CN107742124A - A kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics - Google Patents
A kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics Download PDFInfo
- Publication number
- CN107742124A CN107742124A CN201710868120.7A CN201710868120A CN107742124A CN 107742124 A CN107742124 A CN 107742124A CN 201710868120 A CN201710868120 A CN 201710868120A CN 107742124 A CN107742124 A CN 107742124A
- Authority
- CN
- China
- Prior art keywords
- mrow
- gradient direction
- image
- occurrence matrix
- msub
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/469—Contour-based spatial representations, e.g. vector-coding
- G06V10/473—Contour-based spatial representations, e.g. vector-coding using gradient analysis
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics.This method includes：Predefine one group of shift factor；Gradient magnitude image and gradient direction image are calculated for image to be analyzed；Gradient magnitude image is encoded using local binary patterns algorithm to obtain the gradient magnitude encoded radio of each pixel；To predefined each shift factor, on the basis of gradient direction image, weight computing co-occurrence matrix is encoded to gradient magnitude, obtains weighted gradient direction co-occurrence matrix；All weighted gradient direction co-occurrence matrixs are subjected to vectorization and normalized, obtain weighted gradient direction co-occurrence matrix textural characteristics.The present invention solves the limitation that traditional co-occurrence matrix texture characteristic extracting method is only counted to single image information, realizes the purpose for improving goal description power.
Description
Technical field
The present invention relates to image processing techniques, more particularly to a kind of extraction of weighted gradient direction co-occurrence matrix textural characteristics
Method.
Background technology
Characteristics of image discloses the essential attribute of image, and image feature extraction techniques are always the important of image application field
Research contents.Image characteristics extraction is substantially the process of an exclusion redundancy, and it is successive image segmentation, identification, divided
The premise of the operations such as class, the precision for detecting and identifying in subsequent applications can be effectively improved, and operand can be efficiently reduced,
Improve arithmetic speed.
Texture is a category feature of body surface generally existing, and it reflects the distinctive structural arrangement information of body surface,
There is stronger robustness to brightness and color change.Image texture information occupy in fields such as image procossing, pattern-recognitions and
Its important effect, it has widely in fields such as remote Sensing Image Analysis, medical image analysis, vision-based detection and image retrievals
Using.
Although above thering are many texture analysis methods all to achieve good effect in application, different application field and not
Same image type, specific requirement that analyzing image texture is faced are simultaneously different.Meanwhile the texture with complexity and popularity
Information so that it has certain difficulty in extraction process.Therefore, how to build that a kind of description power is strong, discrimination power is high and
The textural characteristics of robust highly distinguish the difference between of a sort texture image and inhomogeneity texture image, and extraction has
The feature of effect is come to describe texture image be still an extremely challenging problem.
Co-occurrence matrix method is a kind of important statistics class texture characteristic extracting method, and in texture analysis it is the most frequently used,
A kind of method of most study.A kind of statistical method of texture feature extraction earlier occur is proposed by Haralick et al.
Algorithm of co-matrix.Traditional gray level co-occurrence matrixes are mainly by combination condition probability density between image gray levels
Calculate to represent texture.But gray level co-occurrence matrixes statistics be pixel gray value, therefore the algorithm is to picture noise and outer
The illumination variation on boundary is more sensitive, and robustness is not strong.For half-tone information, the gradient information of image is one kind in illumination
Metastable characteristic quantity is remained in that under change, this method for establish co-occurrence matrix on gradient image is gradually closed
Note.However, the existing co-occurrence matrix texture characteristic extracting method based on gradient information is generally only believed single gradient magnitude
Breath or gradient angle information are counted, and do not account for multi information to the superiority in characteristics of image description.
The content of the invention
The technology of the present invention solves problem：Overcome the deficiencies in the prior art, it is proposed that one kind has illumination robustness strong
, the weighted gradient direction co-occurrence matrix texture characteristic extracting method that ga s safety degree is high.
