CN112613583B - High-frequency information extraction clustering method for low-frequency noise face image - Google Patents
High-frequency information extraction clustering method for low-frequency noise face image Download PDFInfo
- Publication number
- CN112613583B CN112613583B CN202110009835.3A CN202110009835A CN112613583B CN 112613583 B CN112613583 B CN 112613583B CN 202110009835 A CN202110009835 A CN 202110009835A CN 112613583 B CN112613583 B CN 112613583B
- Authority
- CN
- China
- Prior art keywords
- sample
- matrix
- clustering
- frequency
- image
- 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.)
- Active
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
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Algebra (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computing Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a high-frequency information extraction clustering method for a low-frequency noise face image, which comprises the following steps of: s1: preprocessing a sample data set to obtain a plurality of sample matrixes; s2: extracting high-frequency information from the plurality of sample matrixes to correspondingly obtain a plurality of sample models which retain high-frequency characteristics; s3: calculating a correlation coefficient matrix between sample models; s4: constructing an initial Laplace matrix according to the correlation coefficient matrix; the initial Laplace matrix is standardized, and a standardized Laplace matrix L is obtained; s5: determining the dimension of the cluster as k 1 The method comprises the steps of carrying out a first treatment on the surface of the Calculating the minimum top k of L 1 The feature vectors corresponding to the feature values are normalized according to the rows to obtain n multiplied by k 1 A feature matrix of the dimension; s6: for n x k 1 And clustering the feature matrix of the dimension to obtain a clustering result. The invention provides a high-frequency information extraction clustering method for a low-frequency noise face image, which solves the problem that the existing clustering method cannot meet the clustering requirement of sparse data with less computing resources.
Description
Technical Field
The invention relates to the technical field of image clustering, in particular to a high-frequency information extraction clustering method for a low-frequency noise face image.
Background
Cluster analysis is an important research topic in the field of artificial intelligence, and clustering is a technology for searching internal structures among data, so that all data entities form clusters. The cluster analysis has wide application in the fields of image recognition, business, biomedical treatment, insurance industry, electronic commerce and the like.
Traditional cluster analysis methods include partition-based, hierarchy-based, density-based, mesh-based, model-based, and the like. Wherein the partition-based clustering method is a top-down method, dividing a given data set of several data objects into a plurality of partitions, such that each partition represents a cluster, wherein the K-Means algorithm is the most classical partition-based clustering method; the hierarchical-based clustering method is to decompose the designated data in a hierarchy according to the requirement of a certain condition; the density-based method is to find a high-density region separated by a low-density region, namely, starting from the density of a distribution domain, clustering is continued when the density of a data object exceeds a certain threshold, wherein the most representative method is a DBSAN algorithm; the grid-based method is to place all clusters on a grid for analysis, divide each attribute into adjacent intervals and create a set of cells; model-based clustering methods adapt, i.e., fit, data to models, which are assumed to be density functions of data objects in spatial distribution, etc. However, the existing clustering method requires more computing resources during clustering, and cannot meet the clustering requirement of sparse data by using fewer computing resources.
In the prior art, as disclosed in 2019-03-08, an image clustering method based on double-graph sparse depth matrix decomposition is disclosed as CN109447147A, which is used for solving the technical problems of low image clustering accuracy and low operation speed of a clustering process in the prior art, but cannot meet the clustering requirement of sparse data by using less computing resources.
Disclosure of Invention
The invention provides a high-frequency information extraction clustering method for a low-frequency noise face image, which aims at overcoming the technical defect that the existing clustering method cannot meet the clustering requirement of sparse data with less computing resources.
In order to solve the technical problems, the technical scheme of the invention is as follows:
a high-frequency information extraction clustering method for a low-frequency noise face image comprises the following steps:
s1: preprocessing a sample data set to obtain a plurality of sample matrixes;
s2: extracting high-frequency information from the plurality of sample matrixes to correspondingly obtain a plurality of sample models which retain high-frequency characteristics;
s3: calculating a correlation coefficient matrix between sample models;
s4: constructing an initial Laplace matrix according to the correlation coefficient matrix;
the initial Laplace matrix is standardized, and a standardized Laplace matrix L is obtained;
s5: determining the dimension of the cluster as k 1 ；
Calculating the minimum top k of L 1 The feature vectors corresponding to the feature values are normalized according to the rows to obtain n multiplied by k 1 A feature matrix of the dimension; wherein n is the number of image samples in the sample dataset;
s6: for n x k 1 And clustering the feature matrix of the dimension to obtain a clustering result.
