CN104463814B - Image enhancement method based on local texture directionality - Google Patents
Image enhancement method based on local texture directionality Download PDFInfo
- Publication number
- CN104463814B CN104463814B CN201410741870.4A CN201410741870A CN104463814B CN 104463814 B CN104463814 B CN 104463814B CN 201410741870 A CN201410741870 A CN 201410741870A CN 104463814 B CN104463814 B CN 104463814B
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- texture
- feature
- local
- 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
Landscapes
- Image Processing (AREA)
Abstract
The invention provides an image enhancement method based on the local texture directionality. The method comprises the two steps of local texture direction judgment and local texture image enhancement, wherein in the local texture direction, an image skeleton map is calculated at first; then feature pixel points are calculated on the map, and twelve kinds of feature pixel points are extracted and divided to eight directions; an integral image of the feature pixel points in the eight directions is calculated; finally, the texture main direction in any region on the image can be judged fast by utilizing the integral image of the feature pixel points; when a local texture image is enhanced, at first, high promotion filtering is carried out on an original image; according to the texture direction of the local pixel points, relief operation and Gauss filter of a bilinear interpolation are executed in an iteration mode for two to three times until the image effect is stable. According to the method, directionality texture features can be effectively enhanced, details such as the direction, the density and the texture width of directionality texture are highlighted, the influence on the image texture from factors such as illumination and noise can be effectively restrained, and therefore the enhanced texture image can be applied to other digital image processing algorithms.
Description
Technical field
The present invention relates to the texture description in digital image processing field and image enhancement technique field, and in particular to a kind of
Image enchancing method based on local grain directivity.
Background technology
Textural characteristics describe the matter of utmost importance as image segmentation with classification, are all the time the focuses of research.As
A kind of important visual cues, it in the picture generally existing but be difficult to describe.People always attempt to find textural characteristics dimension
Number is little, the feature description that robustness is good, separating capacity is strong.Nowadays there are many different texture description methods, it is main to divide
For statistical method, method of geometry, structural approach, model method and signal processing method;Its submethod such as gray level co-occurrence matrixes method,
Fourier's energy conversion method, fractal dimension method, markov random file, wavelet analysises etc..These methods to the feature of texture all
It is described from different perspectives, and all tool has been widely used.
Algorithm for image enhancement has a wide range of applications in Digital Image Processing application, and current main method is divided into spatial domain
Method and frequency domain method.Method based on spatial domain includes greyscale transformation, histogram modification, grad enhancement, medium filtering etc.;
Method based on frequency domain mainly includes high-pass filtering, Laplace operator etc..All kinds of algorithms are carried out from different perspectives to image
Conversion, can strengthen some details of image or feature, or remove our unconcerned contents on image, be the follow-up place of image
Reason offer information.
The content of the invention
In order to solve the problems, such as above-mentioned prior art, it is an object of the invention to provide a kind of be based on local grain side
The image enchancing method of tropism, can effectively strengthen Directional texture feature, the direction of projected direction texture, density, texture
The details such as width, can effectively suppress impact of the factors such as illumination, noise to image texture, make the image of enhanced texture
In being applied to other Digital Image Processing algorithms.