The present invention technical solution be：
A kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics, comprises the following steps：
Step 1: for a selected width image to be analyzed I (x, y), one group of shift factor { (s is predefined1,t1),…,
(sl,tl),…,(sW,tW), wherein, W represents the total number of shift factor, s1,…,sl,…,sWRepresent the picture of image to be analyzed
The displacement of vegetarian refreshments in the horizontal direction, t1,…,tl,…,tWRepresent the displacement of the pixel in the vertical direction of image to be analyzed；
And calculate one group of shift factor { (s1,t1),…,(sl,tl),…,(sW,tW) in each shift factor weighted gradient
Direction co-occurrence matrix, to obtain the co-occurrence matrix set of weighted gradient direction
Step 2: in the weighted gradient direction co-occurrence matrix setIn, will be all
Element in the co-occurrence matrix of weighted gradient direction is sequentially connected in series, to obtain the weighted gradient direction co-occurrence matrix setVector form；
Step 3: to the weighted gradient direction co-occurrence matrix setVector form
It is normalized, to obtain weighted gradient direction co-occurrence matrix textural characteristics.
Further, the normalized uses L2-norm modes.
Further, one group of shift factor { (s of the calculating1,t1),…,(sl,tl),…,(sW,tW) in each position
The weighted gradient direction co-occurrence matrix of the factor is moved, including：
When the shift factor is (sl,tl) when, if the gray value of pixel (x, y) is h in image to be analyzed I (x, y)
(x, y), utilize formulaBy the image to be analyzed I (x, y) be converted to gradient magnitude image G (x,
Y), formula is utilizedThe image to be analyzed I (x, y) is converted into gradient direction image θ (x, y)；Dx=
H (x+1, y)-h (x, y), dy=h (x, y+1)-h (x, y)；
Using local binary patterns algorithm, the gradient magnitude image G (x, y) is converted into gradient magnitude coded image
GLBP(x,y)；
Utilize formulaBy the gradient direction image θ
(x, y) is converted to the gradient direction image al (x, y) of quantization, wherein, K represents the gradient direction quantized value of pixel (x, y)
Maximum；
Utilize formula
, it is (s to calculate the shift factorl,tl) weighted gradient direction co-occurrence matrixWherein, i is represented
The gradient direction quantized value of pixel (x, y), j represent pixel (x+sl,y+tl) gradient direction quantized value, f (GLBP(x,y),
GLBP(x+sl,y+tl)) it is weighting function.
Further, the maximum K=9 of the gradient direction quantized value of the pixel (x, y).
The present invention has the advantages that compared with prior art：
(1), weighted gradient direction co-occurrence matrix textural characteristics proposed by the present invention pass through to gradient magnitude information carry out office
Portion's binary pattern coding, can be good at the partial structurtes feature of phenogram picture, is effectively improved the illumination robustness of feature；
(2), weighted gradient direction co-occurrence matrix textural characteristics proposed by the present invention are by gradient magnitude coding information and gradient
Directional information is combined that to build co-occurrence matrix textural characteristics, to solve traditional co-occurrence matrix texture characteristic extracting method only right
The limitation that single image information is counted.
(3), weighted gradient direction co-occurrence matrix textural characteristics proposed by the present invention consider that single co-occurrence matrix can only be from
The textural characteristics of single dimensional analysis image, can have using different shift factors from multiple dimensioned the characteristics of extracting characteristics of image
Solve the problems, such as that traditional co-occurrence matrix texture characteristic extracting method is limited to goal description power to effect.
Brief description of the drawings
Fig. 1 is a kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics (WGOCM) proposed by the present invention
Flow chart.
Fig. 2 is experimental data set sample image, and the first row is angle cock image, and the second row is non-angle cock image.
Fig. 3 is the influence for analyzing different directions quantized level to weighted gradient direction co-occurrence matrix textural characteristics (WGOCM).
