CN101887590B - Method for displaying visualization organization of digital images - Google Patents

Method for displaying visualization organization of digital images Download PDF

Info

Publication number
CN101887590B
CN101887590B CN 201010209558 CN201010209558A CN101887590B CN 101887590 B CN101887590 B CN 101887590B CN 201010209558 CN201010209558 CN 201010209558 CN 201010209558 A CN201010209558 A CN 201010209558A CN 101887590 B CN101887590 B CN 101887590B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
digital image
step
η
pk
matrix
Prior art date
Application number
CN 201010209558
Other languages
Chinese (zh)
Other versions
CN101887590A (en )
Inventor
于达仁
卿绍伟
贺惠新
Original Assignee
哈尔滨工业大学
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
Grant date

Links

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 anextended 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

一种数字图像可视化组织展示的方法 A method of visualizing a digital image display tissue

技术领域 FIELD

[0001] 本发明涉及信息技术领域,具体涉及一种数字图像可视化组织展示的方法。 [0001] The present invention relates to the field of information technology, particularly to a method of digital image visualizing tissue impressions. 背景技术 Background technique

[0002] 目前对于图像可视化的特征抽取方式,考虑细节和分布的方法通常需要很大的计算量,而直方图很容易就丢弃了图像的空间信息,而简单的分段计算的像素总量值也直接忽略了计算时像素量值的分布特性。 [0002] At present, for image feature extraction visualized manner, and considering the details of the distribution method generally requires a large amount of computation, the histogram it is easy to discard the spatial information of the image, while the total value of the pixel simple segmented calculated also directly overlooked the distribution characteristics of the pixel values ​​of the calculation.

[0003] 对于图像可视化的组织方式,目前一般都基于预定义的类别,各个图像归档到不同的类别中,这样的归类相对生硬,对类别判定的阈值依赖性很大。 [0003] For visual image organization, general are based on predefined categories, each image filing into different categories, such classification is relatively stiff, large categories of dependency determination threshold value.

发明内容 SUMMARY

[0004] 为了解决现有的图像可视化方法计算量大的问题,本发明提供了一种数字图像可视化组织展示的方法。 [0004] In order to solve the problem of large conventional method for calculating the visual image, the present invention provides a method of visualizing a digital image display tissue.

[0005] 本发明的一种数字图像可视化组织展示的方法,它包括以下步骤: [0005] A digital image of the present invention show tissue visualization method, comprising the steps of:

[0006] 步骤一:获取G张未经标记的原始数字图像Pk(m,η),1彡k彡G ; [0006] Step a: acquiring a digital image of the original sheets G Pk (m, η) unlabeled, 1 San San G k;

[0007] 步骤二:提取所述的每张原始数字图像Pk(m,η)的原始特征量FSk(Z)及其扩展图像的扩展特征量冊,(ζ),获取所述每张原始数字图像Pk (m,η)的图像特征量PFk = Fk(I),所述Fk(I)表示第k张原始数字图像Pk (m,η)的原始特征量FSk (ζ)及扩展特征量FEk (ζ)的合并特征量,并用所获取的图像特征量PFk表示原始数字图像Pk (m,n),其中,ζ = 1,2,3..., 24; [0007] Step Two: Original extract the original digital image Pk wherein each said (m, η) amounts FSk (Z) and its extension extended register image feature amount, (ζ), each said acquired raw digital image Pk (m, η) of the image feature amount PFk = Fk (I), the Fk (I) represents a k-th sheet of the original digital image Pk (m, η) of the primitive feature amount FSk (ζ) and extended feature amount FEk the image feature amount ([zeta]) were combined feature amount, and represented by the acquired raw digital image PFk Pk (m, n), where, ζ = 1,2,3 ..., 24;

[0008] 步骤三:用ρ范数距离来度量第b张原始数字图像与第g张原始数字图像之间的 [0008] Step three: by the number of distance ρ between the norm of the measured sheet b of the original digital image and the second sheet of the original digital image g

相似距离伪对,并建立全局 Similar pseudo-distance and establish global

距离矩阵 Distance Matrix

Figure CN101887590BD00071

,其中,A为所述 Wherein, A is a

两个图像特征值PFb和PFg的距离计算权值,取&为5的第1个维度量,5为48维度的特征权向量; Two image distance calculating feature values ​​and weights PFb of PFg, & taken as a first dimension an amount of 5, 5 to 48 weight vector feature dimensions;

[0009] 步骤四:将G张原始数字图像作为G张样本,并从所述G张样本中获取每个样本所归属的聚类中心样本允=a^maxWt/)+"(*,/》唭中,a(k, f)为所述有效性矩阵Acxc 中的第k行元素,r (k,f)为所述反馈矩阵中的第k行元素; [0009] Step Four: the digital image as the original sheets G G sheets sample, and each sample acquired belongs cluster centers allow = a ^ maxWt sample from the sample sheets G /) + "(*, /" in Qi, a (k, f) is a matrix Acxc the validity of the k-th row element, r (k, f) is the k-th row of the feedback element of the matrix;

[0010] 步骤五:依据全局距离矩阵Dexe和所获取的每个样本所归属的聚类中心样本fck获取每个样本的特征矢量E (k),进而从所有样本中检测偏远样本; [0010] Step Five: obtaining a sample based on the cluster center distance matrix Dexe fck global and each sample acquired belongs characteristics of each sample vector E (k), and thus detected in all samples from the remote sample;

[0011] 步骤六:将G张样本变换为G张缩小图像,并生成所述每个缩小图像的角度偏转 [0011] Step Six: the sample sheets G G converted into reduced images, and generates the reduced image for each angular deflection

量却, Amount but,

Figure CN101887590BD00081

为[0,1]上的随机值,以所获得的角度偏转量Ap (k)调 A random value [0, 1], to obtain the deflection angle Ap (k) modulation

整所述缩小图像,并使所述缩小图像在目标展示空间中正对着用户观察面以特征矢量E(k) 展示,以完成数字图像可视化组织展示。 The entire reduced image, and the reduced image in the target presentation space viewed by the user against CKS plane to feature vector E (k) Display, in order to complete the display of digital images visualizing tissue.

[0012] 本发明的有益效果:本发明提供了一种基于视觉感受的快速的自适应图像聚类进而实现可视化组织展示的方法,本发明在对图片组织归类时,将数字图像转化到HIS空间进行处理,使得图像可视化更符合人类的视觉感知特性,在图像可视化组织展示过程中,利用图像特征值表示图像进而排除偏远样本,使得图像可视化过程中的计算量小。 [0012] Advantageous Effects of Invention: The present invention provides a method based on Fast Adaptive Clustering visual perception of the image and thus to visualize the organization of the display, when the present invention is to picture tissue classification, conversion to digital image HIS treatment space, so that the image more consistent with the visual characteristics of human visual perception, the image display process of tissue visualized using image feature values ​​representing an image, ruling out the remote sample, is calculated such that the image of the visualization process in a small amount.

