CN114139618A - Signal dependent noise parameter estimation method based on improved density peak clustering - Google Patents

Signal dependent noise parameter estimation method based on improved density peak clustering Download PDF

Info

Publication number
CN114139618A
CN114139618A CN202111406097.2A CN202111406097A CN114139618A CN 114139618 A CN114139618 A CN 114139618A CN 202111406097 A CN202111406097 A CN 202111406097A CN 114139618 A CN114139618 A CN 114139618A
Authority
CN
China
Prior art keywords
clustering
noise
characteristic data
density
sample
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
CN202111406097.2A
Other languages
Chinese (zh)
Other versions
CN114139618B (en
Inventor
吴俣倩
张钰
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.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi University
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 Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN202111406097.2A priority Critical patent/CN114139618B/en
Publication of CN114139618A publication Critical patent/CN114139618A/en
Application granted granted Critical
Publication of CN114139618B publication Critical patent/CN114139618B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)

Abstract

本发明公开了基于改进密度峰值聚类的信号依赖噪声参数估计方法。通过滑动窗口从含有噪声的图像中提取样本，计算均值、熵和梯度作为特征数据。再输入聚类算法中进行聚类，区分弱纹理样本与强纹理样本。在聚类的过程中，引入了相对密度的概念，通过数据点与周围数据的距离，划分了比较范围，计算每个比较范围中数据点的相对密度，最后选择相对密度高的数据点作为聚类中心，解决了传统的DPC算法在对密度不均匀数据集进行聚类时往往忽视稀疏簇中心，从而影响聚类精度的问题。根据聚类结果，对簇标签为弱纹理的样本进行像素水平估计与噪声估计，最后通过最小二乘法拟合像素值‑噪声方差估计对，得到原始图像的噪声参数估计值，实现去噪的准备工作。The invention discloses a signal-dependent noise parameter estimation method based on improved density peak clustering. Extract samples from noisy images through sliding windows, and calculate mean, entropy and gradient as feature data. Then input into the clustering algorithm for clustering to distinguish weak texture samples from strong texture samples. In the process of clustering, the concept of relative density is introduced, the comparison range is divided according to the distance between the data point and the surrounding data, the relative density of the data points in each comparison range is calculated, and finally the data point with high relative density is selected as the cluster. The cluster center solves the problem that the traditional DPC algorithm often ignores the sparse cluster center when clustering data sets with uneven density, thus affecting the clustering accuracy. According to the clustering results, the pixel level estimation and noise estimation are performed for the samples whose cluster label is weak texture, and finally the pixel value-noise variance estimation pair is fitted by the least square method to obtain the estimated value of the noise parameter of the original image, and the preparation for denoising is realized. Work.