Fig. 4 is the influence for analyzing different weights function pair weighted gradient direction co-occurrence matrix textural characteristics (WGOCM).
Fig. 5 is gradient orientation histogram feature (HOG) compared with weighted gradient direction co-occurrence matrix textural characteristics (WGOCM)
Result.
Fig. 6 is gradient coding histogram feature (GEH) compared with weighted gradient direction co-occurrence matrix textural characteristics (WGOCM)
Result.
Fig. 7 is influence of the different shift factor set of analysis to weighted gradient direction co-occurrence matrix textural characteristics (WGOCM).
Embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is it is understood that described herein
Specific embodiment be used only for explaining the present invention, rather than limitation of the invention.It also should be noted that for the ease of
Describe, part related to the present invention rather than entire infrastructure are illustrate only in accompanying drawing.
Fig. 1 is a kind of flow chart of the extracting method of weighted gradient direction co-occurrence matrix textural characteristics proposed by the present invention.
With reference to figure 1, a kind of extracting method for weighted gradient direction co-occurrence matrix textural characteristics that the embodiment of the present invention proposes, including it is as follows
Step：
Step 1: for a selected width image to be analyzed I (x, y), one group of shift factor { (s is predefined1,t1),…,
(sl,tl),…,(sW,tW), wherein, W represents the total number of shift factor, s1,…,sl,…,sWRepresent the picture of image to be analyzed
The displacement of vegetarian refreshments in the horizontal direction, t1,…,tl,…,tWRepresent the displacement of the pixel in the vertical direction of image to be analyzed；
And calculate one group of shift factor { (s1,t1),…,(sl,tl),…,(sW,tW) in each shift factor weighted gradient
Direction co-occurrence matrix, to obtain the co-occurrence matrix set of weighted gradient directionFurther, institute
Stating the value of shift factor will select according to the characteristic that Texture-period is distributed, for example, for thinner texture, shift factor choosing
Take less value.
Step 2: in the weighted gradient direction co-occurrence matrix setIn, add all
Element in power gradient direction co-occurrence matrix is sequentially connected in series, to obtain the weighted gradient direction co-occurrence matrix setVector form.
Specifically, one group of shift factor { (s of the calculating1,t1),…,(sl,tl),…,(sW,tW) in each displacement
The weighted gradient direction co-occurrence matrix of the factor, including：
When the shift factor is (sl,tl) when, if the gray value of pixel (x, y) is h in image to be analyzed I (x, y)
(x, y), utilize formulaBy the image to be analyzed I (x, y) be converted to gradient magnitude image G (x,
Y), formula is utilizedThe image to be analyzed I (x, y) is converted into gradient direction image θ (x, y)；Dx=h
(x+1, y)-h (x, y), dy=h (x, y+1)-h (x, y)；
Using local binary patterns algorithm, the gradient magnitude image G (x, y) is converted into gradient magnitude coded image
GLBP(x,y)；
Utilize formulaBy the gradient direction image θ
(x, y) is converted to the gradient direction image al (x, y) of quantization, wherein, K represents the gradient direction quantized value of pixel (x, y)
Maximum.Optionally, the maximum K of the gradient direction quantized value of the pixel (x, y) is empirical value, recommendation K=9.
Utilize formula
, it is (s to calculate the shift factorl,tl) weighted gradient direction co-occurrence matrixWherein, i is represented
The gradient direction quantized value of pixel (x, y), j represent pixel (x+sl,y+tl) gradient direction quantized value, f (GLBP(x,y),
GLBP(x+sl,y+tl)) it is weighting function.
Obviously, above-mentioned shift factor (sl,tl) method that calculates weighted gradient direction co-occurrence matrix is applied to { (s1,
t1),…,(sl,tl),…,(sW,tW) in other shift factors when, you can calculate one group of shift factor { (s1,
t1),…,(sl,tl),…,(sW,tW) in each weighted gradient direction co-occurrence matrix to shift factor.