To reach object above, the present invention is employed the following technical solutions:
Based on the image enchancing method of local grain directivity, comprise the steps:
Step 1:Local grain walking direction
1) image outline figure is calculated;
2) feature pixel is calculated on contour images, 12 kinds of feature pixels are extracted, and is divided into 8 directions,
Concrete grammar is:For profile diagram, according to the situation in each pixel 3*3 neighborhoods, the cryptographic Hash of each pixel is asked for, often
The value non-zero i.e. 1 of individual pixel, makes the value of each pixel pixel 3*3 neighborhoods Nei and per phase of one 82 system numbers
Correspondence, the decimal form of this 82 system numbers is the cryptographic Hash of this pixel;For on all images a little, only
The pixel of 12 kinds of great directivity meanings is considered as feature pixel, feature pixel meets following condition:
A) central pixel point is foreground point in 3*3 neighborhoods;
B) in 3*3 neighborhoods, in addition to central pixel point, have and only two foreground point A1, A2
C) in 3*3 fields, around central pixel point along clockwise direction, 1 background dot of A1 to A2 interval Ds, between A2 to A1
Every D2 background dot, the value of D1, D2 or while be 3, or one is 2, one is 4;
According to the cryptographic Hash of each pixel, feature pixel is extracted, for other cryptographic Hash pixels are then ignored;
The concrete grammar that 8 directions divide is:The non-central pixel line angle difference positive with X-axis in 3*3 neighborhoods
For 0 °, 30 °, 45 °, 60 °, 90 °, 120 °, 135 °, 150 °, as 8 directions;
3) the feature pixel integrogram in 8 directions is calculated;
4) texture principal direction in any region is quickly judged on image using feature pixel integrogram:It is located at phase in window
Two kinds of directions that the character pixel points sum in adjacent two directions is maximum are respectively θ1And θ2, its corresponding character pixel points difference
For NUM1And NUM2, principal direction θ of last this feature pixelfinalIt is characterized the maximum adjacent two kinds of directions of pixel number sum
Weighted sum, it is as follows
θfinal=θ1*(NUM1/(NUM1+NUM2))+θ2*(NUM2/(NUM11+NUM2))
Step 2:Local grain image enhaucament
1) high boostfiltering is carried out to image;
2) according to the grain direction of local pixel point, iteration performs the embossed operation of bilinear interpolation and gaussian filtering two
To three times, until image effect is stablized;Iteration performs the embossed operation concrete grammar of bilinear interpolation:To every on original image
Individual pixel presses principal direction θfinalVertical direction, take predeterminable range 2 points make difference, are multiplied by the coefficient more than 10, add
Background constant 128, to choose needed in certain direction away from each pixel and makees two pixels of difference using the method for bilinear interpolation
Value.
Preferably, denoising is carried out to the image outline figure that step 1 is calculated, reduces noise spot and texture principal direction judgement is done
Disturb.
Compared to the prior art the present invention, has the advantage that:
The feature and details of Directional texture on image are strengthened by the inventive method.The directivity of texture is texture image
A basic feature, strengthened by direction characteristic to image, by the unconcerned factor such as illumination on local grain, noise
Suppress, remain details of the topography in grain direction, become image and further recognize, classify, the foundation of judgement.
Enhanced texture image remains the direction characteristic and grain details of texture, it is suppressed that the shadow such as texture color, illumination, noise
Ring.
Description of the drawings
Fig. 1 is 12 kinds of feature pixel schematic diagrams.
Fig. 2 is 8 direction character pixel classification charts.
Fig. 3 is the embossed operation chart of bilinear interpolation.
Fig. 4 is the Prototype drawing of gaussian filtering.
Fig. 5 is original image.
Fig. 6 is enhanced image.
Fig. 7 is the profile diagram of original image.
Fig. 8 is the image after high boostfiltering.
Fig. 9 is the partial enlarged drawing of Fig. 5.
Figure 10 is image enhanced to Fig. 9.
Specific embodiment
Below in conjunction with drawings and the specific embodiments, the present invention is described in further detail.
The present invention is broadly divided into two large divisions:1. local grain walking direction;2. local grain image enhaucament.Local grain
Walking direction key step is divided into profile diagram calculating, and profile diagram denoising, feature pixel is calculated, feature pixel integrogram meter
Calculate, local grain walking direction;Local grain image enhaucament key step is divided into high boostfiltering, and iteration performs bilinear interpolation
Embossed and Gaussian Blur.
Specifically introduce step by step below:
1) local grain walking direction
Local grain walking direction used it is a kind of simple, effectively, quickly, local direction description of robust --- can
In complicated natural image, quickly, texture general direction is judged exactly.The basic orientation of texture is follow-up bilinearity
The embossed operation of interpolation provides foundation.Method calculates contour images with canny operators first;Then denoising is made to profile diagram;
Again the cryptographic Hash of each pixel is calculated according to the situation of each pixel 3*3 neighborhoods in profile binary map;According to each picture
The difference of vegetarian refreshments cryptographic Hash, extracts the pixel of 12 kinds of sign direction meanings;The pixel of 12 kinds of sign direction meanings is sorted out
For 8 kinds of directions, the integrogram of feature point number in 8 directions is calculated;For a certain region on image, by 8 kinds of directions
Integrogram can quickly calculate partial image texture direction.