附图说明 BRIEF DESCRIPTION

[0013] 图1是本发明的一种数字图像可视化组织展示的方法的流程图。 [0013] FIG. 1 is a flow chart showing the organization of a method according to the present invention, a digital image visualization.

具体实施方式 detailed description

[0014] 具体实施方式一:根据说明书附图1具体说明本实施方式,本实施方式所述的一种数字图像可视化组织展示的方法,它包括以下步骤: [0014] In a particular embodiment: According to one embodiment of the present embodiment described in detail accompanying drawings, the present embodiment is a digital embodiment of the image display tissue visualization method, comprising the steps of:

[0015] 步骤一:获取G张未经标记的原始数字图像Pk(m,η),1彡k彡G ; [0015] Step a: acquiring a digital image of the original sheets G Pk (m, η) unlabeled, 1 San San G k;

[0016] 步骤二:提取所述的每张原始数字图像Pk(m,η)的原始特征量FSk(Z)及其扩展图像的扩展特征量冊,(ζ),获取所述每张原始数字图像Pk (m,η)的图像特征量PFk = Fk(I),所述Fk(I)表示第k张原始数字图像Pk (m,η)的原始特征量FSk (ζ)及扩展特征量FEk (ζ)的合并特征量,并用所获取的图像特征量PFk表示原始数字图像Pk (m,n),其中,ζ = 1,2,3..., 24; [0016] Step Two: Original extract the original digital image Pk wherein each said (m, η) amounts FSk (Z) and its extension extended register image feature amount, (ζ), each said acquired raw digital image Pk (m, η) of the image feature amount PFk = Fk (I), the Fk (I) represents a k-th sheet of the original digital image Pk (m, η) of the primitive feature amount FSk (ζ) and extended feature amount FEk the image feature amount ([zeta]) were combined feature amount, and represented by the acquired raw digital image PFk Pk (m, n), where, ζ = 1,2,3 ..., 24;

[0017] 步骤三:用ρ范数距离来度量第b张原始数字图像与第g张原始数字图像之间的 [0017] Step three: by the number of distance ρ between the norm of the measured sheet b of the original digital image and the second sheet of the original digital image g

相似距离 Similar distance

Figure CN101887590BD00082

,并建立全局 And the establishment of a global

距离矩阵 Distance Matrix

Figure CN101887590BD00083

其中,A为所述 Wherein, A is a

Figure CN101887590BD00084

两个图像特征值PFb和PFg的距离计算权值,取&为5的第1个维度量,5为48维度的特征权向量; Two image distance calculating feature values ​​and weights PFb of PFg, & taken as a first dimension an amount of 5, 5 to 48 weight vector feature dimensions;

[0018] 步骤四:将G张原始数字图像作为G张样本,并从所述G张样本中获取每个样本所归属的聚类中心样本允=a^maxWt/) +小,/)}唭中,a(k, f)为所述有效性矩阵Acxc 中的第k行元素,r (k,f)为所述反馈矩阵中的第k行元素; [0018] Step Four: the digital image as the original sheets G G sample sheets, and allowed to acquire the cluster center for each sample belongs sample from the sample sheets G = a ^ maxWt /) + small, /)} Qi in, a (k, f) is a matrix Acxc the validity of the k-th row element, r (k, f) is the k-th row of the feedback element of the matrix;

[0019] 步骤五:依据全局距离矩阵Dexe和所获取的每个样本所归属的聚类中心样本fck 获取每个样本的特征矢量E (k),进而从所有样本中检测偏远样本;[0020] 步骤六:将G张样本变换为G张缩小图像,并生成所述每个缩小图像的角度偏转 [0019] Step Five: obtaining a sample based on the center of the cluster belongs to the global fck Dexe distance matrix for each sample and each sample acquired feature vector E (k), and thus detected in all samples from the remote sample; [0020] step six: Zhang G G sample is converted into reduced images, and generates the reduced image for each angular deflection

量却 But the amount of

Figure CN101887590BD00091

,rand(*)为[0,1]上的随机值,以所获得的角度偏转量Ap(k)调 , Rand (*) is a random value [0, 1], to the angle of deflection of the obtained Ap (k) modulation

整所述缩小图像,并使所述缩小图像在目标展示空间中正对着用户观察面以特征矢量E(k) 展示,以完成数字图像可视化组织展示。 The entire reduced image, and the reduced image in the target presentation space viewed by the user against CKS plane to feature vector E (k) Display, in order to complete the display of digital images visualizing tissue.

[0021] 具体实施方式二:本实施方式是对具体实施方式一的进一步说明,具体实施方式一在步骤二中,提取所述的每张原始数字图像Pk(m,η)的原始特征量FSk(Z)的具体方法为: [0021] DETAILED Embodiment 2: The present embodiment is further described in a specific embodiment, a particular embodiment, in step two, the original features of each of the original digital image Pk extracting said (m, η) amounts FSk (Z) is a specific method:

[0022] 步骤二一一:将原始数字图像Pk(m,η)以RGB颜色坐标作为存储,即Pk(m,η)= (Pr, Pgj I\)m,n,其中,ι ^ m ^ Mk, 1 ^ η ^ Nk,该原始数字图像的尺寸为MkXNk,Mk和Nk均大于或等于1 ; [0022] Step two hundred eleven: the original digital image Pk (m, η) stored as an RGB color coordinates, i.e. Pk (m, η) = (Pr, Pgj I \) m, n, where, ι ^ m ^ Mk, 1 ^ η ^ Nk, the size of the original digital image is MkXNk, Mk and Nk are greater than or equal to 1;

[0023] 步骤二一二:将原始数字图像Pk(m,η) = (PE, Pe,n在HIS空间表示为Pk(m, η) = (Im,n,Hm,n,Sm,η),进而计算该原始数字图像Pk(m,η)在HIS空间的每个空间中的像素 [0023] Step two hundred and twelve: the original digital image Pk (m, η) = (PE, Pe, n as represented in space HIS Pk (m, η) = (Im, n, Hm, n, Sm, η) , then calculate the original digital image Pk (m, η) for each pixel in the space on HIS

概率直方图 Probability histograms

Figure CN101887590BD00092

t = 1,2,... ,mn, θ = I, H, S,其中,S0 (qt)表示在 t = 1,2, ..., mn, θ = I, H, S, wherein, S0 (qt) represents