Step 3: to the weighted gradient direction co-occurrence matrix setVector form
It is normalized, to obtain weighted gradient direction co-occurrence matrix textural characteristics.Preferably, the normalized uses
L2-norm modes.
Specifically, it is a characteristic vector to set u, u 2 rank norms are used | | u | |2To represent, note ε be one for 0 it is small
Constant, utilize formulaL2-norm normalizeds are carried out to u, then can obtain normalized vectorial u '.
Embodiment：
Illustrate that weighted gradient direction co-occurrence matrix is calculated below with the example of a specific angle cock of railway wagon classification
The progressive of method.Fig. 2 is experimental data set sample image, and the first row is angle cock image, and the second row is non-angle cock figure
Picture.The present embodiment specifically comprises the following steps：
1st, WGOCM characteristic parameters are determined：
1.1st, shift factor is determined；
In order to analyze influence of the shift factor to WGOCM features, the present embodiment considers three groups of shift factors：(0,1), (0,
1), (1,2), (0,4) }, { (0,1), (0,2), (1,2), (2,3), (0,4) }.
1.2nd, direction quantized level is determined；
For influence of the analysis directions quantized level to WGOCM features, the present embodiment considers three kinds of direction quantized levels.
1.3rd, weighting function is determined；
In order to analyze influence of the weighting function to WGOCM features, the present embodiment considers three kinds of different weighting functions, they
It is respectively：
f(GLBP(x,y),GLBP(x+sl,y+tl))=max (GLBP(x,y),GLBP(x+sl,y+tl))；
f(GLBP(x,y),GLBP(x+sl,y+tl))=min (GLBP(x,y),GLBP(x+sl,y+tl))；
f(GLBP(x,y),GLBP(x+sl,y+tl))=GLBP(x,y)×GLBP(x+sl,y+tl) in.
2nd, training set and test set are built；
The sample data for learning and identifying as grader using the angle cock data set of TFDS system photographs：Including
2186 positive samples and 4800 negative samples, and all sizes are taken as 192 × 128 pixels.Random selection 1093 is just
Sample and 2400 negative samples are as training set, and remaining sample is as test set.
3rd, experiment is classified to the test image by TFDS data sets, we carry out classification of assessment device using ROC curve
Performance；
Fig. 3 analyzes influence of the change to classifier performance of direction quantized level.We consider three kinds of different directions
Quantized level 6,8,9, it can be seen that increase quantized level can effectively improve the performance of grader.Fig. 4 discusses different weights function
Influence to classifier performance.Three kinds of weighting function f (G as shown in table 1LBP(x,y),GLBP(x+sl,y+tl))=max (GLBP
(x,y),GLBP(x+sl,y+tl))、f(GLBP(x,y),GLBP(x+sl,y+tl))=min (GLBP(x,y),GLBP(x+sl,y+tl))、
f(GLBP(x,y),GLBP(x+sl,y+tl))=GLBP(x,y)×GLBP(x+sl,y+tl) in, it can be seen that weighting function f (GLBP
(x,y),GLBP(x+sl,y+tl))=max (GLBP(x,y),GLBP(x+sl,y+tl)) best performance.
Fig. 5 gives HOG and WGOCM result of the comparison, the parameter setting of feature：Direction quantized level is 9, WGOCM's
Shift factor is (0,1), and WGOCM weighting function is f (GLBP(x,y),GLBP(x+sl,y+tl))=max (GLBP(x,y),GLBP
(x+sl,y+tl)).It can be seen that WGOCM classification performance is substantially better than HOG.Fig. 6 gives WGOCM and GEH comparative result,
The parameter setting of feature：Direction quantized level is that 9, WGOCM shift factor takes two groups, be respectively (0,1) and (0,0), (1,
2),(0,4)}.In order to distinguish, shift factor is designated as WGOCM' for the WGOCM features of { (0,0), (1,2), (0,4) }.It can see
It is best to go out the classifying quality for the WGOCM' features that shift factor is { (0,0), (1,2), (0,4) }, shift factor is (0,1)
The classifying quality of WGOCM features takes second place, and GEH classifying quality is worst.As can be seen that the number by increasing co-occurrence matrix, can
Effectively to improve the ga s safety degree of feature.Fig. 7 analyzes influence of the different shift factors combinations to WGOCM features.We with
Machine have selected several different shift factor combinations, as shown in table 1.As can be seen from Figure 7 by increasing the number of co-occurrence matrix
Mesh, it can effectively improve the classification rate of feature.But when the increase of co-occurrence matrix quantity to a certain extent when, the classification rate of feature will
Tend towards stability.