Preferably, the sample data set is a mat file generated by compressing pixel points of n image samples.
Preferably, in step S1, the preprocessing of the sample data set is specifically: filling and restoring each vector in the sample data set into a corresponding sample matrix according to the length and the width of the image sample;
wherein the value of each pixel point in the sample matrix is between 0 and 255.
Preferably, in step S2, the high-frequency information extraction of the sample matrix includes the steps of:
s2.1: performing two-dimensional Fourier transform on the sample matrix to obtain an original frequency spectrum;
s2.2: symmetrically transforming the original frequency spectrum to obtain a ffftshift image;
s2.3: designing a Butterworth filter, and passing the ffftshift image through the Butterworth filter;
s2.4: the ffftshift image passing through the butterworth filter is subjected to inverse fourier transform to obtain a sample model retaining high-frequency characteristics.
Preferably, the method further comprises transforming the values of the sample matrix before performing the two-dimensional fourier transform on the sample matrix.
Preferably, the order of the butterworth filter is 2 nd order, and the cut-off frequency is 5% of the width of the image sample.
Preferably, in step S3, a correlation coefficient matrix between each sample model is determined by:
the sample model is regarded as an energy signal sequence, the correlation between the two sample models is calculated to be the similarity between the two energy signals, the comparison result value is the cross correlation coefficient of the two energy signal sequences, the value is from 0 to 1, and the higher the value is, the more similar the sample result is; and obtaining a correlation coefficient matrix by circularly calculating the mutual correlation coefficients between every two.
Preferably, after obtaining the correlation coefficient matrix, the method further includes: the first K largest values in the correlation coefficient matrix are truncated as a reduced correlation coefficient matrix W.
Preferably, in step S4, the method specifically includes the following steps:
s4.1: obtaining a degree matrix D according to the simplified correlation coefficient matrix W;
s4.2: construction of an initial Laplace matrix L 1 ＝D-W；
S4.3: building a normalized laplacian matrix l=d -1/2 L 1 D -1/2 。
Preferably, in step S6, the K-Means algorithm is used for n K 1 Clustering the feature matrix of the dimension to obtain a clustering result; the clustering result is cluster division c=c 1 ，C 2 ，...，C k2 The method comprises the steps of carrying out a first treatment on the surface of the Where k2 is the number of clusters.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the invention provides a high-frequency information extraction clustering method for a low-frequency noise face image, which only reserves the most important features by extracting the high-frequency information of a sample matrix, reduces the computing resources required by clustering and realizes that less computing resources are used for meeting the clustering requirement of sparse data.
Drawings
FIG. 1 is a flow chart of the steps performed in the technical scheme of the invention;
FIG. 2 is a partial image sample obtained by filling and restoring in the present invention;
FIG. 3 is a diagram of a ffftshift image of the present invention that has been symmetrically transformed;
FIG. 4 is a sample image before high frequency information extraction in the present invention;
fig. 5 is a sample image after high frequency information extraction in the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
for the purpose of better illustrating the embodiments, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the actual product dimensions;
it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
As shown in fig. 1, a high-frequency information extraction clustering method for a low-frequency noise face image includes the following steps:
s1: preprocessing a sample data set to obtain a plurality of sample matrixes;
s2: extracting high-frequency information from the plurality of sample matrixes to correspondingly obtain a plurality of sample models which retain high-frequency characteristics;
s3: calculating a correlation coefficient matrix between sample models;
s4: constructing an initial Laplace matrix according to the correlation coefficient matrix;
the initial Laplace matrix is standardized, and a standardized Laplace matrix L is obtained;
s5: determining the dimension of the cluster as k 1 ；
Calculating the minimum top k of L 1 The feature vectors corresponding to the feature values are normalized according to the rows to obtain n multiplied by k 1 A feature matrix of the dimension; where n is the sample datasetThe number of middle image samples;
s6: for n x k 1 And clustering the feature matrix of the dimension to obtain a clustering result.
Example 2
More specifically, the sample data set is a mat file generated by compressing pixels of n image samples.
In this embodiment, the sample data set is set to be a 2200×260 mat file, where 260 is the number of image samples and 2200 is a vector formed by stretching sample data in the sample data set.
More specifically, in step S1, the preprocessing of the sample data set is specifically: filling and restoring each vector in the sample data set into a corresponding sample matrix according to the length and the width of the image sample;
wherein the value of each pixel point in the sample matrix is between 0 and 255.