1.1) image outline figure is calculated
The edge of image refers to the significant part of image regional area brightness flop, and the gray scale section in the region typically can be with
Regard a step as, both having to go to the toilet in the buffer area of very little from a gray value, to change to another gray scale difference larger for play
Gray value.The most information of image has been concentrated in the marginal portion of image, and the determination and extraction of image edge is for whole image
The identification of scene is very important with understanding, while being also the key character that Image is relied on.For with directivity
The image of texture, the edge graph of image has obvious feature.Here image outline figure is calculated using Canny operators, such as Fig. 7 institutes
Show.
Its key step is as follows:Gauss operator is smoothed to image, effectively suppresses noise present in image;Use single order
Difference convolution mask, calculates image gradient amplitude and direction;Non-maxima suppression (NMS), retains partial gradient maximum of points, suppression
Manufacture-illegal maximum gradient value;Dual threashold value-based algorithm detects and connects image outline, retains strong edge and the weak edge being connected with strong edge.
The profile diagram that Canny operators are calculated, extracts feature pixel and provides foundation for next step.
1.2) profile diagram denoising
For the profile diagram for calculating uses medium filtering, the noise of some less details in profile diagram is filtered.
Reduce the interference that noise spot is judged texture principal direction.
1.3) feature pixel is calculated
For the two-value profile diagram after denoising, according to the situation in each pixel 3*3 neighborhoods, each pixel is asked for
Cryptographic Hash.The value non-zero i.e. 1 of each pixel, makes the value and one 82 system numbers of each pixel pixel 3*3 neighborhoods Nei
Per it is corresponding, the decimal form of this 82 system numbers is the cryptographic Hash of this pixel.
For institute a little, only wants to consider the pixel of 12 kinds of great directivity meanings as character pixel on all images
Point, feature pixel meets following condition:
A) central pixel point is foreground point in 3*3 neighborhoods;
B) in 3*3 neighborhoods, in addition to central pixel point, have and only two foreground point A1, A2
C) in 3*3 fields, around central pixel point along clockwise direction, 1 background dot of A1 to A2 interval Ds, between A2 to A1
Every D2 background dot, the value of D1, D2 or while be 3, or one is 2, one is 4.
As shown in Figure 1 (black color dots represent foreground point, and white point represents background dot).According to the cryptographic Hash of each pixel,
We can extract these our feature pixel interested, for other cryptographic Hash pixels are then ignored.
As shown in Fig. 2 for the 12 kinds of pixels for extracting, being classified as 8 directions, non-central pixel in 3*3 neighborhoods
The line angle positive with X-axis is respectively 0 °, 30 °, 45 °, 60 °, 90 °, 120 °, 135 °, 150 °, used as 8 directions.
1.4) local grain direction calculating
In order to quickly calculate image in any region for the statistical information of feature pixel, image local area is needed
The feature pixel statistical value in interior which direction is more.The feature pixel integrogram with regard to 8 directions is first calculated, for each
The pixel in direction, calculates SUM [i] [j], in representative image by (0,0), (i, 0), (0, j), what (i, j) this four points were surrounded
In rectangular area, the number of this kind of direction pixel.Calculating for SUM [i] [j] with common integrogram recurrence formula,
As follows, Pixel [i] [j] represents the pixel whether pixel (i, j) is this kind of direction, and it is otherwise 0 for 1 to be then:
SUM [i] [j]=SUM [i-1] [j]+SUM [i] [j-1]-SUM [i-1] [j-1]+Pixel [i] [j]
For the feature pixel integrogram in calculated 8 kinds of directions, can be used to quickly calculate 8 kinds in any region
The statistical value of feature pixel.It is special to count all directions in the window ranges of d in the length of side centered on a certain pixel (i, j)
Pixel number is levied, the character pixel points formula of a direction is as follows:
Total=SUM [i+d/2] [j+d/2]-SUM [i+d/2] [j-d/2]-SUM [i-d/2] [j+d/2]+SUM [i-d/
2][j-d/2]
The character pixel points and two kinds of maximum directions for being located at adjacent two direction in window are respectively θ 1 and θ 2, its correspondence
Character pixel points be respectively NUM1, NUM2, the direction of last this pixel is the weighted sum in this adjacent two kinds of direction, as follows
It is shown
θ final=θ 1* (NUM1/ (NUM1+NUM2))+θ 2* (NUM2/ (NUM1+NUM2))
As it is assumed that local grain direction is consistent, so carrying out an average filter again to the direction of each pixel
Ripple, makes local grain direction as far as possible consistent.