Table 1
By the identification problem of angle cock, we demonstrate the validity of WGOCM features proposed by the invention.To light
According to fairly obvious target identification problem is changed, WGOCM features have all shown excellent ga s safety degree.
The technical scheme of the present embodiment, which provides, a kind of to be combined gradient magnitude coding information with Gradient direction information
Co-occurrence matrix texture characteristic extracting method, solve traditional co-occurrence matrix texture characteristic extracting method and only single image is believed
Cease the limitation counted.On this basis, it is contemplated that single co-occurrence matrix can only be from the line of single dimensional analysis image
Feature is managed, can be from multiple dimensioned the characteristics of extracting characteristics of image using different shift factors, the present invention proposes one kind will be multiple common
Raw matrix is combined into the texture characteristic extracting method of co-occurrence matrix set, efficiently solves traditional co-occurrence matrix textural characteristics and carries
The problem of taking method limited to goal description power.Angle cock of railway wagon classification problem is finally applied to, demonstrates carried side
The validity of method.
Pay attention to, above are only presently preferred embodiments of the present invention and institute's application technology principle.It will be appreciated by those skilled in the art that
The invention is not restricted to specific embodiment described here, can carry out for a person skilled in the art various obvious changes,
Readjust and substitute without departing from protection scope of the present invention.Therefore, although being carried out by above example to the present invention
It is described in further detail, but the present invention is not limited only to above example, without departing from the inventive concept, also
Other more equivalent embodiments can be included, and the scope of the present invention is determined by scope of the appended claims.
Claims (4)
1. a kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics, it is characterised in that comprise the following steps：
Step 1: for a width image to be analyzed I (x, y), one group of shift factor { (s is predefined1,t1),…,(sl,tl),…,
(sW,tW), wherein, W represents the total number of shift factor, s1,…,sl,…,sWRepresent the pixel of image to be analyzed in level
Displacement on direction, t1,…,tl,…,tWRepresent the displacement of the pixel in the vertical direction of image to be analyzed；And described in calculating
One group of shift factor { (s1,t1),…,(sl,tl),…,(sW,tW) in each shift factor weighted gradient direction symbiosis square
Battle array, to obtain the co-occurrence matrix set of weighted gradient direction
Step 2: in the weighted gradient direction co-occurrence matrix setIn, by all weighting ladders
Element in degree direction co-occurrence matrix is sequentially connected in series, to obtain the weighted gradient direction co-occurrence matrix setVector form；
Step 3: to the weighted gradient direction co-occurrence matrix setVector form carry out
Normalized, to obtain weighted gradient direction co-occurrence matrix textural characteristics.
2. the extracting method of weighted gradient direction co-occurrence matrix textural characteristics according to claim 1, it is characterised in that institute
State normalized and use L2-norm modes.