θ空间中原始数字图像Pk(m,η)的像素值为qt的像素个数, Θ number of pixels in the original digital image space Pk (m, η) of the pixel values ​​qt,

Figure CN101887590BD00093

为在θ空间中原 Central space as θ

始数字图像Pk(m,η)的所有像素个数; All the number of pixels beginning digital image Pk (m, η) of;

[0024] 步骤二一三:在HIS空间的每个空间中,根据原始数字图像Pk(m,η)的所有像素值qt组成的值域范围,将该值域范围平均划分为8个区段,进而获取原始数字图像Pk(m,η)在每个空间的特征量 [0024] Step two hundred thirteen: in each spatial HIS space, according to all the pixels of the original digital image Pk (m, η) value range value qt composition, the average value range is divided into eight segments , then obtain the original digital image Pk (m, η) of each feature amount space

Figure CN101887590BD00094

,W= 1,2,...,8, qfflin = min(qt), Qfflax = , W = 1,2, ..., 8, qfflin = min (qt), Qfflax =

max(qt),dar = (qmax_qmin)/8,并将获取的原始数字图像Pk(m,η)在I空间的特征量用H1(W) 表示、在H空间的特征量用Hh(W)表示,在S空间的特征量用Hs(W)表示; max (qt), dar = (qmax_qmin) / 8, the original digital image Pk (m, η) is shown in the acquired feature quantity with a space I H1 (W), in the feature quantity space H with Hh (W) He represents, expressed by Hs (W) in the feature amount S of the space;

[0025] 步骤二一四:根据所获取的原始数字图像Pk(m,η)在HIS空间的每个空间中的8 个特征量获取原始数字图像Pk(m,η)在HIS空间的M个原始特征量FSk(Z)。 [0025] Step two hundred fourteen: 8 in each spatial feature amount space HIS acquiring raw digital image Pk (m, η) M th HIS space in the raw digital image Pk (m, η) obtained primitive feature amount FSk (Z).

[0026] 具体实施方式三:本实施方式是对具体实施方式二的进一步说明,具体实施方式二中在步骤二一二中,将原始数字图像Pk(m,n)=扎乂,! [0026] DETAILED Embodiment 3: The present embodiment is further described two specific embodiments, specific embodiments of two hundred and twelve in step II, the original digital image Pk (m, n) = qe bar,! ^^在! ^^ in! ^空间表示为? ^ Space is represented as? ^!!!,!!) =(Im,n,Hm,n,Sm, n)的具体方法为: ^ !!!, !!) = (Im, n, Hm, n, Sm, n) is a specific method:

[0027] 首先计算 [0027] First Calculation

Figure CN101887590BD00095

进而获取Hm,„和^i1,n, And then obtain Hm, "and ^ i1, n,

[0028]当 Pk = min (PE, PG, Pb)时,则 [0028] When Pk = min (PE, PG, Pb), the

[0029] [0029]

Figure CN101887590BD00096

[0030]当Pg = min (PE, PG, Pb)时,则 [0030] When Pg = min (PE, PG, Pb), the

Figure CN101887590BD00101

[0032]当I3b = min (PE, P,, PB)时,则 [0032] When I3b = min (PE, P ,, PB), then

[0033] [0033]

Figure CN101887590BD00102

[0034] 最终完成将原始数字图像Pk (m, n) = (PE, Pe,PB)m,n在HIS空间表示为Pk(m, n)= [0034] finalize the original digital image Pk (m, n) = (PE, Pe, PB) m, n as represented in space HIS Pk (m, n) =

(工111,n,Hm,η,Sni7 η) O (Engineering 111, n, Hm, η, Sni7 η) O

[0035] 具体实施方式四:本实施方式是对具体实施方式一、二或三的进一步说明,具体实施方式一、二或三中,在步骤二中,提取所述每张原始数字图像I\(m,η)的扩展图像的扩展特征量FEk (ζ)的具体方法为: [0035] DETAILED DESCRIPTION 4: The embodiment further illustrate specific embodiments one, two or three, a particular embodiment, the two or three, in step two, each of said original digital image to extract the I \ DETAILED method extends the feature amount FEk (m, η) extended image ([zeta]) is:

[0036]步骤: [0036] Step:

:将原始数字图像Pk(m,η)以RGB颜色坐标作为存储,即Pk(m,η)= : The original digital image Pk (m, η) stored as an RGB color coordinates, i.e. Pk (m, η) =

(PE, Pg, I\)m,n,其中,1 ^ m ^ Mk, 1 ^ η ^ Nk,原始数字图像的尺寸为MkXNk,Mk和Nk均大于或等于1,并将所述原始数字图像Pk(m,η)扩展为扩展图像A(u,ν), (PE, Pg, I \) m, n, where, 1 ^ m ^ Mk, 1 ^ η ^ Nk, the size of the original digital image is MkXNk, Mk and Nk are greater than or equal to 1, and the original digital image Pk (m, η) for the extension extended image A (u, ν),

[0037] [0037]

Figure CN101887590BD00103

[0038] 步骤二二二:对所述原始数字图像Pk(m,n)的扩展图像Qk (u,ν)作平滑处理,获得所述扩展图像Qk(u,ν)的平滑结果 [0038] Step two hundred twenty-two: the original digital image Pk (m, n) of the extended image Qk (u, ν) for smoothing process, smoothing the result of the extended image obtained Qk (u, ν) of

[0039] [0039]

Figure CN101887590BD00104

[0040] 步骤二二三:获取所述原始数字图像Pk(m,η)的扩展图像Qk(u,ν)的平滑结果 [0040] Step two hundred twenty-three: acquiring extended image Qk (u, ν) of the original digital image Pk (m, η) of the smoothed results

Yk(m, η)在HIS空间的每个空间中的像素概率直方图' Yk (m, η) of each spatial pixel probability space histogram HIS '

Figure CN101887590BD00105

' '

其中,S0〒s(q' t)表示在θ空间中平滑结果Yk (m,n)的像素值为q' t的像素个数, Wherein S0〒s (q 't) represents the spatial smoothing θ results Yk (m, n) is a pixel value q' t is the number of pixels,

Figure CN101887590BD00106

为在θ空间中所述平滑结果Yk(m,η)的所有像素个数;[0041] 步骤二二四:在HIS空间的每个空间中,根据所述平滑结果Yk(m,n)的所有像素值q' ^且成的值域范围,将所述所有像素值q' 值域范围平均划分为8个区段,进而获取所述平滑结果Yk(m,η)在每个空间的特征量 For all the number of pixels in said spatial smoothing θ results Yk (m, η); and [0041] Step two hundred twenty-four: HIS each space in the space, according to the smoothing result Yk (m, n) of all pixel values ​​of q 'and ^ into value range, all of the pixel values ​​q' value ranges equally divided into eight segments, and further wherein obtaining the smoothed result Yk (m, η) in each space the amount