Description

Signal dependent noise parameter estimation method based on improved density peak clustering
Technical Field
The invention belongs to the technical field of image noise signal processing, and particularly relates to a signal dependent noise parameter estimation method based on improved density peak clustering.
Background
The image is an important carrier for information recording and transmission in modern society, but the image is inevitably affected by noise in the processes of acquisition, storage and transmission, so that the quality of the image is reduced. The noise level is an important parameter of image processing optimization algorithms such as image denoising, image compression, image splicing and the like, so that accurately estimating the noise level has a very important meaning for image denoising. With the development of CMOS image sensors, in order to reduce the influence of noise, some manufacturers directly embed a noise reduction module in an image sensor chip, which can effectively suppress noise independent of a signal whose noise component is becoming a main noise source of the CMOS image sensor. At present, most of estimation of signal dependent noise is based on weak texture image blocks, and the method has the difficulty that the weak texture image blocks are selected, then a pair of pixel values and noise variance are estimated for each image block, and finally, noise parameters are fitted. The weak texture image blocks may be obtained using a clustering algorithm.
Clustering is a process of grouping different objects according to similarity, and objects with high similarity are grouped into the same group. Clustering analysis, which is the main content of unsupervised learning, has been an important component in the fields of machine learning, pattern recognition, data mining, and the like. At present, many clustering algorithms have appeared, and can be roughly divided into a partition clustering method, a hierarchical clustering method, a density clustering method, a model clustering method, a grid clustering method and the like.
In 2014, Rodriguez and Laio proposed a density peak Clustering algorithm (DPC) based on density and distance on Science, which can search and find density peaks quickly, then determine cluster centers, and assign the remaining points to corresponding clusters. The DPC algorithm has the advantages of low parameter requirement, non-iteration and the like, and can effectively detect the clusters of any shapes from a large-scale data set with low calculation complexity. However, the existing density peak clustering algorithm is more prone to select the clustering center of the dense region, and the clustering of the sparse region is often ignored. For a data set with large density difference, the DPC algorithm often determines density points of a sparse region as abnormal values or allocates the abnormal values to adjacent dense clusters in a staggered manner, so that the cluster center of the sparse cluster is ignored, and thus the inaccuracy of a clustering result is caused. Some existing DPC-based improved clustering algorithms often improve the clustering effect by introducing K Nearest Neighbor (KNN), but the parameter k is a numerical value set artificially and subjectively according to past experience, and the size of the k value often has a large influence on the clustering result, so that the clustering accuracy is further influenced.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a signal dependent noise parameter estimation method based on improved density peak value clustering, relative density definition is introduced, a clustering center is determined according to picture characteristic distribution, parameter values do not need to be defined artificially, clustering precision is improved, and accurate estimation of noise parameters is realized.
The signal dependent noise parameter estimation method based on the improved density peak clustering specifically comprises the following steps:
step one, extracting a plurality of samples with the same size from an original image containing noise by using a sliding window. And calculating the mean value, entropy and gradient of each sample as the characteristic data of the sample. And (4) forming a feature data set by the feature data of all the samples, and then carrying out normalization processing.
Step two, calculating the following parameters aiming at the feature data set X after normalization processing in the step one:
characteristic data xiEuclidean distance d from other characteristic dataij：
dij＝||xi-xj||
② characteristic data xiLocal density of (p)i：
Figure BDA0003372860730000021
③ characteristic data xiDistance δ from nearest higher density feature datai：
Figure BDA0003372860730000022
Characteristic data xiDistance θ from nearest lower density feature datai：
Figure BDA0003372860730000023
Wherein x isi、xj∈X，dcThe cutoff distance is a value at 1% to 2% after arranging the distances between all feature data in ascending order.
Step three, dividing the characteristic data x according to the calculation result of the step twoiRelative density comparison range η ofiAnd calculating the relative density ρ thereofi'：