3. the extracting method of weighted gradient direction co-occurrence matrix textural characteristics according to claim 1, it is characterised in that institute
State and calculate one group of shift factor { (s1,t1),…,(sl,tl),…,(sW,tW) in each shift factor weighted gradient direction
Co-occurrence matrix, including：
When the shift factor is (sl,tl) when, if the gray value of pixel (x, y) is h (x, y) in image to be analyzed I (x, y),
Utilize formulaThe image to be analyzed I (x, y) is converted into gradient magnitude image G (x, y), utilized
FormulaThe image to be analyzed I (x, y) is converted into gradient direction image θ (x, y)；Dx=h (x+1,
Y)-h (x, y), dy=h (x, y+1)-h (x, y)；
Using local binary patterns algorithm, the gradient magnitude image G (x, y) is converted into gradient magnitude coded image GLBP(x,
y)；
Utilize formulaBy the gradient direction image θ (x, y)
The gradient direction image al (x, y) of quantization is converted to, wherein, K represents the maximum of the gradient direction quantized value of pixel (x, y)
Value；
Utilize formula
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>s</mi>
<mi>l</mi>
</msub>
<mo>,</mo>
<msub>
<mi>t</mi>
<mi>l</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mi>x</mi>
</munder>
<munder>
<mo>&Sigma;</mo>
<mi>y</mi>
</munder>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>G</mi>
<mrow>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
</mrow>
</msub>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
</mrow>
<mo>)</mo>
<mo>,</mo>
<msub>
<mi>G</mi>
<mrow>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
</mrow>
</msub>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mo>+</mo>
<msub>
<mi>s</mi>
<mi>l</mi>
</msub>
<mo>,</mo>
<mi>y</mi>
<mo>+</mo>
<msub>
<mi>t</mi>
<mi>l</mi>
</msub>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mi>i</mi>
<mi>f</mi>
<mi> </mi>
<mi>a</mi>
<mi>l</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>i</mi>
<mi> </mi>
<mi>a</mi>
<mi>n</mi>
<mi>d</mi>
<mi> </mi>
<mi>a</mi>
<mi>l</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>+</mo>
<msub>
<mi>s</mi>
<mi>l</mi>
</msub>
<mo>,</mo>
<mi>y</mi>
<mo>+</mo>
<msub>
<mi>t</mi>
<mi>l</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>j</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
<mi>o</mi>
<mi>t</mi>
<mi>h</mi>
<mi>e</mi>
<mi>r</mi>
<mi>w</mi>
<mi>i</mi>
<mi>s</mi>
<mi>e</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
,
It is (s to calculate the shift factorl,tl) weighted gradient direction co-occurrence matrixWherein, i represents pixel
The gradient direction quantized value of (x, y), j represent pixel (x+sl,y+tl) gradient direction quantized value, f (GLBP(x,y),GLBP(x+
sl,y+tl)) it is weighting function.
4. the extracting method of weighted gradient direction co-occurrence matrix textural characteristics according to claim 3, it is characterised in that institute
State the maximum K=9 of the gradient direction quantized value of pixel (x, y).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710868120.7A CN107742124A (en) | 2017-09-22 | 2017-09-22 | A kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710868120.7A CN107742124A (en) | 2017-09-22 | 2017-09-22 | A kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107742124A true CN107742124A (en) | 2018-02-27 |
Family
ID=61236148
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710868120.7A Pending CN107742124A (en) | 2017-09-22 | 2017-09-22 | A kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107742124A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109754390A (en) * | 2018-12-11 | 2019-05-14 | 西北大学 | A kind of non-reference picture quality appraisement method based on mixing visual signature |
| CN111028210A (en) * | 2019-11-25 | 2020-04-17 | 北京航天控制仪器研究所 | Deep neural network glass tube end surface defect detection method based on sparse automatic encoder |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567735A (en) * | 2010-12-30 | 2012-07-11 | 中国科学院电子学研究所 | Method for automatically picking up control point sections of remote sensing images |
| CN106778833A (en) * | 2016-11-28 | 2017-05-31 | 北京航天控制仪器研究所 | Small object loses the automatic identifying method of failure under a kind of complex background |
-
2017
