CN111291805A - Color texture image classification method based on complete extreme value non-negative dense micro-block difference - Google Patents

Color texture image classification method based on complete extreme value non-negative dense micro-block difference Download PDF

Info

Publication number
CN111291805A
CN111291805A CN202010076384.0A CN202010076384A CN111291805A CN 111291805 A CN111291805 A CN 111291805A CN 202010076384 A CN202010076384 A CN 202010076384A CN 111291805 A CN111291805 A CN 111291805A
Authority
CN
China
Prior art keywords
color
channel
texture image
difference
channels
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
CN202010076384.0A
Other languages
Chinese (zh)
Other versions
CN111291805B (en
Inventor
董永生
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.)
Henan University of Science and Technology
Original Assignee
Henan University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Henan University of Science and Technology filed Critical Henan University of Science and Technology
Priority to CN202010076384.0A priority Critical patent/CN111291805B/en
Publication of CN111291805A publication Critical patent/CN111291805A/en
Application granted granted Critical
Publication of CN111291805B publication Critical patent/CN111291805B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features

Abstract

The invention discloses a color texture image classification method based on complete extreme value non-negative dense micro-block difference, which relates to computer vision and mode identification, and has the beneficial effects that the method builds a color texture descriptor for color texture classification by modeling the characteristics in the color channels of the color texture image and the characteristics among the color channels and fusing the local characteristics and the global characteristics of color textures: the classification method of the present invention is effective and excellent as compared with other representative methods as proved by experiments.