2) local grain image enhaucament
Local grain image enhaucament, makees image to texture and increases according to the local grain direction calculated in previous step
By force.Because the particularity of Directional texture, the embossed operation of class, enhancing side are performed along local grain vertical direction iteration here
Tropism texture, retains Directional texture feature, suppresses other features in addition to grain direction, reaches the purpose of image enhaucament.
2.1) high boostfiltering
High boostfiltering is a kind of common image sharpening method, to effect after artwork filtering as shown in figure 8, its is main
Step is:1. original image is obscured.2. broad image is deducted from original image and obtains a template image, its computing formula is:3. the coefficient that template image is multiplied by more than 10 is then added on original image.This
Plant sharpening mode to be mainly sharpened image border, it is main to giving reservation where the conversion of color continuous uniform on image
The advantage wanted can be seldom to produce noise while image sharpening is ensured.Its formula is:G (x, y)=f (x, y)+k*
gmask(x, y).Wherein:F (x, y) represents original image,Represent the image after obscuring, gmask(x, y) represents template image,
G (x, y) represents the image after high boostfiltering, and k is a larger coefficient.
2.2) bilinear interpolation is embossed
Make the embossed operation with bilinear interpolation on image after sharpening.To each pixel on image by certain side
To, take certain distance 2 points make difference, are multiplied by the coefficient more than 10, along with background constant 128 (image is gray-scale maps, as
Vegetarian refreshments maximum is for 255).The larger gray value of difference is so caused, difference is more projected, so as to strengthen the effect of local grain
Really, the factors such as illumination, noise are made to be inhibited.Make difference by 2 points of fixed-direction certain distance a pixel is chosen
When, because two pixel point coordinates are not necessarily integer, 2 points for choosing differ and are surely fetched directly into from image, if choosing
Take nearer integral point approximately to replace, then can make it is embossed after image produce more noise.Here inserted using bilinearity and referred to
Method choose away from each pixel need in certain direction make difference two pixel point values.
Detail is as follows:
A. for the pixel in image, according to pixel grain direction θ calculated in local grain walking directionfinal
Vertical direction, with predeterminable range d, selection needs to make two point coordinates b and c of difference, as shown in figure 3, the following institute of computing formula
X is stated, y represents respective pixel point coordinates.
B.x=a.x-d*sin θ.
B.y=a.y+d*cos θ
C.x=a.x+d*sin θ.
C.y=a.y-d*cos θ
B. for the b for calculating, c point coordinates may not be integer, it is impossible to directly pixel gray value be obtained from image,
Here being obtained using the method for bilinear interpolation --- the gray value for obtaining most adjoining four pixels is multiplied by respective weights system
Number, i.e., for point b, c on figure, by the gray value on correspondingly four summits of white edge its value is calculated, each pixel is multiplied by with
The related proportionality coefficient of the distance of impact point is sued for peace again.
C. for the b for obtaining, the value that 2 points of c asks for poor absolute value, is multiplied by a larger proportionality coefficient k, adds
Value after background constant 128, as rearmost point a are embossed, formula is as follows:
ANS=MIN (255, MAX (0, (B_value-C_value) * k+128)))
2.3) gaussian filtering
Can also there are many noises in the image later for embossed operation, here using Gaussian filter to picture noise
Suppressed.Gaussian filter is the linear smoothing filter that a class selects weights according to the shape of Gaussian function, such as Fig. 4 institutes
Show.Gaussian filter is highly effective for the noise for suppressing Normal Distribution.It is flat by being weighted to entire image
, the value of each pixel, obtains after being all weighted averagely by other pixel values in itself and neighborhood.
After gaussian filtering, general image can tend to fuzzy again, repeat execution step 2.2 and step 2.3, this process iteration
Perform 2 to 3 times, make picture quality strengthen quality tend towards stability, local original image as shown in figure 9, strengthen after image such as Figure 10 institutes
Show.
The present invention proposes a kind of algorithm for image enhancement based on local direction texture, can effectively strengthen Directional texture
The details such as feature, the direction of projected direction texture, density, texture width, can effectively suppress the factors pair such as illumination, noise
The impact of image texture, the image for making enhanced texture is applied in other Digital Image Processing algorithms, before and after texture strengthens
Image is as shown in Figure 5 and Figure 6.