[0042] [0042]

Figure CN101887590BD00111

[0043] [0043]

=min(q' t) , q' max = max(q' t) , d' ar = (q' max_q' min)/8,并将获取 = Min (q 't), q' max = max (q 't), d' ar = (q 'max_q' min) / 8, and acquires

的所述每一个平滑结果Yk(m,η)在I空间的特征量用Htsi(W)表示、在H空间的特征量用H¥sh(w)表示,在S空间的特征量用H 平滑s (w) ; Each of the smoothing result Yk (m, η) represented by Htsi (W) in the feature quantity space I, in the feature quantity space H ¥ sh (w) is represented by H, the feature amount space S with a smooth H s (w);

[0044] 步骤二二五:根据步骤二二四所获取的平滑结果Yk(m,η)在HIS空间的每个空间中的8个特征量获取原始数字图像I\(m,η)在HIS空间的M个扩展特征量FEk(Z)。 [0044] Step two hundred twenty-five: 8 in each spatial feature amount space HIS acquiring raw digital image I \ (m, η) according to the smoothed result Yk (m, η) Step two hundred twenty-four acquired HIS M extended spatial feature amount FEk (Z).

[0045] 具体实施方式五:本实施方式是对具体实施方式一至四中任意一个实施方式的进一步说明,具体实施方式一至四中在步骤四中,从所述G张样本中获取每个样本所归属的 [0045] DETAILED DESCRIPTION five: The present embodiment is further described in specific embodiments one to four of any one embodiment, one to four specific embodiments, the obtaining samples from each of the sample sheets G in the Step 4 ownership

聚类中心样本 Sample cluster center

[0046] 步骤四 [0046] Step Four

Figure CN101887590BD00112

的具体方法为 The specific method

获取第b张原始数字图像与第g张原始数字图像 Obtaining sheet b of the original digital image and the second sheet of the original digital image g

之间的相似性 Similarity between

Figure CN101887590BD00113

,并建立全局相似矩阵 And the establishment of global similarity matrix

Figure CN101887590BD00114

,其中,Sim(i, j)为所述全局相似矩阵Sexe中 Wherein, Sim (i, j) is the global similarity matrix in Sexe

第i行第j列的元素; I-th row j-th column element;

[0047] 步骤四二:建立反馈矩阵&xe,并令所述反馈矩阵的所有元素r (i,j)的初始值均为0 ; [0047] Step forty-two: establishing feedback matrix & xe, and so that all the elements of the feedback matrix of the r (i, j) of the initial values ​​are 0;

[0048] 步骤四三:建立有效性矩阵Aexe,并令所述有效性矩阵Aexe的所有元素a(i,j)的初始值均为0 ; [0048] Step forty-three: establish the validity of the matrix Aexe, and to make all the elements of the matrix Aexe validity of a (i, j) of the initial values ​​are 0;

[0049] 步骤四四:对反馈矩阵和有效性矩阵Aexe进行LT次迭代,进而获取LT次迭代更新后的反馈矩阵仏㈣和效性矩阵Aexe,所述每次迭代的具体过程为: [0049] Step four four: the effectiveness of the feedback matrix and matrix Aexe LT iterations performed, thereby obtaining feedback matrix Fo and (iv) the effectiveness of LT matrix Aexe update iterations, each iteration of the specific process is:

[0050] 步骤四四一:将建立的反馈矩阵仏㈣中的所有元素r(i,j)更新为r(i. j)* = Xrtmp(i,j) + (l"A)rold(i, j),其中,λ为更新系数,rold(i, j)为上一次迭代更新获得的 [0050] Step four hundred forty-one: all the elements of r (i, j) feedback matrix Fo iv established is updated to r (i j.) * = Xrtmp (i, j) + (l "A) rold (i , j), where, λ is the update coefficient, rold (i, j) is obtained in the last iteration updates

Figure CN101887590BD00121

[0051] atmp(i, j')的初始值为0,然后执行步骤四四二; [0051] atmp (i, j ') of the initial value is 0, then step four hundred forty-two;

[0052] 步骤四四二:将建立的有效性矩阵Aexe中的所有元素a(i,j)更新为a(i. j)* = λ atmp (i,j) + (1- λ ) aold (i,j),其中,(i,j)为上一次迭代更新获得的a (i,j), [0052] Step four hundred forty-two: The effectiveness of all of the elements in the matrix Aexe establish a (i, j) is updated to a (. I j) * = λ atmp (i, j) + (1- λ) aold ( i, j), where, (i, j) is obtained in the last iteration updates a (i, j),

[0053] [0053]

Figure CN101887590BD00122

[0054] 步骤四五:根据所述的进行LT次更新后的反馈矩阵和效性矩阵Aexe获得每个样本所归属的聚类中心样本 [0054] Step forty-five: obtaining a sample belongs to each cluster center sample according to the said feedback matrix update is performed and the effectiveness of LT matrix Aexe

Figure CN101887590BD00123

[0055] 具体实施方式六:本实施方式是对具体实施方式一至五中任意一个实施方式的进一步说明,具体实施方式一至五中在步骤五中,依据全局距离矩阵Dexe和所获取的每个样本所归属的聚类中心样本fck获取每个样本的特征矢量E (k),进而从所有样本中检测偏远样本的具体方法为: [0055] DETAILED DESCRIPTION VI: The present embodiment is further described in specific embodiments to any one of first to fifth embodiment, the first to fifth specific embodiment in step 5, based on the overall distance matrix Dexe and each sample acquired sample belongs to the cluster center for each sample acquired fck feature vector E (k), and thus the specific method of detecting a sample isolated from all samples as follows:

[0056] 步骤五——:将所述全局距离矩阵Dexe中各元素Dist(PFb,PFg)更新为 [0056] Step Five: - the global distance matrix Dexe each element Dist (PFb, PFg) is updated to

Figure CN101887590BD00124

进而获取所述全局距离矩阵Dexe的更新矩阵DAdexe ; Further acquiring the global update matrices DAdexe Dexe the distance matrix;

[0057] 步骤五一二:利用多维尺度变换MDS算法,从所述全局距离矩阵Dexe的更新矩阵DAdexe中提取出每个样本在三维空间表示时的特征矢量E (k) = [xk,yk,zk]; [0057] Step five hundred twelve: MDS multidimensional scaling algorithm extracting from the global update matrices from the matrix Dexe DAdexe in the feature vector E (k) = [xk, yk represents the time of each sample in three-dimensional space, zk];

Ect {x) = median{xx ,x2,...,xG) Ect {x) = median {xx, x2, ..., xG)

[0058] 步骤五一三:获取所有样本的数据中心 [0058] Step five hundred thirteen: acquiring data centers all samples

Figure CN101887590BD00125

[0059] 步骤五一四:获取每个样本与所述数据中心E。 [0059] Step five hundred fourteen: obtaining a sample of each of the data center E. t的距离 The distance t

Figure CN101887590BD00126

,并获取dmed (3) = median (dk (3)); And acquires dmed (3) = median (dk (3));

[0060] 步骤五一五:判断dk(3)≥5*dmed(3),如果是,则判定所述样本为偏远样本。 [0060] Step five hundred fifteen: Analyzing dk (3) ≥5 * dmed (3), and if so, determining that the remote sample is a sample.