Description

Color texture image classification method based on complete extreme value non-negative dense micro-block difference
Technical Field
The invention belongs to the technical field of computer vision and pattern recognition, and particularly relates to a color texture image classification method based on complete extremum non-negative dense micro-block difference.
Background
As a classical problem in the field of computer vision and pattern recognition, the texture representation method is an important step in the texture classification process. Most texture representation methods focus on gray-scale texture images, ignoring color information. However, in the real world, color is also information having discriminative power. It is of great significance to construct a representation of a color texture image.
The method for expressing the color texture is constructed, intuitively, the method for expressing the gray texture can be applied to each color channel of the color texture, and finally, the characteristics of the three channels are fused. However, each color channel in the color texture image is not independent, and therefore, information between color channels needs to be captured. At present, most color texture representation methods cannot simultaneously explore information in and between color channels, and also lack global features of color textures.
Based on the problem, a texture classification method based on complete extreme value non-negative dense micro-block difference is constructed, in the process of representing color textures, the color channel internal features and the color channel inter-features after dense micro-block difference modeling are fused through non-negative and extreme value operations, and then the color textures and the global features of a color texture image are combined together to form complete extreme value non-negative dense micro-block difference (CEN-DMD) and used for color texture classification.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a color texture image classification method based on complete extremum non-negative dense micro-block difference, and solve the problems existing in the existing color texture representation method.
The technical scheme adopted by the invention for solving the technical problems is as follows: the color texture image classification method based on complete extremum non-negative dense micro-block difference comprises the following steps:
the method comprises the following steps: modeling the color intra-channel features and the color inter-channel features of the color texture image: firstly, for a given color texture image, dividing the color texture image into image blocks with the size of L x L according to the step size of 2 pixels; secondly, sampling micro blocks in the image block, extracting texture features by calculating the difference of the average values of pixels between the micro blocks in pairs, modeling features in a color channel of the color texture image, and selecting the micro block pairs for difference from the same color channel; for inter-color-channel modeling of color texture images, then the pair of tiles used for the difference is selected from within the opposite color channel, while non-negative operations are used in the tile difference process;
step two: after modeling a color texture image, capturing difference features in color channels and difference features among the color channels, fusing the maximum value of the difference features in the channels and the maximum value of the difference features among the channels, and constructing an extreme value non-negative dense micro-block difference descriptor EN-DMD, wherein the definition of the EN-DMD is as follows:
Figure BDA0002378593110000021
where max () is the max operation, u, v, w represent color channels, IpRepresents a texture image patch, p represents patch,
Figure BDA0002378593110000022
represents IpThe difference of the tiles within the color channel u,
Figure BDA0002378593110000023
represents IpThe difference of the tiles within a color channel v,
Figure BDA0002378593110000024
represents IpThe difference of the tiles within a color channel w,
Figure BDA0002378593110000025
represents IpThe difference in the tiles between color channel u and color channel v,
Figure BDA0002378593110000026
represents IpThe difference in the tiles between color channel v and color channel w,
Figure BDA0002378593110000027
represents IpDifferential microblock between color channel w and color channel u;
step three: the method for constructing the complete extreme value non-negative dense micro-block differential CEN-DMD by fusing the global characteristics of the color textures comprises the following steps:
(1) the global features of the color texture image include global features within a color channel and global features between color channels, I for the color texture imageu,IvandIwRespectively representing image information of three color channels, wherein the mean value and the variance of pixels in each color channel are used as global features in the color channels, and the global features among the color channels are captured through the mean value and the variance of differences among the color channels of the texture image;
(2) global feature G in color channel by combining extreme non-negative microblock differential EN-DMD with color texture imageIntraAnd colorInter-channel global feature GInterAnd constructing a complete extreme value non-negative dense micro block differential CEN-DMD: CEN-DMD [ -EN-DMD, GIntra，GInter]；
Step four: the CEN-DMD is Fisher vector encoded to obtain a color texture descriptor, and finally, a linear SVM classifier is used to perform color texture image classification.
The method for modeling the characteristics in the color channel of the color texture image in the first step comprises the following steps: for image blocks of size L on color channel c
Figure BDA0002378593110000028
Color in-channel features of
Figure BDA0002378593110000029
The calculation is as follows:
Figure BDA00023785931100000210
wherein the content of the first and second substances,
Figure BDA00023785931100000211
wherein s represents the size of the micro-tile,
Figure BDA0002378593110000031
representing a coordinate x on a particular color channelc＝(a，b)TAverage pixels of the tile,/denotes absolute value operation, c ∈ { u, v, w } represents a color channel.
The method for modeling the characteristics among the color channels of the color texture image in the first step of the invention comprises the following steps: for a color texture image block of size L x L, the features between the color channels between color channels c and c