Claims (2)
1. the image enchancing method of local grain directivity is based on, it is characterised in that:Comprise the steps:
Step 1:Local grain walking direction
1) image outline figure is calculated;
2) feature pixel is calculated on contour images, 12 kinds of feature pixels are extracted, and is divided into 8 directions, specifically
Method is:For profile diagram, according to the situation in each pixel 3*3 neighborhoods, the cryptographic Hash of each pixel is asked for, each picture
The value non-zero i.e. 1 of vegetarian refreshments, makes the value of each pixel pixel 3*3 neighborhoods Nei corresponding with per of one 82 system numbers,
The decimal form of this 82 system numbers is the cryptographic Hash of this pixel;For on all contour images a little, only
The pixel of 12 kinds of great directivity meanings is considered as feature pixel, feature pixel meets following condition:
A) central pixel point is foreground point in 3*3 neighborhoods;
B) in 3*3 neighborhoods, in addition to central pixel point, have and only two foreground point A1, A2
C) in 3*3 fields, around central pixel point along clockwise direction, 1 background dot of A1 to A2 interval Ds, A2 to A1 interval Ds 2
Individual background dot, the value of D1, D2 or while be 3, or one be 2, one be 4;
According to the cryptographic Hash of each pixel, feature pixel is extracted, for other cryptographic Hash pixels are then ignored;
The concrete grammar that 8 directions divide is:The non-central foreground pixel point line angle difference positive with X-axis in 3*3 neighborhoods
For 0 °, 30 °, 45 °, 60 °, 90 °, 120 °, 135 °, 150 °, as 8 directions;
3) the feature pixel integrogram in 8 directions is calculated;
4) texture principal direction in any region is quickly judged on image using feature pixel integrogram:It is located at adjacent two in window
Two kinds of directions that the character pixel points sum in direction is maximum are respectively θ1And θ2, its corresponding character pixel points is respectively
NUM1And NUM2, principal direction θ of last this feature pixelfinalIt is characterized the maximum adjacent two kinds of directions of pixel number sum
Weighted sum, it is as follows
θfinal=θ1*(NUM1/(NUM1+NUM2))+θ2*(NUM2/(NUM1+NUM2))
Step 2:Local grain image enhaucament
1) high boostfiltering is carried out to image;
2) according to the grain direction of local pixel point, iteration performs the embossed operation of bilinear interpolation and gaussian filtering two to three
It is secondary, until image effect is stablized;Iteration performs the embossed operation concrete grammar of bilinear interpolation:To each picture on original image
Vegetarian refreshments presses principal direction θfinalVertical direction, take predeterminable range 2 points make difference, the coefficient more than 10 are multiplied by, along with background
Constant 128, to choose needed in certain direction away from each pixel and makees two pixel point values of difference using the method for bilinear interpolation.
2. the image enchancing method based on local grain directivity according to claim 1, it is characterised in that:To step 1
The image outline figure of calculating carries out denoising.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410741870.4A CN104463814B (en) | 2014-12-08 | 2014-12-08 | Image enhancement method based on local texture directionality |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410741870.4A CN104463814B (en) | 2014-12-08 | 2014-12-08 | Image enhancement method based on local texture directionality |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104463814A CN104463814A (en) | 2015-03-25 |
CN104463814B true CN104463814B (en) | 2017-04-19 |
Family
ID=52909802
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410741870.4A Active CN104463814B (en) | 2014-12-08 | 2014-12-08 | Image enhancement method based on local texture directionality |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104463814B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105488758B (en) * | 2015-11-30 | 2018-07-20 | 河北工业大学 | A kind of image-scaling method based on perception of content |
CN107368835B (en) * | 2016-05-12 | 2020-11-03 | 北京君正集成电路股份有限公司 | Method and device for acquiring integral sum of pixel points |
CN106169177A (en) * | 2016-06-27 | 2016-11-30 | 北京金山安全软件有限公司 | Image buffing method and device and electronic equipment |
CN107103619B (en) * | 2017-04-19 | 2022-03-08 | 腾讯科技(上海)有限公司 | Method, device and system for processing hair texture direction |
CN109615607B (en) * | 2018-11-09 | 2023-05-16 | 福建和盛高科技产业有限公司 | Noise detection method based on single image custom features |