[0061] 本实施方式中,在可视化组织展示时,也可加入偏远样本。 [0061] In the present embodiment, when the visual display tissue, may also be added to the remote sample.

[0062] 具体实施方式七:本实施方式是对具体实施方式一至五中任意一个实施方式的进一步说明,具体实施方式一至五中在步骤五中,依据全局距离矩阵Dexe和所获取的每个样本所归属的聚类中心样本fck获取每个样本的特征矢量E (k),进而从所有样本中检测偏远样本的具体方法为: [0062] Seventh Embodiment: This embodiment is further described in specific embodiments to any one of first to fifth embodiment, the first to fifth specific embodiment in step 5, based on the overall distance matrix Dexe and each sample acquired sample belongs to the cluster center for each sample acquired fck feature vector E (k), and thus the specific method of detecting a sample isolated from all samples as follows:

[0063] 步骤五二一:将所述全局距离矩阵Dexe中各元素Dist (PFb, PFg)更新为 [0063] Step five hundred twenty-one: the global distance matrix Dexe each element Dist (PFb, PFg) is updated to

Figure CN101887590BD00131

进而获取所述全局距离矩阵Dexe的更新[Dist\PFb,PFg), fcb^fcg Further for acquiring the update of the global distance matrix Dexe [Dist \ PFb, PFg), fcb ^ fcg

矩阵DAdexe ; Matrix DAdexe;

[0064] 步骤五二二:利用多维尺度变换MDS算法,从所述全局距离矩阵Dexe的更新矩阵DAdexe中提取出每个样本用二维空间表示时的特征矢量E (k) = [xk,yk]; [0064] Step five hundred twenty-two: MDS multidimensional scaling algorithm extracts a feature vector E (k) = [xk when each sample is represented by a two-dimensional space from the global update matrices DAdexe the distance matrix Dexe, yk ];

[0065] 步骤五二三:获取所有样本的数据中心 [0065] Step five hundred twenty-three: acquiring data centers all samples

Figure CN101887590BD00132

[0066] 步骤五二四:获取每个样本与所述数据中心E。 [0066] Step five hundred twenty-four: obtaining a sample of each of the data center E. t的距离 The distance t

Figure CN101887590BD00133

,并获取dmed (2) = median (dk (2)); And acquires dmed (2) = median (dk (2));

[0067] 步骤五二五:判断≥5*CLJ2),如果是,则判定所述样本为偏远样本。 [0067] Step five hundred twenty-five: Analyzing ≥5 * CLJ2), and if so, determining that the remote sample is a sample.

[0068] 本实施方式中,在可视化组织展示时,也可加入偏远样本。 [0068] In the present embodiment, when the visual display tissue, may also be added to the remote sample.

Claims (1)