Figure BDA0002378593110000032
The calculation is as follows:
Figure BDA0002378593110000033
wherein the content of the first and second substances,
Figure BDA0002378593110000034
represents IpThe difference in microblocks between color channels between color channel c and color channel c ', c ∈ { u, v, w } and c' ∈ { u, v, w } representing color channels, the set of sampling points XcAnd YcDistributed by t
Figure BDA0002378593110000035
And (4) generating.
The second step of the invention is that the global characteristic G in the color channel of the color texture imageIntraIs defined as follows:
GIntra＝[avg(Iu)，avg(Iv)，avg(Iw)，var(Iu)，var(Iv)，var(Iw)]，
wherein, avg (I)u)，avg(Iv)，avg(Iw) Respectively representing the mean value, var (I), of the pixels in a single color channel of the color texture imageu)，var(Iv)，var(Iw) Respectively representing the variance of a pixel in a single color channel of the color texture image.
Global feature G between color channels in step two of the present inventionInterIs defined as follows:
GInter＝[avg(|Iu-Iv|)，avg(|Iv-Iw|)，avg(|Iw-Iu|)， var(|Iu-Iv|)，var(|Iv-Iw|)，var(|Iw-Iu|)]，
wherein, avg (| I)u-Iv|)，avg(|Iv-Iw|)，avg(|Iw-Iu|) mean of the differences between the color channels of the respective color texture images, var (| I)u-Iv|)，var(|Iv-Iw|)，var(|Iw-Iu|) the variance of the difference between the color channels of the color texture image, u, v, w represent the three color channels of the color texture image, respectively.
The invention has the beneficial effects that: firstly, constructing a non-negative multi-resolution DMD characteristic for representing texture information, wherein the multi-resolution of the characteristic represents that micro-block differential characteristics with different scales are connected after non-negative operation; secondly, the invention measures the significance of the difference features by using the maximum value of the difference features in the color of the color texture image and the maximum value of the difference features between channels; thirdly, the invention proposes a complete extreme non-negative dense difference (CEN-DMD) for representing color texture, wherein the integrity of the CEN-DMD includes features in color channels and features between color channels, and includes local features and global features of the color texture; fourth, the classification results on five published color texture datasets (CURET, Colored Brodatz, VisTex, USPTex and KTH-TIPS) demonstrate that the classification method of the present invention is efficient and superior compared to other representative methods.
Drawings
FIG. 1 is a diagram of the color intra-channel features and inter-channel features of a dense micro-block differential modeled color texture image in accordance with the present invention;
FIG. 2 is a flow chart of a complete extremum non-negative dense microblock differential CEN-DMD method of the present invention;
FIG. 3 is a comparison graph of classification accuracy of CEN-DMD of the present invention in three different color spaces on three texture data sets;
FIG. 4 is a comparison graph of classification accuracy of the CEN-DMD single scale representation and multi-scale representation of the present invention on four texture datasets.
Detailed Description
The following description of specific embodiments (examples) of the present invention are provided in connection with examples to enable those skilled in the art to better understand the present invention.
The method constructs the color texture descriptor for color texture classification by modeling the features in the color channels and the features among the color channels of the color texture image and fusing the local features and the global features of the color textures. First in the image block, dense micro-block differencing is used to model the intra-color-channel and inter-color-channel features of the color texture, and non-negative operations are used in the differencing process. Then, the maximum value of the difference within the color channel and the maximum value of the difference between the color channels are fused to realize that the features within the color channel and between the color channels can be captured simultaneously. Furthermore, multi-resolution representation of color textures is achieved by connecting differential features on different scale tiles. Meanwhile, the complete extreme value non-negative dense micro-block difference (CEN-DMD) is constructed by fusion with the global characteristics of the color texture image. The Fisher vector is used to encode the CEN-DMD to obtain the color texture descriptor. Finally, a linear SVM classifier is used for testing the classification performance of the proposed method, and a large number of experimental results verify the effectiveness of the method.
The method comprises the following steps: modeling color intra-channel and inter-color-channel features of a color texture image
First, for a given color texture image it is divided into image blocks of size L x L in 2 pixel steps. Next, the image block is subjected to micro-block sampling, and texture features are extracted by calculating the difference in average values of pixels between pairs of micro-blocks. Modeling features in color channels of the color texture image, wherein the micro-block pairs used for difference are all selected from the same color channel; modeling between color channels of a color texture image, the pair of tiles used for differencing is selected from within the opposite color channel, while non-negative operations are used in the tile differencing process. For each color channel of the color texture image block, the selection of the micro-block pairs is performed by sampling point sets
Figure BDA0002378593110000051
And
Figure BDA0002378593110000052
where n represents the number of sample points and c ∈ { u, v, w) represents a color channel.
For image blocks of size L on color channel c
Figure BDA0002378593110000053
The inter-color-channel feature of (a) is calculated as follows:
Figure BDA0002378593110000054
wherein
Figure BDA0002378593110000055
Wherein s represents the size of the micro-tile,
Figure BDA0002378593110000056
representing a coordinate x on a particular color channelc＝(a，b)TAverage pixel of the tile,/denotes absolute value operation, c ∈ { u, v, w } denotes color channel, IpRepresents a texture image block patch, which is represented by the initials p for simplicity.