- 2017-09-22 CN CN201710868120.7A patent/CN107742124A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567735A (en) * | 2010-12-30 | 2012-07-11 | 中国科学院电子学研究所 | Method for automatically picking up control point sections of remote sensing images |
| CN106778833A (en) * | 2016-11-28 | 2017-05-31 | 北京航天控制仪器研究所 | Small object loses the automatic identifying method of failure under a kind of complex background |
Non-Patent Citations (6)
| Title |
|---|
| FERNANDO ROBERTI DE SIQUEIRA 等: "Multi-scale gray level co-occurrence matrices for texture description", 《NEUROCOMPUTING》 * |
| LIU LIU 等: "Automated Visual Inspection System for Bogie Block Key Under Complex Freight Train Environment", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
| LORIS NANNI 等: "Improving the descriptors extracted from the co-occurrence matrix using preprocessing approaches", 《EXPERT SYSTEMS WITH APPLICATIONS》 * |
| 吴刚 等: "融合典型纹理特征的粒子滤波目标跟踪方法", 《计算机工程与应用》 * |
| 吴志斌: "基于视频序列分析的乳腺癌良恶性方法的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 张波: "基于机器视觉的复杂工况下驾驶人疲劳状态检测方法研究", 《中国博士学位论文全文数据库 工程科技II辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109754390A (en) * | 2018-12-11 | 2019-05-14 | 西北大学 | A kind of non-reference picture quality appraisement method based on mixing visual signature |
| CN109754390B (en) * | 2018-12-11 | 2023-04-07 | 西北大学 | No-reference image quality evaluation method based on mixed visual features |
| CN111028210A (en) * | 2019-11-25 | 2020-04-17 | 北京航天控制仪器研究所 | Deep neural network glass tube end surface defect detection method based on sparse automatic encoder |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106845478B (en) | A kind of secondary licence plate recognition method and device of character confidence level | |
| Babu et al. | Statistical features based optimized technique for copy move forgery detection | |
| CN104778457B (en) | Video face identification method based on multi-instance learning | |
| CN104573729B (en) | A kind of image classification method based on core principle component analysis network | |
| CN102779273B (en) | A kind of face identification method based on local contrast pattern | |
| Walia et al. | Fusion of handcrafted and deep features for forgery detection in digital images | |
| CN107103317A (en) | Fuzzy license plate image recognition algorithm based on image co-registration and blind deconvolution | |
| CN105488536A (en) | Agricultural pest image recognition method based on multi-feature deep learning technology | |
| CN108596154A (en) | Classifying Method in Remote Sensing Image based on high dimensional feature selection and multi-level fusion | |
| CN103679187B (en) | Image-recognizing method and system | |
| Ye et al. | Detecting USM image sharpening by using CNN | |
| CN102663436A (en) | Self-adapting characteristic extracting method for optical texture images and synthetic aperture radar (SAR) images | |
| CN107092876A (en) | The low-light (level) model recognizing method combined based on Retinex with S SIFT features | |
| CN106971158A (en) | A kind of pedestrian detection method based on CoLBP symbiosis feature Yu GSS features | |
| CN104778475A (en) | Image classification method based on maximum frequent visual word of annular region | |
| CN111832650A (en) | Image classification method based on generation of confrontation network local aggregation coding semi-supervision | |
| CN109614866A (en) | Method for detecting human face based on cascade deep convolutional neural networks | |
| CN111783885A (en) | Millimeter wave image quality classification model construction method based on local enhancement | |
| Satpathy et al. | Extended histogram of gradients feature for human detection | |
| CN107742124A (en) | A kind of extracting method of weighted gradient direction co-occurrence matrix textural characteristics | |
| Sah et al. | Text and non-text recognition using modified HOG descriptor | |
| CN107358244B (en) | A kind of quick local invariant feature extracts and description method | |
| CN103902965B (en) | Spatial domain symbiosis image representing method and its application in image classification, identification | |
| Jingyi et al. | Classification of images by using TensorFlow | |
| CN105512682B (en) | A kind of security level identification recognition methods based on Krawtchouk square and KNN-SMO classifier |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180227 |
|
| RJ01 | Rejection of invention patent application after publication |