  1. 1. ー种数字图像可视化组织展示的方法,其特征在于它包括以下步骤: 步骤ー:获取G张未经标记的原始数字图像Pk(m,n),1彡k彡G ; 步骤ニ:提取所述的每张原始数字图像Pk (m,n)的原始特征量FSk (z)及其扩展图像的扩展特征量FEk (z),获取所述每张原始数字图像Pk (m,n)的图像特征量PFk = Fk (1),所述Fk(I)表示第k张原始数字图像Pk (m,n)的原始特征量FSk (z)及扩展特征量rak(Z)的合并特征量,并用所获取的图像特征量PFk表示原始数字图像Pk (m,n),其中,z = 1,2,3...,24; 步骤三:用P范数距离来度量第b张原始数字图像与第g张原始数字图像之间的相似距离,并建立全局距离 A method ー kind digital display image visualizing tissue, characterized in that it comprises the following steps: Step ー: acquiring a digital image of the original sheets G Pk (m, n) unlabeled, 1 San San G k; ni steps: Extraction each of the original digital image Pk (m, n) of the primitive feature amount FSk (z) and its extension extended FEk image feature amount (z), each of the acquired raw digital image Pk (m, n) of the image feature amount PFk = Fk (1), said Fk (I) represents a k-th sheet of the original digital image Pk (m, n) of the primitive feature amount FSk (z) and extended feature amount rak (Z) combined feature quantity, and said original digital image Pk (m, n), where, z = 1,2,3 ..., 24 with the image feature amount acquired PFk; step three: P norm used to measure the distance b of the original digital image sheet similar to the distance between the g Zhang original digital image, and the establishment of a global distance
    Figure CN101887590BC00021
    ,其中, ,among them,
    Figure CN101887590BC00022
    为所述两个图像特征量PFb和PFg的距离计算权值,取み为茂的第1个维度量,5为48维度的特征权向量;步骤四:将G张原始数字图像作为G张样本,并从所述G张样本中获取每个样本所归属的聚类中心样本 To calculate the distance of the two weights image feature amount and PFb PFg, whichever first dimension Mi Mao amount, 5 to 48 weight vector feature dimensions; Step four: sheets G as the original digital image sample sheets G and obtaining the samples of each cluster center samples from the home sample sheets G
    Figure CN101887590BC00023
    ,a(k,f)为有效性矩阵Aexe中的第k行元素,r(k,f)为反馈矩阵も㈣中的第k行元素;步骤五:依据全局距离矩阵Dtjxtj和所获取的每个样本所归属的聚类中心样本fck获取每个样本的特征矢量E (k),进而从所有样本中检测偏远样本;步骤六:将G张样本变换为G张縮小图像,并生成所述每个縮小图像的角度偏转量 , A (k, f) is the k-th row element of the matrix Aexe effectiveness in, r (k, f) is the feedback element of the k-th row in the matrix mo iv; Step Five: Global distance matrix based on the acquired Dtjxtj and each sample belongs to the cluster center for each sample acquired fck sample feature vector E (k), and thus detected in all samples from the remote sample; step six: Zhang G G sample is converted into reduced images, and generating each of a narrow angle of deflection of the image
    Figure CN101887590BC00024
    上的随机值,以所获得的角度偏转量Ap(k)调整所述縮小图像,并使所述縮小图像在目标展示空间中正对着用户观察面以特征矢量E(k) 展示,以完成数字图像可视化组织展示;在步骤四中,从所述G张样本中获取每个样本所归属的聚类中心样本 On the random value, the amount of deflection angle obtained Ap (k) adjusting said reduced image, and the reduced image to the viewing surface facing the user feature vector E (k) appear on the target CKS display space to complete the digital the image display tissue visualization; in the step 4, obtaining samples of each cluster center samples from the home sample sheets G
    Figure CN101887590BC00025
    的具体方法为:步骤四一:获取第b张原始数字图像与第g张原始数字图像之间的相似性没 The specific method is: Step forty-one: Get no similarity between the first sheet b of the original digital image and the second sheet of the original digital image g
    Figure CN101887590BC00026
    并建立全局相似矩阵 And the establishment of global similarity matrix
    Figure CN101887590BC00031
    ,其中,Sim(i, j)为所述全局相似矩阵SGxG中第i行第j列的元素;步骤四二:建立反馈矩阵RGxG。 Wherein, Sim (i, j) is the global similarity SGxG matrix element in row i and column j; Step forty-two: establishing feedback matrix RGxG. 并令所述反馈矩阵RGxG的所有元素r(i,j)的初始值均为0;步骤四三:建立有效性矩阵AGxG,并令所述有效性矩阵AGxG的所有元素a(i,j)的初始值均为0 ;步骤四四:对反馈矩阵;和有效性矩阵AGxG进行LT次迭代,进而获取LT次迭代更新后的反馈矩阵仏㈣和有效性矩阵AGxG,所述每次迭代的具体过程为:步骤四四一:将建立的反馈矩阵仏㈣中的所有元素r(i,j)更新为r(i. j)*= λ rtmp (i, j) + (l"A)rold(i, j),其中,λ为更新系数,rold(i, j)为上一次迭代更新获得的r(i,j), And to make all the elements of the feedback matrix RGxG of r (i, j) of the initial values ​​are 0; forty-three steps: establish the validity of the matrix AGxG, and to make all the elements of the matrix AGxG validity of a (i, j) the initial values ​​are 0; step four four: feedback matrix; and effectiveness of LT matrix AGxG iterations performed, thereby obtaining feedback matrix Fo and (iv) the effectiveness of LT matrix AGxG updating iterations, each iteration of the specific process: step four hundred forty-one: all the elements of r (i, j) feedback matrix Fo iv established is updated to r (i j.) * = λ rtmp (i, j) + (l "a) rold ( i, j), where, [lambda] is the update coefficient, rold (i, j) to update the previous iteration r (i, j) obtained,
    Figure CN101887590BC00032
    atmp(i,j')的初始值为0,然后执行步骤四四二;步骤四四二:将建立的有效性矩阵AGxG中的所有元素a(i,j)更新为a(i. j)* =atmp(i, j) + (lA)aold(i, j),其中,aold(i,j)为上一次迭代更新获得的a (i,j), Initial atmp (i, j ') value is 0, then step four hundred forty-two; Step four hundred forty-two: all elements to establish the validity of the matrix AGxG in a (i, j) is updated to a (i j.) * = atmp (i, j) + (lA) aold (i, j), wherein, aold (i, j) is the previous iteration updates a (i, j) obtained,
    Figure CN101887590BC00033
    步骤四五:根据所述的进行LT次更新后的反馈矩阵和有效性矩阵AGxG获得每个样本所归属的聚类中心样本 Forty-five steps of: obtaining a sample belongs to each cluster center matrix and the effectiveness of the feedback sample matrix AGxG after the update is performed LT
    Figure CN101887590BC00034
    在步骤五中,依据全局距离矩阵DGxG和所获取的每个样本所归属的聚类中心样本fck 获取每个样本的特征矢量E (k),进而从所有样本中检测偏远样本的具体方法有两种,第一种方法为:步骤五——:将所述全局距离矩阵DGxG中各元素Dist (PFb,PFg)更新为 In step 5, based on the overall distance matrix DGxG and each sample acquired sample belongs fck cluster center for each sample acquired feature vector E (k), and thus detected in all samples from remote sample specific method has two species, the first method is: step five: - the global distance matrix DGxG each element Dist (PFb, PFg) is updated to
    Figure CN101887590BC00041
    ,进而获取所述全局距离矩阵Dexe的更新矩阵DAdexe ;步骤五一二:利用多维尺度变换MDS算法,从所述全局距离矩阵Dexe的更新矩阵DAdexe 中提取出每个样本在三维空间表示时的特征矢量E (k) = [xk, yk, zk];步骤五一三:获取所有样本的数据中心 Thus obtaining the global update matrices DAdexe Dexe the distance matrix; five hundred twelve steps: MDS multidimensional scaling algorithm extracting from the global update matrices from the matrix Dexe DAdexe when the characteristic of each sample is shown in three-dimensional space vector E (k) = [xk, yk, zk]; five hundred thirteen steps: acquiring data centers all samples
    Figure CN101887590BC00042
    步骤五一四:获取每个样本与所述数据中心E。 Step five hundred fourteen: obtaining a sample of each of the data center E. t的距离 The distance t
    Figure CN101887590BC00043
    ,并获取dmed ⑶=median (dk ⑶);步骤五一五:判断dk(3)彡5*dmed(3),如果是,则判定所述样本为偏远样本; 第二种方法为:步骤五二一:将所述全局距离矩阵Dexe中各元素Dist (PFb,PFg)更新为 And acquires dmed ⑶ = median (dk ⑶); five hundred fifteen steps: Analyzing dk (3) San 5 * dmed (3), and if so, determining that the remote sample is a sample; second method: Step Five twenty-one: the global distance matrix Dexe each element Dist (PFb, PFg) is updated to
    Figure CN101887590BC00044
    ,进而获取所述全局距离矩阵Dexe的更新矩阵DAdGXG ;步骤五二二:利用多维尺度变换MDS算法,从所述全局距离矩阵Dexe的更新矩阵DAdexe 中提取出每个样本用二维空间表示时的特征矢量E (k) = [xk,yk];步骤五二三:获取所有样本的数据中心 Thus obtaining the global update matrices DAdGXG from the matrix Dexe; Step five hundred twenty-two: MDS multidimensional scaling algorithm extracting from the global update matrices from the matrix Dexe DAdexe in each sample is represented by a two-dimensional space of feature vector E (k) = [xk, yk]; five hundred twenty-three steps of: acquiring data centers all samples