For a color texture image block of size L × L, the features between color channels c and c' are calculated as follows:
Figure BDA0002378593110000057
wherein, IpRepresents a texture image block patch, which, for the sake of simplicity, is represented by the initials p,
Figure BDA0002378593110000058
represents IpThe difference in microblocks between color channels between color channel c and color channel c ', c ∈ { u, v, w } and c' ∈ { u, v, w } representing color channels. Set of sampling points XcAnd YcDistributed by t
Figure BDA0002378593110000059
And (4) generating.
Step two: constructing extreme non-negative dense micro-block differential descriptor (extreme non-negative DMD, EN-DMD)
After modeling the color texture image, the difference features within the color channels and the difference features between the color channels may be captured. Fusing the maximum value of the difference characteristic in the channel and the maximum value of the difference characteristic between the channels to construct an extreme non-negative dense micro-block difference descriptor (EN-DMD), wherein the EN-DMD is defined as follows:
Figure BDA00023785931100000510
where max () is the max operation, u, v, w represent color channels, IpRepresents a texture image patch, p represents patch,
Figure BDA0002378593110000061
represents IpThe difference of the tiles within the color channel u,
Figure BDA0002378593110000062
represents IpThe difference of the tiles within a color channel v,
Figure BDA0002378593110000063
represents IpThe difference of the tiles within a color channel w,
Figure BDA0002378593110000064
represents IpThe difference in the tiles between color channel u and color channel v,
Figure BDA0002378593110000065
represents IpThe difference in the tiles between color channel v and color channel w,
Figure BDA0002378593110000066
represents IpThe difference in tiles between color channel w and color channel u.
Step three: and (3) constructing complete extreme value non-negative dense micro-block difference (CEN-DMD) by fusing global features of the color textures: the global features of the color texture image include global features within a color channel and global features between color channels, I for the color texture imageu,Ivand IwRepresenting the image information of three color channels respectively, the mean and variance of the pixels in each color channel being used as the global feature, color, in the color channelColor intra-channel global feature G of color texture imageIntraIs defined as follows:
GIntra＝[avg(Iu)，avg(Iv)，avg(Iw)，var(Iu)，var(Iv)，var(Iw)]
wherein, avg (I)u)，avg(Iv)，avg(Iw) Respectively representing the mean value, var (I), of the pixels in a single color channel of the color texture imageu)，var(Iv)，var(Iw) Respectively representing the variance of a pixel in a single color channel of the color texture image.
The inter-color-channel global features of the color texture image are captured by means of the mean and variance of the differences between the color channels of the texture image. Inter-color-channel global feature GInterIs defined as follows:
GInter＝[avg(|Iu-Iv|)，avg(|Iv-Iw|)，avg(Iw-Iu|)， var(|Iu-Iv|)，var(|Iv-Iw|)，var(|Iw-Iu|)]
wherein, avg (| I)u-Iv|)，avg(|Iv-Iw|)，avg(|Iw-Iu|) mean of the differences between the color channels of the respective color texture images, var (| I)u-Iv|)，var{|Iv-Iw|)，var(|Iw-Iu|) variance of differences between color channels of the respective color texture images. u, v, w represent the three color channels of the color texture image, respectively.
By combining extreme non-negative microblock differences (EN-DMD) with intra-color-channel and inter-color-channel global features of a color texture image, complete extreme non-negative dense microblock differences (CEN-DMD) are constructed, which are defined as follows:
CEN-DMD＝[EN-DMD，GIntra，GInter]；
step four: and encoding the CEN-DMD by a Fisher vector to obtain a color texture descriptor. Finally, a linear SVM classifier is used to perform color texture image classification.
Simulation classification experiment
The method is used for classifying color texture images based on complete extreme value non-negative dense micro-block difference, and classification experiments are carried out on five public texture data sets including CURET, ColoredBrodatz, VisTex, USPTex and KTH-TIPS, wherein the color space selects RGB color space, the number of sampling points n is 80, the size L of an image block is 15, and the size s of a micro-block is represented in a multi-resolution mode from 1 to 5. The specific experiment is as follows:
(1) texture data set: CURET: including 61 texture classes, we used 46 samples for training and 46 individual samples for testing. Colored Brodatz: including 40 texture classes, each 640 x 640 texture class was divided into 16 160 x 160 non-overlapping samples, 8 for training and 8 for testing. VisTex: containing 40 color texture classes, each 512 x 512 texture image was divided into 16 128 x 128 non-overlapping samples, 8 for training and 8 for testing. USPTex: consisting of 191 texture classes, each class having 12 samples, 6 samples for training and 6 samples for testing. KTH-TIPS texture images containing different illumination angles and different scales are composed of 10 texture classes, each class has 81 samples, 40 of them are used for training and 41 are used for testing.
(2) Investigation of parameter sensitivity: classification experiments were performed on five standard texture datasets in the three RGB, HSV and YCbCr spaces, respectively, and the experimental results are shown in fig. 3, with CEN-DMD performing best in the RGB color space.
(3) Multi-resolution experiments: comparative experiments on CEN-DMD single and multi-scale were performed on CURET, VisTex, USPTex and KTH-TIPS texture datasets, respectively. The size of the image blocks is 15 x 15. At single scale, the microblock size is 5 x 5. In multi-scale, the microblock size is from 1 x 1 to 5 x 5. That is, the total of 80 samples is 16 samples per microblock scale. Experimental results as shown in fig. 4, the classification accuracy of the multi-scale is higher than that of the single scale.
Comparative experiment: comparative experiments were performed on five standard texture datasets (KTH-TIPS, VisTex, CURET, USPTex, ColoredBrodatz) with 13 representative color texture classification methods, the results of which are shown in Table 1. The experimental result shows that the classification method of the Qiyou is higher in classification precision.
Table 114 classification precision table of color texture representation method on five texture data sets
Figure BDA0002378593110000081