ηi＝max(θi,δi)
Figure BDA0003372860730000031
Wherein N isiIs a relative density comparison range etaiThe number of characteristic data in the table.
Figure BDA0003372860730000032
Representing characteristic data xjWherein the distance characteristic data xiThe more recent, the feature data xiThe greater the relative density contribution of (a), the greater the weight occupied. Defining feature data xiIs the ratio of its local density to the cumulative average of the local densities of the surrounding feature data.
Step four, using the characteristic data x obtained in the step threeiRelative density of (g)i' alternative local Density ρiRecalculating the feature data xiDistance δ from nearest higher density feature datai. Then, the feature data x is calculatediCluster value of gammai：
γi＝ρi'·δi
Calculating to obtain a clustering value gammaiSorting according to descending order, selecting two feature data with the highest clustering value as clustering centers, distributing other feature data to clusters with the clustering centers closer to each other, and outputting each feature dataAnd finishing the clustering process by using the cluster labels corresponding to the characteristic data.
Step five, calculating the estimated pixel level of the weak texture sample according to the clustering result of the step four
Figure BDA0003372860730000033
And variance of noise
Figure BDA0003372860730000034
Figure BDA0003372860730000035
wk＝1～B_weak_n
Figure BDA0003372860730000036
wk∈[1,B_weak_n]
Where N is the side length of the sample, wk denotes the index of the weak texture sample, B _ bead _ N is the number of weak texture samples, Iwk(m, n) represents a weak texture sample IwkThe pixel value of the mth row and the nth column.
Figure BDA0003372860730000038
Is the minimum variance direction vector of the weak texture samples, and the covariance matrix CPCorrelation of minimum eigenvalues of:
Figure BDA0003372860730000037
step six, configuring the noise in the original image into a Poisson-Gaussian noise model, and then carrying out a total noise method sigma of (p, q) positions in the original image2(p, q) is:
σ2(p,q)＝ax(p,q)+b
where x (p, q) represents the noise-free pixel value at the (p, q) position in the original image. a. b is a noise parameter. Fitting weak texture samples I using least squareswkPixel value-to-noise variance estimation pair of
Figure BDA0003372860730000041
Obtaining the estimated value of the noise parameter of the original image
Figure BDA0003372860730000042
And
Figure BDA0003372860730000043
the invention has the following beneficial effects:
1. according to the method, for a data set with uneven density distribution of characteristic data of signal-related noise of a CMOS image sensor, in the clustering process of samples, relative density is defined, the density is calculated according to the data distribution characteristics of the data set, so that the central point of a sparse cluster can have a larger density value, the density value of a sparse cluster in the data set with large density difference is prevented from being always lower than that of a dense cluster, and the sparse cluster is prevented from being taken as an outlier or wrongly distributed to other high-density clusters.
2. The comparison areas are divided according to the data distribution characteristics of the data sets, and subjectively determined parameter values are not introduced, so that different data sets have different relative density comparison ranges in a self-adaptive manner, and the influence of artificial experience judgment on experimental results is avoided.
3. The accuracy of the clustering result is improved by introducing the relative density, so that a more accurate weak texture region is selected, the noise estimation of the whole image is further carried out, the estimation accuracy is improved, and the later-stage image denoising is facilitated.
Detailed Description
The invention is further explained in the following with reference to the accompanying drawings; the raw image used in this embodiment is derived from an image captured by a CMOS image sensor, and noise in the raw image is a signal-dependent noise component generated by a noise reduction module of the CMOS image sensor during the capturing process.
The signal dependent noise parameter estimation method based on the improved density peak clustering specifically comprises the following steps:
step one, using a sliding window with the size of 16 x 16, sliding the distance of one pixel point at a time according to the sequence from top to bottom and from left to right, and extracting n samples with the size of 16 x 16 from an original image containing noise. Calculating the mean, entropy, and gradient of each sample as the characteristic data of the sample:
Figure BDA0003372860730000044
Figure BDA0003372860730000045
Figure BDA0003372860730000046
where, gray represents the magnitude of the gray value, and p (gray) represents the probability that the gray value is gray. W, H is the width and height of the sample, pw,hThe pixel value of the (w, h) position in the sample.