    Figure CN101887590BC00045
    步骤五二四:获取每个样本与所述数据中心E。 Step five hundred twenty-four: obtaining a sample of each of the data center E. t的距离 The distance t
    Figure CN101887590BC00046
    ,并获取dmed(2) = median (dk O));步骤五二五:判断彡5*CLJ2),如果是,则判定所述样本为偏远样本。 And acquires dmed (2) = median (dk O)); five hundred twenty-five steps: Analyzing San 5 * CLJ2), and if so, determining that the remote sample is a sample. 2.根据权利要求1所述的一种数字图像可视化组织展示的方法,其特征在于在步骤二中,提取所述的每张原始数字图像Pk(m,η)的原始特征量FSk(ζ)的具体方法为:步骤二一一:将原始数字图像Pk(m,n)以RGB颜色坐标作为存储,即Pk(m,n) = (PE,PG, I\)m,n,其中,1彡m彡Mk,1彡η彡Nk,该原始数字图像的尺寸为MkXNk,Mk和Nk均大于或等于1 ;步骤二一二:将原始数字图像Pk(m,n)=扎,? 2. A digital image of the display method of claim 1, visualization tissue, characterized in that in step two, each primitive features extracted according to the original digital image Pk (m, η) amounts FSk (ζ) the specific method is: two hundred and eleven steps: the original digital image Pk (m, n) is stored as an RGB color coordinates, i.e. Pk (m, n) = (PE, PG, I \) m, n, where 1 San San Mk m, 1 San η San Nk, the size of the original digital image is MkXNk, Mk and Nk are greater than or equal to 1; two hundred twelve steps: the original digital image Pk (m, n) = bar,? ^,⑴^在! ^, ⑴ ^ in! ^空间表示为? ^ Space is represented as? ^!!!,η)= (Iffl,n, Hm,n,^1J,进而计算该原始数字图像Pk(m,η)在HIS空间的每个空间中的像素概率直方图 ^ !!!, η) = (Iffl, n, Hm, n, ^ 1J, and then calculate the original digital image Pk (m, η) of each spatial pixel probability space histogram HIS
    Figure CN101887590BC00047
    其中,^⑷表示在Θ空间中原始数字图像Pk(m,n)的像素值为qt的像素个数 Wherein, ^ ⑷ space Θ represents the number of pixels in the original digital image Pk (m, n) is a pixel value qt
    Figure CN101887590BC00051
    为在θ空间中原始数字图像Pk (m, η)的所有像素个数;步骤二一三:在HIS空间的每个空间中,根据原始数字图像Pk(m,n)的所有像素值(^组成的值域范围,将该值域范围平均划分为8个区段,进而获取原始数字图像Pk(m,n)在每个空间的特征量 Is the number of all pixels of the original digital image Pk (m, η) in the θ space; two hundred thirteen step: in each spatial HIS space, according to all the pixel values ​​of the original digital image Pk (m, n) of (^ value range of the composition, the average value range is divided into eight segments, thereby obtaining the feature quantity of the original digital image Pk (m, n) in each space
    Figure CN101887590BC00052
    dar = (qmax_qmin)/8,并将获取的原始数字图像Pk(m,η)在I空间的特征量用H1(W)表示、在H空间的特征量用Hh(W)表示,在S空间的特征量用Hs(W)表示;步骤二一四:根据所获取的原始数字图像Pk(m,η)在HIS空间的每个空间中的8个特征量获取原始数字图像Pk(m,η)在HIS空间的M个原始特征量FSk(Z)。 dar = (qmax_qmin) / 8, the original digital image Pk (m, η) is represented by the acquired H1 (W) in the feature quantity space I, H represents the feature quantity space with Hh (W), the space S the feature amount is represented by Hs (W); two hundred fourteen steps: Get 8 HIS each spatial feature amount space according to the original digital image Pk (m, η) of the acquired raw digital image Pk (m, η ) the M original feature amount space HIS FSk (Z). 3.根据权利要求2所述的一种数字图像可视化组织展示的方法,其特征在于在步骤二一二中,将原始数字图像I\(m,n)=扎乂,! 3. The method of claim 2 a digital visual image display as claimed in claim tissue, characterized in that in step two hundred twelve, the original digital image I \ (m, n) = qe bar,! ^^在! ^^ in! ^空间表示为? ^ Space is represented as? !^,!!)= (Im,n,Hm, n,Sffl,n)的具体方法为:首先计算 ! ^, !!) = (Im, n, Hm, n, Sffl, n) is a specific method: first calculating
    Figure CN101887590BC00053
    进而获取Hm, η和^un,当PE = min(PE,PG,PB)时,则 Further obtaining Hm, η and ^ un, while when PE = min (PE, PG, PB), the
    Figure CN101887590BC00054
    当PG = min(PE,PG,PB)时,则 When PG = min (PE, PG, PB), the
    Figure CN101887590BC00055
    当Pb = min(PE,PG,PB)时,则 When Pb = min (PE, PG, PB), the
    Figure CN101887590BC00056
    最终完成将原始数字图像Pk(m,η)=扎,? The final completion of the original digital image Pk (m, η) = bar,? ^⑴^在!^空间表示为Pk(m,n) = (Iffl,η,Hm,η,Sni7》°4.根据权利要求1所述的一种数字图像可视化组织展示的方法,其特征在于在步骤二中,提取所述每张原始数字图像Pk (m,η)的扩展图像的扩展特征量FEk(Z)的具体方法为:步骤二二一:将原始数字图像Pk(m,n)以RGB颜色坐标作为存储,即Pk(m,n) = (PE,PG, I\)m,n,其中,1彡m彡Mk,1彡η彡Nk,原始数字图像的尺寸为MkXNk,Mk和Nk均大于或等于1,并将所述原始数字图像Pk(m,η)扩展为扩展图像A(u,ν), ^ ^ At ⑴! ^ Space is represented as Pk (m, n) = (Iffl, η, Hm, η, Sni7 "° 4. The method of claim 1 a digital visual image display as claimed in claim tissue, wherein in step two, the amount of extracted FEk (Z) wherein said extended specific method for each of the original digital image Pk (m, η) of the image is expanded: two two step one: the original digital image Pk (m, n) RGB color coordinates as storage, i.e. Pk (m, n) = (PE, PG, I \) m, n, where 1 San m San Mk, 1 San η San Nk of, the size of the original digital image is MkXNk, Mk and Nk equal to or greater than 1, and the original digital image Pk (m, η) for the extension extended image A (u, ν),
    Figure CN101887590BC00061
    ,0<w<M + l,0<v<iV + l步骤二二二:对所述原始数字图像Pk(m,η)的扩展图像(4(u,v)作平滑处理,获得所述扩展图像(4(u,ν)的平滑结果 , 0 <w <M + l, 0 <v <iV + l Step two hundred twenty-two: extended image of the original digital image Pk (m, η) of (4 (u, v) smoothing processing for obtaining the extended image (4 (u, ν) smoothing results
    Figure CN101887590BC00062
    步骤二二_获取所述原始数字图像Pk (m,η)的扩展图像Qk (u,ν)的平滑结果Yk(m,n)在HIS空间的每个空间中的像素概率直方图 Step twenty-two probability pixel of the original digital image acquiring _ Pk (m, η) of the extended image Qk (u, ν) smoothing results Yk (m, n) space in each spatial histogram HIS
    Figure CN101887590BC00063
    其中,Sf(q' t)表示在θ空间中平滑结果Yk(m,η)的像素值为q' t的像素个数 Wherein, Sf (q 't) represents the spatial smoothing θ results Yk (m, η) of a pixel value q' t is the number of pixels
    Figure CN101887590BC00064
    为在θ空间中所述平滑结果Yk(m,η)的所有像素个数;步骤二二四:在HIS空间的每个空间中,根据所述平滑结果Yk(m,n)的所有像素值q' t 组成的值域范围,将所述所有像素值q' t的值域范围平均划分为8个区段,进而获取所述平滑结果Yk(m,η)在每个空间的特征量 Θ is the number of all pixels in said spatial smoothing result Yk (m, η); and two hundred twenty-four steps: in each spatial HIS space, according to the result of smoothing all the pixel values ​​Yk (m, n) of 't composed value range, all of the pixel values ​​q' q t is the average value range is divided into eight segments, thereby obtaining a smoothed result Yk (m, η) of each feature amount space
    Figure CN101887590BC00065
    q' min = min(q' t),q' max = max(q' t), d' ar = (q' max_q' min)/8,并将获取的所述每一个平滑结果Yk(m,η)在I空间的特征量用Htsi(W)表示、在H空间的特征量用H^w H(w)表示,在S空间的特征量用Hws(W)表示;步骤二二五:根据步骤二二四所获取的平滑结果Yk(m,η)在HIS空间的每个空间中的8个特征量获取原始数字图像Pk(m,η)在HIS空间的M个扩展特征量FEk(Z)。 The q 'min = min (q' t), q 'max = max (q' t), d 'ar = (q' max_q 'min) / 8, each of the acquired smoothed result Yk (m, [eta]) represents the I feature quantity space by Htsi (W), is represented by H ^ w H (w) in the feature quantity space is H, represented by Hws (W) in the feature quantity space S; step two hundred twenty-five: the smoothing the result Yk (m, η) obtained in step two hundred twenty-four 8 HIS each spatial feature amount space in the original digital image acquired Pk (m, η) in the m spread on HIS feature amount FEk (Z ).
CN 201010209558 2010-06-25 2010-06-25 Method for displaying visualization organization of digital images CN101887590B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201010209558 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
CN 201010209558 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 true CN101887590B (en) 2011-12-07

