample fc that each sample belonged to that is obtained kObtain the eigenvector E (k) of each sample, and then the concrete grammar of the remote sample of detection is from all samples:
Step 5 is one by one: with described overall distance matrix D G * GIn each element Dist (PF b, PF g) be updated to
Figure FSA00000176816400064
And then obtain described overall distance matrix D G * GRenewal matrix D Ad G * G
Step May Day two: utilize multi-dimentional scale conversion MDS algorithm, from described overall distance matrix D G * GRenewal matrix D Ad G * GIn extract eigenvector E (k)=[x of each sample when three dimensions is represented k, y k, z k];
Step May Day three: the data center that obtains all samples
Figure FSA00000176816400071
Step May Day four: obtain each sample and the described E of data center CtDistance
Figure FSA00000176816400072
And obtain d Med(3)=median (d k(3));
The step 5 First Five-Year Plan: judge d k(3) 〉=5*d Med(3), if judge that then described sample is remote sample.
7. the method for a kind of displaying visualization organization of digital images according to claim 1 is characterized in that in step 5, according to overall distance matrix D G * GWith the cluster centre sample fc that each sample belonged to that is obtained kObtain the eigenvector E (k) of each sample, and then the concrete grammar of the remote sample of detection is from all samples:
Step 521: with described overall distance matrix D G * GIn each element Dist (PF b, PF g) be updated to
Figure FSA00000176816400073
And then obtain described overall distance matrix D G * GRenewal matrix D Ad G * G
Step 5 two or two: utilize multi-dimentional scale conversion MDS algorithm, from described overall distance matrix D G * GRenewal matrix D Ad G * GIn extract eigenvector E (k)=[x of each sample when representing with two-dimensional space k, y k];
Step 5 two or three: the data center that obtains all samples
Figure FSA00000176816400074
Step 5 two or four: obtain each sample and the described E of data center CtDistance
Figure FSA00000176816400075
And obtain d Med(2)=median (d k(2));
Step 5 two or five: judge d k(2) 〉=5*d Med(2), if judge that then described sample is remote sample.
CN2010102095582A 2010-06-25 2010-06-25 Method for displaying visualization organization of digital images Active CN101887590B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2010102095582A CN101887590B (en) 2010-06-25 2010-06-25 Method for displaying visualization organization of digital images

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2010102095582A CN101887590B (en) 2010-06-25 2010-06-25 Method for displaying visualization organization of digital images

Publications (2)

Publication Number Publication Date
CN101887590A true CN101887590A (en) 2010-11-17
CN101887590B CN101887590B (en) 2011-12-07

Family

ID=43073500

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2010102095582A Active CN101887590B (en) 2010-06-25 2010-06-25 Method for displaying visualization organization of digital images

Country Status (1)

Country Link
CN (1) CN101887590B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102607641A (en) * 2011-12-27 2012-07-25 哈尔滨工业大学 Cluster anomaly detection method of combustion gas turbine
CN102706563A (en) * 2012-06-14 2012-10-03 哈尔滨工业大学 Detection method for neighbor abnormities of gas turbine
CN103268327A (en) * 2013-04-28 2013-08-28 浙江工业大学 High-dimensional service data oriented hybrid visualization method
CN108346172A (en) * 2018-02-26 2018-07-31 中译语通科技股份有限公司 Multidimensional space data dot matrix VR methods of exhibiting and system, computer program
CN109886309A (en) * 2019-01-25 2019-06-14 成都浩天联讯信息技术有限公司 A method of digital picture identity is forged in identification
CN111259917A (en) * 2020-02-20 2020-06-09 西北工业大学 Image feature extraction method based on local neighbor component analysis

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1818974A (en) * 2006-03-08 2006-08-16 杭州电子科技大学 Multi-modality medical data three-dimensional visual method
WO2006095289A1 (en) * 2005-03-11 2006-09-14 Koninklijke Philips Electronics, N.V. System and method for volume rendering three-dimensional ultrasound perfusion images
CN101576913A (en) * 2009-06-12 2009-11-11 中国科学技术大学 Automatic clustering, visual and retrieval system for tongue picture based on self-organizing map neural network