Claims (5)

1. The color texture image classification method based on the complete extremum nonnegative dense micro-block difference is characterized by comprising the following steps of: the method comprises the following steps:
the method comprises the following steps: modeling the color intra-channel features and the color inter-channel features of the color texture image: firstly, for a given color texture image, dividing the color texture image into image blocks with the size of L x L according to the step size of 2 pixels; secondly, sampling micro blocks in the image block, extracting texture features by calculating the difference of the average values of pixels between the micro blocks in pairs, modeling features in a color channel of the color texture image, and selecting the micro block pairs for difference from the same color channel; for inter-color-channel modeling of color texture images, then the pair of tiles used for the difference is selected from within the opposite color channel, while non-negative operations are used in the tile difference process;
step two: after modeling a color texture image, capturing difference features in color channels and difference features among the color channels, fusing the maximum value of the difference features in the channels and the maximum value of the difference features among the channels, and constructing an extreme value non-negative dense micro-block difference descriptor EN-DMD, wherein the definition of the EN-DMD is as follows:
Figure FDA0002378593100000011
where max () is the max operation, u, v, w represent color channels, IpRepresents a texture image patch, p represents patch,
Figure FDA0002378593100000012
represents IpThe difference of the tiles within the color channel u,
Figure FDA0002378593100000013
represents IpThe difference of the tiles within a color channel v,
Figure FDA0002378593100000014
represents IpThe difference of the tiles within a color channel w,
Figure FDA0002378593100000015
represents IpThe difference in the tiles between color channel u and color channel v,
Figure FDA0002378593100000016
represents IpThe difference in the tiles between color channel v and color channel w,
Figure FDA0002378593100000017
represents IpDifferential microblock between color channel w and color channel u;
step three: the method for constructing the complete extreme value non-negative dense micro-block differential CEN-DMD by fusing the global characteristics of the color textures comprises the following steps:
(1) the global features of the color texture image include global features within a color channel and global features between color channels, I for the color texture imageu,Ivand IwRespectively representing image information of three color channels, wherein the mean value and the variance of pixels in each color channel are used as global features in the color channels, and the global features among the color channels are captured through the mean value and the variance of differences among the color channels of the texture image;
(2) global feature G in color channel by combining extreme non-negative microblock differential EN-DMD with color texture imageIntraAnd global feature G between color channelsInterAnd constructing a complete extreme value non-negative dense micro block differential CEN-DMD: CEN-DMD [ -EN-DMD, GIntra，GInter]；
Step four: the CEN-DMD is Fisher vector encoded to obtain a color texture descriptor, and finally, a linear SVM classifier is used to perform color texture image classification.
2. The method of claim 1, wherein the method comprises: in the first step, the color channel internal feature modeling method of the color texture image comprises the following steps: for image blocks of size L on color channel c
Figure FDA0002378593100000021
Color in-channel features of
Figure FDA0002378593100000022
The calculation is as follows:
Figure FDA0002378593100000023
wherein the content of the first and second substances,
Figure FDA0002378593100000024
wherein s represents the size of the micro-tile,
Figure FDA0002378593100000025
representing a coordinate x on a particular color channelc＝(a，b)TAverage pixels of the tile,/denotes absolute value operation, c ∈ { u, v, w } represents a color channel.
3. The method of claim 1, wherein the method comprises: in the first step, the method for modeling the characteristics among the color channels of the color texture image comprises the following steps: for a color texture image block of size L x L, the features between the color channels between color channels c and c
Figure FDA0002378593100000026
The calculation is as follows:
Figure FDA0002378593100000027
wherein the content of the first and second substances,
Figure FDA0002378593100000028
represents IpInter-color-channel microblocks between color channel c and color channel c ', c ∈ { u, v, w } and c' ∈ { u, v, w } representing color channels, the set of sampling points XcAnd YcDistributed by t
Figure FDA0002378593100000029
And (4) generating.
4. The method of claim 1, wherein the method comprises: the second step is that the global characteristic G in the color channel of the color texture imageIntraIs defined as follows:
CIntra＝[avg(Iu)，avg(Iv)，avg(Iw)，var(Iu)，var(Iv)，var(Iw)，
wherein, avg (I)u)，avg(Iv)，avg(Iw) Respectively representing the mean value, var (I), of the pixels in a single color channel of the color texture imageu)，var(Iv)，var(Iw) Respectively representing the variance of a pixel in a single color channel of the color texture image.
5. The method of claim 1, wherein the method comprises: in the second step, global characteristics G between color channelsInterIs defined as follows:
GInter＝[avg(|Iu-Iv|)，avg(|Iv-Iw|)，avg(|Iw-Iu|)，var(|Iu-Iv|)，var(|Iv-Iw|)，var|Iw-Iu|)，
wherein, avg (| I)u-Iv|)，avg(|Iv-Iw|)，avg(|Iw-Iu|) mean of the differences between the color channels of the respective color texture images, var (| I)u-Iv|)，var(|Iv-Iw|)，var(|Iw-Iu|) the variance of the difference between the color channels of the color texture image, u, v, w represent the three color channels of the color texture image, respectively.
CN202010076384.0A 2020-01-23 2020-01-23 Color texture image classification method based on complete extreme value non-negative dense micro-block difference Active CN111291805B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010076384.0A CN111291805B (en) 2020-01-23 2020-01-23 Color texture image classification method based on complete extreme value non-negative dense micro-block difference