In the specific implementation, the image is a standard rectangle, and the length and width of the image are expressed by rows and columns of the matrix. The sample data set is a matrix formed by stretching a standard matrix corresponding to each image sample into a column combination. In this embodiment, the length and width dimensions of the image are set to 55×40, and fig. 2 is a partial image sample obtained by restoration, where 10 samples in each row are the same cluster.
More specifically, in step S2, the high-frequency information extraction of the sample matrix includes the steps of:
s2.1: performing two-dimensional Fourier transform on the sample matrix to obtain an original frequency spectrum, wherein the high frequency is in the middle, and the low frequency is around;
s2.2: symmetrically transforming the original frequency spectrum, as shown in fig. 3, to obtain ffftshift (moving zero frequency point to middle of frequency spectrum) images with low frequency in the middle and high frequency around;
s2.3: designing a Butterworth filter, and passing the ffftshift image through the Butterworth filter;
s2.4: the ffftshift image passing through the Butterworth filter is subjected to Fourier inverse transformation to obtain a sample model with high-frequency characteristics reserved, the high-frequency characteristics, namely the outline shape of the image, are reserved, the low-frequency characteristics are filtered, and the interference of the low frequency band on subsequent cluster analysis is reduced, as shown in figures 4-5.
In the implementation process, the sample matrix has higher dimensionality, which means that the data is sparse at the same time, and the influence of noise is large when cluster analysis is carried out. The image sample comprises a low-frequency part and a high-frequency part, the color of the low-frequency part is gradually changed, namely the gray level is slowly changed, the frequency of the high-frequency part is quickly changed, the gray level is obviously and severely changed, the gray level value of the edge of one image and background part is quickly changed, the image contour of the sample is highlighted, namely the high-frequency part of the image represents the rough information of the image contour.
More specifically, the method further comprises transforming the values of the sample matrix before performing the two-dimensional fourier transform on the sample matrix.
In the specific implementation process, the accuracy can be ensured by carrying out the value transformation, and the loss of the numerical value during the two-dimensional Fourier transformation (FFT 2) transformation is avoided.
More specifically, the butterworth filter has an order of 2, and the cut-off frequency is 5% of the image sample width.
In a specific implementation, the ffftshift image is passed through a butterworth filter, so that the frequency response curve in the passband is maximally flat, and gradually drops to zero in the passband. The low frequency band in the image sample is filtered, most of the high frequency band is reserved, the possibility of dimension reduction is realized for cluster analysis, the interference of low frequency characteristics is reduced, and the accuracy of clustering is improved.
More specifically, in step S3, a correlation coefficient matrix between each sample model is determined by:
regarding the sample model as an energy signal sequence, calculating the correlation between the two sample models, namely comparing the similarity between the two energy signals, wherein the comparison result value is the cross-correlation coefficient of the two energy signal sequencesWherein gamma is xy As the inner product of sequence X and sequence Y,γ xx and gamma yy Respectively, the inner products of the sequence X and the sequence Y to obtain a cross-correlation coefficient rho xy In order to be the correlation between the two corresponding sample models, the value is from 0 to 1, and the higher the value is, the more similar the sample results are; and obtaining a correlation coefficient matrix by circularly calculating the mutual correlation coefficients between every two.
In the specific implementation process, a mode of calculating and obtaining a correlation coefficient matrix by using a Gaussian kernel function is avoided, and the operation efficiency and accuracy are improved.
More specifically, the method further comprises the following steps of: the first K largest values in the correlation coefficient matrix are truncated as a reduced correlation coefficient matrix W.
In a specific implementation process, the sample data set of the embodiment includes 260 images, and the number of clusters is only 10, which means that the number of the same kind of each image sample is about tens, therefore, the correlation coefficient matrix can be simplified, only the front K term with the largest similarity is taken, the value of the front K term with the largest correlation is set as 0 in the correlation coefficient matrix, and the calculation efficiency in the clustering process can be improved by setting the proper K value, and the accuracy of the clustering is also guaranteed sufficiently. In order to obtain the best clustering effect, the present embodiment sets the K value to 260 as full connection.
More specifically, in step S4, the following steps are specifically included:
s4.1: obtaining a degree matrix D according to the simplified correlation coefficient matrix W;
s4.2: construction of an initial Laplace matrix L 1 ＝D-W；
S4.3: building a normalized laplacian matrix l=d -1/2 L 1 D -1/2 。
In the implementation process, the value of each row of the correlation matrix W is accumulated and placed on a diagonal line, and other values are assigned to 0, so as to obtain a degree matrix D.