I.e. feature data x for each samplei’Has three attributes of mean, entropy, and gradient, xi'＝{xi1,xi2,xi3}. And (4) forming a feature data set by the feature data of all the samples, and performing normalization processing.
Step two, calculating the following parameters aiming at the feature data set X after normalization processing in the step one:
characteristic data xiEuclidean distance d from other characteristic dataij：
dij＝||xi-xj||
② characteristic data xiLocal density of (p)i：
Figure BDA0003372860730000051
③ characteristic data xiDistance δ from nearest higher density feature datai：
Figure BDA0003372860730000052
Characteristic data xiDistance θ from nearest lower density feature datai：
Figure BDA0003372860730000053
Wherein x isi、xj∈X，dcThe size is a value at 1% to 2% after arranging the distances between all feature data in ascending order, as the cutoff distance. When the feature data xiIs the global lowest, thetai＝δi。
Step three, dividing the characteristic data x according to the calculation result of the step twoiRelative density comparison range η ofiAnd calculating the relative density ρ thereofi'：
ηi＝max(θi,δi)
When etai＝δiTime, description feature data xiSurrounded by feature data of lower density than the feature data, and more likely to become the center point.
When etai＝θiTime, description feature data xiSurrounded by feature data with a higher density than the density, the probability of becoming a center point is small.
Figure BDA0003372860730000054
Wherein N isiIs a relative density comparison range etaiThe number of characteristic data in the table.
Figure BDA0003372860730000055
Representing characteristic data xjWherein the distance characteristic data xiThe more recent, the feature data xiThe greater the relative density contribution of (A), is occupiedThe greater the weight. Defining feature data xiIs the ratio of its local density to the cumulative average of the local densities of the surrounding feature data. Thus, the density value calculated by each point is the relative size, not the absolute size, compared with the density of the points around the point, the local structure of the cluster is better considered, and even the cluster center of the sparse cluster can obtain a larger density value, so that the sparse cluster is better identified.
Step four, using the characteristic data x obtained in the step threeiRelative density of (g)i' alternative local Density ρiRecalculating the feature data xiDistance δ from nearest higher density feature datai. A traditional DPC algorithm needs to draw a decision diagram according to rho and delta, and then manually selects a point with larger rho and delta as a clustering center. The subjective factors of the method account for a great proportion, and the condition that the clustering center cannot be easily selected by naked eyes exists, so that the domino effect is directly caused in the subsequent distribution step, the final clustering result is influenced, and therefore the product of the calculated relative density and the calculated distance is selected as the characteristic data xiCluster value of gammai：
γi＝ρi'·δi
Calculating to obtain a clustering value gammaiAnd sorting according to a descending order, selecting two feature data with the highest clustering value as clustering centers, distributing other feature data into clusters with the clustering centers which are closer to each other, outputting a cluster label corresponding to each feature data, dividing the cluster label into a weak texture sample and a strong texture sample, and finding the weak texture sample after finishing the clustering process.
Step five, calculating the estimated pixel level of the weak texture sample according to the clustering result of the step four
Figure BDA0003372860730000061
And variance of noise
Figure BDA0003372860730000062
Figure BDA0003372860730000063
wk＝1～B_weak_n
Figure BDA0003372860730000064
wk∈[1,B_weak_n]
Where N is the side length of each sample, N is 16 in this embodiment, wk denotes the index of the weak texture sample, B _ bead _ N is the number of weak texture samples, Iwk(m, n) represents a weak texture sample IwkThe pixel value of the mth row and the nth column.
Figure BDA0003372860730000065
Is the minimum variance direction vector of the weak texture samples, and the covariance matrix CPCorrelation of minimum eigenvalues of:
Figure BDA0003372860730000066
step six, configuring the noise in the original image into a Poisson-Gaussian noise model, and then carrying out a total noise method sigma of (p, q) positions in the original image2(p, q) is:
σ2(p,q)＝ax(p,q)+b
where x (p, q) represents the noise-free pixel value at the (p, q) position in the original image. a. b is a noise parameter. Fitting weak texture samples I using least squareswkPixel value-to-noise variance estimation pair of
Figure BDA0003372860730000067
Obtaining the estimated value of the noise parameter of the original image
Figure BDA0003372860730000071
And
Figure BDA0003372860730000072
after determining the noise estimate in the image, the magnitude of the noise component in the image may be determined for use in subsequent CMOS image sensor signalsThe number depends on the removal of noise.