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010076384.0A CN111291805B (en) 2020-01-23 2020-01-23 Color texture image classification method based on complete extreme value non-negative dense micro-block difference

Publications (2)

Publication Number Publication Date
CN111291805A true CN111291805A (en) 2020-06-16
CN111291805B CN111291805B (en) 2023-03-24

Family

ID=71023369

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010076384.0A Active CN111291805B (en) 2020-01-23 2020-01-23 Color texture image classification method based on complete extreme value non-negative dense micro-block difference

Country Status (1)

Country Link
CN (1) CN111291805B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2435160A1 (en) * 2002-09-12 2004-03-12 Michael Schroder Texture-based colour correction
US20120155766A1 (en) * 2010-12-17 2012-06-21 Sony Corporation Patch description and modeling for image subscene recognition
CN109740572A (en) * 2019-01-23 2019-05-10 浙江理工大学 A kind of human face in-vivo detection method based on partial color textural characteristics
CN110490210A (en) * 2019-08-23 2019-11-22 河南科技大学 A kind of color texture classification method based on compact interchannel t sample differential

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2435160A1 (en) * 2002-09-12 2004-03-12 Michael Schroder Texture-based colour correction
US20120155766A1 (en) * 2010-12-17 2012-06-21 Sony Corporation Patch description and modeling for image subscene recognition
CN109740572A (en) * 2019-01-23 2019-05-10 浙江理工大学 A kind of human face in-vivo detection method based on partial color textural characteristics
CN110490210A (en) * 2019-08-23 2019-11-22 河南科技大学 A kind of color texture classification method based on compact interchannel t sample differential

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
肖淑苹等: "基于小波变换和支持向量机的彩色纹理识别", 《微电子学与计算机》 *
郝连旺等: "基于多颜色空间特征融合的彩色白细胞图像识别", 《生物医学工程学杂志》 *

Also Published As

Publication number Publication date
CN111291805B (en) 2023-03-24

Similar Documents

Publication Publication Date Title
CN108492343B (en) Image synthesis method for training data for expanding target recognition
CN108280397B (en) Human body image hair detection method based on deep convolutional neural network
CN109410168B (en) Modeling method of convolutional neural network for determining sub-tile classes in an image
Rathore et al. Colour based image segmentation using L* a* b* colour space based on genetic algorithm
CN110717526A (en) Unsupervised transfer learning method based on graph convolution network
CN105893925A (en) Human hand detection method based on complexion and device
Kadam et al. Detection and localization of multiple image splicing using MobileNet V1
CN109447123B (en) Pedestrian re-identification method based on label consistency constraint and stretching regularization dictionary learning
CN107808129A (en) A kind of facial multi-characteristic points localization method based on single convolutional neural networks
CN111652273B (en) Deep learning-based RGB-D image classification method
CN108921120B (en) Cigarette identification method suitable for wide retail scene
WO2021082168A1 (en) Method for matching specific target object in scene image
CN107527054B (en) Automatic foreground extraction method based on multi-view fusion
CN112101262B (en) Multi-feature fusion sign language recognition method and network model
CN106355607B (en) A kind of width baseline color image template matching method
JP4553300B2 (en) Content identification device
CN112163401A (en) Chinese character font generation method based on compressed and excited GAN network
CN111507357A (en) Defect detection semantic segmentation model modeling method, device, medium and equipment
CN106203448B (en) A kind of scene classification method based on Nonlinear Scale Space Theory
CN110827304A (en) Traditional Chinese medicine tongue image positioning method and system based on deep convolutional network and level set method
CN108491863A (en) Color image processing method based on Non-negative Matrix Factorization and convolutional neural networks
CN103839074A (en) Image classification method based on matching of sketch line segment information and space pyramid
CN110490210B (en) Color texture classification method based on t sampling difference between compact channels
CN107103325B (en) Tissue pathology image classification method
CN111291805B (en) Color texture image classification method based on complete extreme value non-negative dense micro-block difference

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