More specifically, in step S6, the K-Means algorithm is used on n×k 1 Clustering the feature matrix of the dimension to obtain a clustering result; the clustering result is cluster division c=c 1 ，C 2 ，...，C k2 The method comprises the steps of carrying out a first treatment on the surface of the Where k2 is the number of clusters.
In an implementation, the dimension k of the cluster is reduced 1 The efficiency of the operation can be improved. Setting k in this embodiment 1 The method comprises the steps of (1) calculating the feature vectors corresponding to the first 14 minimum feature values of L, and normalizing the feature vectors according to rows to obtain a 260×14 feature matrix, wherein each row of the feature matrix is taken as a 14-dimensional sample, so that dimension reduction operation of samples from 2200 to 14 is realized, and the total number of the samples is 260; the cluster number k2 is set to 10, resulting in cluster division c=c 1 ，C 2 ，...，C 10 。
Example 3
Table 1 is a comparison of the effects of the three clustering methods.
TABLE 1
The sample data set used is 10 persons, 26 images are obtained, no labels exist, namely 260 image samples are obtained, the category number is set to be 10, and the height and width of the images are respectively set to be 55 and 40. As can be seen from Table 1, the indexes such as the accuracy of the high-frequency information extraction clustering method for the low-frequency noise face image are greatly improved compared with the traditional K-means algorithm.
It is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (6)
1. A high-frequency information extraction clustering method for a low-frequency noise face image is characterized by comprising the following steps of:
s1: preprocessing a sample data set to obtain a plurality of sample matrixes; the sample data set is a mat file generated by compressing pixel points of n image samples;
s2: extracting high-frequency information from the plurality of sample matrixes to correspondingly obtain a plurality of sample models which retain high-frequency characteristics; the method specifically comprises the following steps:
s2.1: performing two-dimensional Fourier transform on the sample matrix to obtain an original frequency spectrum;
s2.2: symmetrically transforming the original frequency spectrum to obtain a ffftshift image;
s2.3: designing a Butterworth filter, and passing the ffftshift image through the Butterworth filter;
s2.4: performing Fourier inverse transformation on the ffftshift image passing through the Butterworth filter to obtain a sample model with high-frequency characteristics reserved;
s3: calculating a correlation coefficient matrix between sample models;
s4: the first K maximum values in the correlation coefficient matrix are intercepted to be used as a simplified correlation coefficient matrix W, and a degree matrix D is obtained according to the simplified correlation coefficient matrix W;
construction of an initial Laplace matrix L 1 ＝D-W；
And normalizing the initial Laplace matrix to obtain a normalized Laplace matrix L=D -1/2 L 1 D -1/2 ；
S5: determining the dimension of the cluster as k 1 ；
Calculating the minimum top k of L 1 The feature vectors corresponding to the feature values are normalized according to the rows to obtain n multiplied by k 1 A feature matrix of the dimension; wherein n is the number of image samples in the sample dataset;
s6: for n x k 1 And clustering the feature matrix of the dimension to obtain a clustering result.
2. The method for extracting and clustering high-frequency information for low-frequency noise face images according to claim 1, wherein in step S1, preprocessing the sample dataset is specifically: filling and restoring each vector in the sample data set into a corresponding sample matrix according to the length and the width of the image sample;
wherein the value of each pixel point in the sample matrix is between 0 and 255.
3. The method of claim 1, further comprising transforming the values of the sample matrix before performing a two-dimensional fourier transform on the sample matrix.
4. The method for extracting and clustering high-frequency information for low-frequency noise face images according to claim 1, wherein the order of the butterworth filter is 2 nd order, and the cut-off frequency is 5% of the width of the image sample.
5. The high-frequency information extraction clustering method for low-frequency noise face images according to claim 1, wherein in step S3, a correlation coefficient matrix between each sample model is determined by:
the sample model is regarded as an energy signal sequence, the correlation between the two sample models is calculated to be the similarity between the two energy signals, the comparison result value is the cross correlation coefficient of the two energy signal sequences, the value is from 0 to 1, and the higher the value is, the more similar the sample result is; and obtaining a correlation coefficient matrix by circularly calculating the mutual correlation coefficients between every two.