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006095289A1 (en) * 2005-03-11 2006-09-14 Koninklijke Philips Electronics, N.V. System and method for volume rendering three-dimensional ultrasound perfusion images
CN1818974A (en) * 2006-03-08 2006-08-16 杭州电子科技大学 Multi-modality medical data three-dimensional visual method
CN101576913A (en) * 2009-06-12 2009-11-11 中国科学技术大学 Automatic clustering, visual and retrieval system for tongue picture based on self-organizing map neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
《Human Vision and Electronic Imaging: Models, Methods, and Applications(1990)》 19901231 Haim Levkowitz等 ICONOGRAPHIC INTEGRATED DISPLAYS OF MULTIPARAMETER SPATIAL DISTRIBUTIONS 345-355 1-7 第1249卷, 2 *
《航天医学与医学工程》 20011231 秦斌杰等 医学图像三维可视化 452-455 1-7 第14卷, 第6期 2 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102607641A (en) * 2011-12-27 2012-07-25 哈尔滨工业大学 Cluster anomaly detection method of combustion gas turbine
CN102706563A (en) * 2012-06-14 2012-10-03 哈尔滨工业大学 Detection method for neighbor abnormities of gas turbine
CN103268327A (en) * 2013-04-28 2013-08-28 浙江工业大学 High-dimensional service data oriented hybrid visualization method
CN103268327B (en) * 2013-04-28 2016-06-15 浙江工业大学 Mixing method for visualizing towards high-dimensional service data
CN108346172A (en) * 2018-02-26 2018-07-31 中译语通科技股份有限公司 Multidimensional space data dot matrix VR methods of exhibiting and system, computer program
CN108346172B (en) * 2018-02-26 2021-10-08 中译语通科技股份有限公司 Multidimensional space data dot matrix VR display method and system
CN109886309A (en) * 2019-01-25 2019-06-14 成都浩天联讯信息技术有限公司 A method of digital picture identity is forged in identification
CN111259917A (en) * 2020-02-20 2020-06-09 西北工业大学 Image feature extraction method based on local neighbor component analysis

Also Published As

Publication number Publication date
CN101887590B (en) 2011-12-07

Similar Documents

Publication Publication Date Title
CN107194872B (en) Remote sensed image super-resolution reconstruction method based on perception of content deep learning network
CN101887590B (en) Method for displaying visualization organization of digital images
CN100550037C (en) Utilize and improve Hausdorff apart from the method for extracting the identification human ear characteristic
Wang et al. Color image segmentation using pixel wise support vector machine classification
WO2018023734A1 (en) Significance testing method for 3d image
CN103177446B (en) Based on the accurate extracting method of display foreground of neighborhood and non-neighborhood smoothing prior
CN104008538B (en) Based on single image super-resolution method
CN102509099B (en) Detection method for image salient region
Zhang et al. Dual-mode artificially-intelligent diagnosis of breast tumours in shear-wave elastography and B-mode ultrasound using deep polynomial networks
CN102509104B (en) Confidence map-based method for distinguishing and detecting virtual object of augmented reality scene
CN103116763A (en) Vivo-face detection method based on HSV (hue, saturation, value) color space statistical characteristics
CN101944178B (en) Significant region extraction method for intelligent monitoring
CN101276417A (en) Method for filtering internet cartoon medium rubbish information based on content
CN103714349A (en) Image recognition method based on color and texture features
CN105678261B (en) Based on the direct-push Method of Data with Adding Windows for having supervision figure
CN104598871A (en) Correlation regression based face age calculating method
Liu et al. Road centerlines extraction from high resolution images based on an improved directional segmentation and road probability
CN105488475B (en) Method for detecting human face in mobile phone
CN110705565A (en) Lymph node tumor region identification method and device
CN103390170A (en) Surface feature type texture classification method based on multispectral remote sensing image texture elements
CN115861409B (en) Soybean leaf area measuring and calculating method, system, computer equipment and storage medium
CN103778430B (en) Rapid face detection method based on combination between skin color segmentation and AdaBoost
CN109146925A (en) Conspicuousness object detection method under a kind of dynamic scene
CN114999637B (en) Pathological image diagnosis method and system based on multi-angle coding and embedded mutual learning
CN107256421B (en) Method for rapidly counting rice and wheat seeds

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C53 Correction of patent for invention or patent application
CB03 Change of inventor or designer information

Inventor after: Yu Daren

Inventor after: He Huixin

Inventor after: Qing Shaowei

Inventor before: Yu Daren

Inventor before: He Huixin

COR Change of bibliographic data

Free format text: CORRECT: INVENTOR; FROM: YU DAREN HE HUIXIN TO: YU DAREN HE HUIXIN QING SHAOWEI

C14 Grant of patent or utility model
GR01 Patent grant