Family

ID=43073500

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201010209558 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)

Families Citing this family (3)

* 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
CN103268327B (en) * 2013-04-28 2016-06-15 浙江工业大学 The method for mixing high-dimensional visualization of the data service

Citations (2)

* 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
CN101576913A (en) 2009-06-12 2009-11-11 中国科学技术大学;中国人民解放军第二军医大学 Automatic clustering, visual and retrieval system for tongue picture based on self-organizing map neural network

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080219526A1 (en) * 2005-03-11 2008-09-11 Koninklijke Philips Electronics, N.V. System and Method for Volume Rendering Three-Dimensional Ultrasound Perfusion Images

Patent Citations (2)

* 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
CN101576913A (en) 2009-06-12 2009-11-11 中国科学技术大学;中国人民解放军第二军医大学 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
Haim Levkowitz等.ICONOGRAPHIC INTEGRATED DISPLAYS OF MULTIPARAMETER SPATIAL DISTRIBUTIONS.《Human Vision and Electronic Imaging: Models, Methods, and Applications(1990)》.1990,第1249卷345-355.
秦斌杰等.医学图像三维可视化.《航天医学与医学工程》.2001,第14卷(第6期),452-455.

Also Published As

Publication number Publication date Type
CN101887590A (en) 2010-11-17 application

Similar Documents

Publication Publication Date Title
CN103077529A (en) System for analyzing plant leaf features based on image scanning
CN101667293A (en) Method for conducting high-precision and steady registration on diversified sensor remote sensing images
JP2004062605A (en) Scene identification method and device, and program
CN103116763A (en) Vivo-face detection method based on HSV (hue, saturation, value) color space statistical characteristics
CN102750705A (en) Optical remote sensing image change detection based on image fusion
CN101639947A (en) Image-based plant three-dimensional shape measurement and reconstruction method and system
CN102156985A (en) Method for counting pedestrians and vehicles based on virtual gate
JP2004078939A (en) Object identification method, apparatus and program
CN101562675A (en) No-reference image quality evaluation method based on Contourlet transform
CN102314610A (en) Object-oriented image clustering method based on probabilistic latent semantic analysis (PLSA) model
CN101916436A (en) Multi-scale spatial projecting and remote sensing image fusing method
CN101403795A (en) Remote sensing survey method and system for estimating tree coverage percentage of city
CN102708370A (en) Method and device for extracting multi-view angle image foreground target
CN102945378A (en) Method for detecting potential target regions of remote sensing image on basis of monitoring method
CN101916435A (en) Method for fusing multi-scale spectrum projection remote sensing images
US20130342694A1 (en) Method and system for use of intrinsic images in an automotive driver-vehicle-assistance device
CN101710418A (en) Interactive mode image partitioning method based on geodesic distance
CN101853509A (en) SAR (Synthetic Aperture Radar) image segmentation method based on Treelets and fuzzy C-means clustering
CN102509279A (en) Self-adapting shaded-area detail reproduction method for tongue image with sufficient root-part illumination
CN103455991A (en) Multi-focus image fusion method
CN102393966A (en) Self-adapting image compressive sampling method based on multi-dimension saliency map
CN101777129A (en) Image matching method based on feature detection
US20130155235A1 (en) Image processing method
JP2006244119A (en) Retrieval device for texture image, retrieval method, retrieval program and recording medium with its program recorded
CN103279939A (en) Image stitching processing system

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance
C53 Change or modification
COR Bibliographic change or correction in the description

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

C14 Granted