6. The method according to claim 1, wherein in step S6, the K-Means algorithm is used for n×k 1 Clustering the feature matrix of the dimension to obtain a clustering resultThe method comprises the steps of carrying out a first treatment on the surface of the The clustering result is cluster division c=c 1 ，C 2 ，...，C k2 The method comprises the steps of carrying out a first treatment on the surface of the Where k2 is the number of clusters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110009835.3A CN112613583B (en) | 2021-01-05 | 2021-01-05 | High-frequency information extraction clustering method for low-frequency noise face image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110009835.3A CN112613583B (en) | 2021-01-05 | 2021-01-05 | High-frequency information extraction clustering method for low-frequency noise face image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112613583A CN112613583A (en) | 2021-04-06 |
| CN112613583B true CN112613583B (en) | 2023-07-21 |
Family
ID=75253318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110009835.3A Active CN112613583B (en) | 2021-01-05 | 2021-01-05 | High-frequency information extraction clustering method for low-frequency noise face image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112613583B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113313394A (en) * | 2021-06-02 | 2021-08-27 | 马鞍山钢铁股份有限公司 | Coke quality evaluation method |
| CN117576493B (en) * | 2024-01-16 | 2024-04-02 | 武汉明炀大数据科技有限公司 | Cloud storage compression method and system for large sample data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563287A (en) * | 2017-06-16 | 2018-01-09 | 江汉大学 | Face identification method and device |
| CN108647726A (en) * | 2018-05-11 | 2018-10-12 | 南京理工大学 | A kind of image clustering method |
| CN109190661A (en) * | 2018-07-26 | 2019-01-11 | 安徽师范大学 | A kind of Spectral Clustering based on difference secret protection |
| CN110349101A (en) * | 2019-06-26 | 2019-10-18 | 上海集成电路研发中心有限公司 | A kind of multiple dimensioned united Wavelet image denoising method |
-
2021
- 2021-01-05 CN CN202110009835.3A patent/CN112613583B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563287A (en) * | 2017-06-16 | 2018-01-09 | 江汉大学 | Face identification method and device |
| CN108647726A (en) * | 2018-05-11 | 2018-10-12 | 南京理工大学 | A kind of image clustering method |
| CN109190661A (en) * | 2018-07-26 | 2019-01-11 | 安徽师范大学 | A kind of Spectral Clustering based on difference secret protection |
| CN110349101A (en) * | 2019-06-26 | 2019-10-18 | 上海集成电路研发中心有限公司 | A kind of multiple dimensioned united Wavelet image denoising method |
Non-Patent Citations (1)
| Title |
|---|
| 融合多尺度统计信息模糊C均值聚类与Markov随机场的小波域声纳图像分割;夏平;任强;吴涛;雷帮军;;兵工学报(05);第111-119页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112613583A (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108805002B (en) | Monitoring video abnormal event detection method based on deep learning and dynamic clustering | |
| Montazer et al. | An improved radial basis function neural network for object image retrieval | |
| CN104268593B (en) | The face identification method of many rarefaction representations under a kind of Small Sample Size | |
| Jalba et al. | Morphological hat-transform scale spaces and their use in pattern classification | |
| CN107977661B (en) | Region-of-interest detection method based on FCN and low-rank sparse decomposition | |
| CN112613583B (en) | High-frequency information extraction clustering method for low-frequency noise face image | |
| Valliammal et al. | Leaf image segmentation based on the combination of wavelet transform and k means clustering | |
| DE112009005002T5 (en) | Techniques for recognizing video copies | |
| CN106803105B (en) | Image classification method based on sparse representation dictionary learning | |
| CN108985346B (en) | Existing exploration image retrieval method fusing low-level image features and CNN features | |
| CN113077452B (en) | Apple tree pest and disease detection method based on DNN network and spot detection algorithm | |
| CN111127407B (en) | Fourier transform-based style migration forged image detection device and method | |
| CN111553417B (en) | Image data dimension reduction method and system based on discriminant regularization local preserving projection | |
| CN116310452B (en) | Multi-view clustering method and system | |
| Dhua et al. | Segmentation of images using density-based algorithms | |
| Zhang et al. | A generative adversarial network with dual discriminators for infrared and visible image fusion based on saliency detection | |
| CN112818779B (en) | Human behavior recognition method based on feature optimization and multiple feature fusion | |
| da Silva et al. | Satellite image segmentation using wavelet transforms based on color and texture features | |
| CN114495037A (en) | Video prediction method and system based on key points and Kalman filtering | |
| Rajesh et al. | Undecimated dual tree complex wavelet transform based face recognition | |
| CN112862700A (en) | Noise attention-based blind denoising method for hyperspectral remote sensing image | |
| CN111160397A (en) | Multi-scale visual dictionary generation method and system | |
| Kaur et al. | No-Reference Blur Assessment Based on Re-Blurring Using Markov Basis. | |
| CN110188608B (en) | Interactive behavior analysis method based on adaptive order | |
| Chavan et al. | A Review on Feature Extraction Methods for Image Analysis |
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 |