Claims (3)

1. The signal dependent noise parameter estimation method based on the improved density peak value clustering is characterized by comprising the following steps: the method specifically comprises the following steps:
the method comprises the following steps of firstly, extracting a plurality of samples with the same size from an original image containing noise by using a sliding window; calculating the mean value, entropy and gradient of each sample as the characteristic data of the sample; normalizing the characteristic data of all samples to form a characteristic data set X;
step two, calculating the following parameters aiming at the characteristic data set X:
characteristic data xiEuclidean distance d from other characteristic dataij：
dij＝||xi-xj|| (1)
② characteristic data xiLocal density of (p)i：
Figure FDA0003372860720000011
③ characteristic data xiDistance δ from nearest higher density feature datai：
Figure FDA0003372860720000012
Characteristic data xiDistance θ from nearest lower density feature datai：
Figure FDA0003372860720000013
Wherein x isi、xj∈X，dcThe size of the cut-off distance is a value of 1% -2% of the distances between all the characteristic data after the distances are arranged in ascending order;
step three, dividing the characteristic data x according to the calculation result of the step twoiRelative density comparison range η ofiAnd calculating the relative density ρ thereofi′：
ηi＝max(θi,δi) (5)
Figure FDA0003372860720000021
Wherein N isiIs a relative density comparison range etaiThe number of the characteristic data in the database;
Figure FDA0003372860720000022
representing characteristic data xjThe weight of (c);
step four, using the characteristic data x obtained in the step threeiRelative density of (g)i' alternative local Density ρiRecalculating the feature data x according to equation (3)iDistance δ from nearest higher density feature datai(ii) a Then, the feature data x is calculatediCluster value of gammai：
γi＝ρi'·δi (7)
Calculating to obtain a clustering value gammaiSorting in a descending order, selecting two feature data with the highest clustering values as clustering centers for clustering, and then outputting a cluster label corresponding to each feature data to finish the clustering process;
step five, calculating the estimated pixel level of the weak texture sample according to the clustering result of the step four
Figure FDA0003372860720000023
And variance of noise
Figure FDA0003372860720000024
Figure FDA0003372860720000025
Figure FDA0003372860720000026
Where N is the side length of the sample, wk denotes the index of the weak texture sample, B _ bead _ N is the number of weak texture samples, Iwk(m, n) represents a weak texture sample IwkThe pixel values of the mth row and the nth column;
Figure FDA0003372860720000027
is the minimum variance direction vector of the weak texture samples, and the covariance matrix CPCorrelation of minimum eigenvalues of:
Figure FDA0003372860720000028
step six, configuring the noise in the original image into a Poisson-Gaussian noise model, and then carrying out a total noise method sigma of (p, q) positions in the original image2(p, q) is:
σ2(p,q)＝ax(p,q)+b (11)
wherein x (p, q) represents a noise-free pixel value at a position (p, q) in the original image; a. b is a noise parameter; fitting weak texture samples I using least squareswkPixel value-to-noise variance estimation pair of
Figure FDA0003372860720000029
Obtaining the estimated value of the noise parameter of the original image
Figure FDA00033728607200000210
And
Figure FDA00033728607200000211
2. the method of claim 1 for signal dependent noise parameter estimation based on improved density peak clustering, wherein: step one, using a sliding window with the size of 16 x 16, sliding the distance of one pixel point at a time from top to bottom and from left to right, and extracting a plurality of samples with the size of 16 x 16 from an original image containing noise.
3. A signal dependent noise parameter estimation method based on improved density peak clustering according to claim 1 or 2, characterized by: calculating the mean, entropy, and gradient of each sample as the characteristic data of the sample:
Figure FDA0003372860720000031
Figure FDA0003372860720000032
Figure FDA0003372860720000033
wherein, gray represents the magnitude of the gray value, and p (gray) represents the probability of the gray value being gray; w, H is the width and height of the sample, pw,hThe pixel value of the (w, h) position in the sample.
CN202111406097.2A 2021-11-24 2021-11-24 Signal-dependent noise parameter estimation method based on improved density peak clustering Active CN114139618B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111406097.2A CN114139618B (en) 2021-11-24 2021-11-24 Signal-dependent noise parameter estimation method based on improved density peak clustering

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111406097.2A CN114139618B (en) 2021-11-24 2021-11-24 Signal-dependent noise parameter estimation method based on improved density peak clustering

Publications (2)

Publication Number Publication Date
CN114139618A true CN114139618A (en) 2022-03-04
CN114139618B CN114139618B (en) 2025-02-11

Family

ID=80391343

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111406097.2A Active CN114139618B (en) 2021-11-24 2021-11-24 Signal-dependent noise parameter estimation method based on improved density peak clustering

Country Status (1)