CN111178289B (en) * | 2019-12-31 | 2023-08-25 | 张杰辉 | Method and system for shortening iris recognition time consumption |
CN112329796B (en) * | 2020-11-12 | 2023-05-23 | 北京环境特性研究所 | Infrared imaging cloud detection method and device based on visual saliency |
CN113743508B (en) * | 2021-09-06 | 2024-07-02 | 合肥工业大学 | Outlier identification method in mixed Gaussian pulse noise image |
CN115526810B (en) * | 2022-11-07 | 2023-04-07 | 青岛理工大学 | Underwater image enhancement method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103440625A (en) * | 2013-08-16 | 2013-12-11 | 浙江大学 | Hyperspectral image processing method based on textural feature strengthening |
CN103955929A (en) * | 2014-04-29 | 2014-07-30 | 北京工商大学 | Method and device for judging image local edge mode and non-edge mode |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2000357206A (en) * | 1999-06-15 | 2000-12-26 | Nippon Telegr & Teleph Corp <Ntt> | Method for processing line drawing and recording medium where line drawing processing program is recorded |
JP5941674B2 (en) * | 2011-12-28 | 2016-06-29 | オリンパス株式会社 | Cell contour forming apparatus and method, and cell contour forming program |
-
2014
- 2014-12-08 CN CN201410741870.4A patent/CN104463814B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103440625A (en) * | 2013-08-16 | 2013-12-11 | 浙江大学 | Hyperspectral image processing method based on textural feature strengthening |
CN103955929A (en) * | 2014-04-29 | 2014-07-30 | 北京工商大学 | Method and device for judging image local edge mode and non-edge mode |
Non-Patent Citations (3)
Title |
---|
Extracting image orientation feature by using integration operator;Xudong Jiang;《ScienceDirect》;20071231;第705-717页 * |
基于Haar特性的LBP纹理特性;周书仁等;《软件学报》;20131231;第24卷(第8期);第1909-1926页 * |
基于纹理的自适应提升小波框架的图像增强;李广琼等;《计算机应用与软件》;20140331;第31卷(第3期);第191-194页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104463814A (en) | 2015-03-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104463814B (en) | Image enhancement method based on local texture directionality | |
CN106530237A (en) | Image enhancement method | |
CN108038833B (en) | Image self-adaptive sharpening method for gradient correlation detection and storage medium | |
CN109377450B (en) | Edge protection denoising method | |
CN105374015A (en) | Binary method for low-quality document image based on local contract and estimation of stroke width | |
JPH06348842A (en) | Noise reducing filter | |
CN108765325A (en) | Small unmanned aerial vehicle blurred image restoration method | |
CN104966285B (en) | A kind of detection method of salient region | |
CN102289668A (en) | Binaryzation processing method of self-adaption word image based on pixel neighborhood feature | |
CN105405138B (en) | Waterborne target tracking based on conspicuousness detection | |
CN109961416B (en) | Business license information extraction method based on morphological gradient multi-scale fusion | |
CN109544583B (en) | Method, device and equipment for extracting interested area of leather image | |
Krishnan et al. | A survey on different edge detection techniques for image segmentation | |
CN103310439B (en) | The most steady extremal region detection method of a kind of image based on metric space | |
CN113034452A (en) | Weldment contour detection method | |
CN113592776A (en) | Image processing method and device, electronic device and storage medium | |
CN104424641A (en) | Detection method for image fuzzy tampering | |
CN114418904B (en) | Infrared image enhancement method based on improved histogram equalization and enhanced high-pass filtering | |
CN102930538B (en) | The self-adaptive projection method method that Gauss potential and spatial histogram merge | |
CN104050472B (en) | A kind of adaptive global threshold method of Binary Sketch of Grey Scale Image | |
CN112633070A (en) | High-resolution remote sensing image building extraction method and system | |
CN104700418A (en) | Method for estimating precise circumferences of target boundaries based on gray level information | |
CN105303538A (en) | Gauss noise variance estimation method based on NSCT and PCA | |
CN110807747B (en) | Document image noise reduction method based on foreground mask | |
Das et al. | Histogram equalization techniques for contrast enhancement: a review |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20190312 Address after: 405, 4th floor, No. 2 Building, 28 Shangdi Sixth Street, Haidian District, Beijing 100085 Patentee after: Beijing Jieshitong Technology Development Co., Ltd. Address before: 710049 No. 28, Xianning Road, Xi'an, Shaanxi Patentee before: Xi'an Jiaotong University |