Country Link
CN (1) CN114139618B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114613456A (en) * 2022-03-07 2022-06-10 哈尔滨理工大学 A Hardness Prediction Method for High-Entropy Alloys Based on Improved Density Peak Clustering Algorithm
CN115112032A (en) * 2022-06-06 2022-09-27 上海工程技术大学 A method for measuring the pitch of spherical grid components
CN115563522A (en) * 2022-12-02 2023-01-03 湖南工商大学 Clustering method, device, equipment and medium of traffic data
CN116342915A (en) * 2023-05-30 2023-06-27 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) A deep image clustering method, system and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170178309A1 (en) * 2014-05-15 2017-06-22 Wrnch Inc. Methods and systems for the estimation of different types of noise in image and video signals
CN110046639A (en) * 2019-01-10 2019-07-23 湖南理工学院 A kind of Hyperspectral imaging noise label detection method based on super-pixel weight density
CN110503625A (en) * 2019-07-02 2019-11-26 杭州电子科技大学 A Parameter Estimation Method of CMOS Image Signal Correlation Noise

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170178309A1 (en) * 2014-05-15 2017-06-22 Wrnch Inc. Methods and systems for the estimation of different types of noise in image and video signals
CN110046639A (en) * 2019-01-10 2019-07-23 湖南理工学院 A kind of Hyperspectral imaging noise label detection method based on super-pixel weight density
CN110503625A (en) * 2019-07-02 2019-11-26 杭州电子科技大学 A Parameter Estimation Method of CMOS Image Signal Correlation Noise

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张钰等: "去除冗余背景的手势连续轮廓提取算法与实验", RESEARCH AND EXPLORATION IN LABORATORY, 30 September 2017 (2017-09-30), pages 1 - 7 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114613456A (en) * 2022-03-07 2022-06-10 哈尔滨理工大学 A Hardness Prediction Method for High-Entropy Alloys Based on Improved Density Peak Clustering Algorithm
CN115112032A (en) * 2022-06-06 2022-09-27 上海工程技术大学 A method for measuring the pitch of spherical grid components
CN115563522A (en) * 2022-12-02 2023-01-03 湖南工商大学 Clustering method, device, equipment and medium of traffic data
CN115563522B (en) * 2022-12-02 2023-04-07 湖南工商大学 Clustering method, device, equipment and medium of traffic data
CN116342915A (en) * 2023-05-30 2023-06-27 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) A deep image clustering method, system and storage medium
CN116342915B (en) * 2023-05-30 2024-06-25 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) A deep image clustering method, system and storage medium

Also Published As

Publication number Publication date
CN114139618B (en) 2025-02-11

Similar Documents

Publication Publication Date Title
CN114139618B (en) Signal-dependent noise parameter estimation method based on improved density peak clustering
CN115082419B (en) Blow-molded luggage production defect detection method
CN115082467B (en) Building material welding surface defect detection method based on computer vision
CN115082683B (en) Injection molding defect detection method based on image processing
CN114723705B (en) Cloth flaw detection method based on image processing
CN103544499B (en) The textural characteristics dimension reduction method that a kind of surface blemish based on machine vision is detected
CN115187548A (en) An artificial intelligence-based defect detection method for mechanical parts
CN114998198A (en) Injection molding surface defect identification method
CN112561849B (en) Wafer defect detection method
CN116205919A (en) Hardware part production quality detection method and system based on artificial intelligence
JP2002288658A (en) Object extraction apparatus and method based on region feature value matching of region-divided video
CN115100199B (en) Method for detecting wafer low-texture defects
CN117522864B (en) European pine plate surface flaw detection method based on machine vision
CN115375690B (en) Classification and identification method for greasy tongue coating
CN114820625B (en) Automobile top block defect detection method
CN108154158B (en) A Building Image Segmentation Method for Augmented Reality Applications
CN113657216A (en) Method for separating tree crown and wood point of tree in point cloud scene based on shape characteristics
CN116486091B (en) Fan blade defect area rapid segmentation method and system based on artificial intelligence
CN115456944A (en) Ceramic substrate appearance defect intelligent identification method based on target detection
CN115294109A (en) Real wood board production defect identification system based on artificial intelligence, and electronic equipment
CN103942526B (en) Linear feature extraction method for discrete data point set
CN102708367A (en) Image identification method based on target contour features
CN110930425A (en) Damaged target detection method based on neighborhood vector inner product local contrast image enhancement
CN109166093B (en) An image salient region detection method
CN117975175B (en) Plastic pipeline appearance defect detection method based on machine vision

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