CN108052945A - A kind of similar determination method of improved pictures of LBP - Google Patents
A kind of similar determination method of improved pictures of LBP Download PDFInfo
- Publication number
- CN108052945A CN108052945A CN201711305720.9A CN201711305720A CN108052945A CN 108052945 A CN108052945 A CN 108052945A CN 201711305720 A CN201711305720 A CN 201711305720A CN 108052945 A CN108052945 A CN 108052945A
- Authority
- CN
- China
- Prior art keywords
- lbp
- values
- diagonal
- dct
- pixel
- 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.)
- Withdrawn
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
- G06V10/20—Image preprocessing
- G06V10/32—Normalisation of the pattern dimensions
-
- 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
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of similar determination method of the improved pictures of LBP, the first step, compressed picture sizes;Second step, by compressed picture gray processing;3rd step extracts diagonal gray scale;Cornerwise each pixel is carried out dct transform by the 4th step;5th step calculates the LBP values of diagonal pixels point;6th step, establishes the hash of LBP, calculates the average of diagonal L BP, each LBP values and average comparison, more than or equal to being denoted as 1, less than being denoted as 0;It is hashed using uniformity, obtains 15 hashed values based on LBP;7th step, comparison judges, the hashed value of the LBP of different pictures is compared, less than or equal to 2, then judges approximation.
Description
Technical field
The present invention relates to image processing field, the similar determination method of more particularly to a kind of improved pictures of LBP.
Background technology
LBP (Local Binary Pattern, local binary patterns) is that one kind is used for describing image local textural characteristics
Operator;Original LBP operator definitions are in the window of 3*3, using window center pixel as threshold value, by 8 adjacent pixels
Gray value compared with it, if surrounding pixel values are more than center pixel value, the position of the pixel is marked as 1, no
It is then 0.In this way, 8 points in 3*3 neighborhoods through compare can generate 8 bits (be typically converted into decimal number i.e. LBP codes,
Totally 256 kinds) to get to the LBP values of the window center pixel, and reflect with this value the texture information in the region.Such as Fig. 1
It is shown.
The greatest drawback of basic LBP operators is that it covers only the zonule in the range of a radii fixus, this is aobvious
The needs of different sizes and frequency texture cannot so be met.
Circular LBP operators:In order to adapt to the textural characteristics of different scale, and reach the requirement of gray scale and rotational invariance,
Ojala etc. improves LBP operators, by 3 × 3 neighborhood extendings to arbitrary neighborhood, and with circle shaped neighborhood region instead of square
Neighborhood, improved LBP operators allow have any number of pixels in the circle shaped neighborhood region that radius is R.So as to obtain such as
Radius is the LBP operators containing P sampled point in the border circular areas of R, shown in Fig. 2.
LBP invariable rotary patterns:From the definition of LBP as can be seen that LBP operators are that gray scale is constant, but it is not rotation
Constant.The rotation of image will obtain different LBP values.
Maenpaa et al. is again extended LBP operators, it is proposed that has the LBP operators of rotational invariance, i.e., constantly
Rotation circle shaped neighborhood region obtains a series of LBP values of original definitions, takes LBP value of its minimum value as the neighborhood.
Fig. 3 gives the process schematic for the LBP for asking for invariable rotary, the digital representation operator pair in figure below operator
The LBP values answered, 8 kinds of LBP patterns shown in figure, by the processing of invariable rotary, what is finally obtained has rotational invariance
LBP values are 15.That is, the LBP patterns of the corresponding invariable rotary of 8 kinds of LBP patterns in figure are all 00001111.
LBP equivalent formulations:One LBP operator can generate different binary modes, for the border circular areas that radius is R
The interior LBP operators containing P sampled point will generate 2 kinds of patterns of P.It will be apparent that with the increase of sampling number in neighborhood collection,
The species of binary mode sharply increases.Such as:20 sampled points in 5 × 5 neighborhoods, there is a 220=1,048,576 kind two into
Molding formula.So many binary pattern no matter for texture extraction or the identification for texture, classification and the access of information
All it is unfavorable.Meanwhile excessive schema category is unfavorable for the expression of texture.For example, by LBP operators for texture point
When class or recognition of face, the information of image is expressed frequently with the statistic histogram of LBP patterns, and more schema category will make
It is excessive to obtain data volume, and histogram is excessively sparse.Therefore, it is necessary to carry out dimensionality reduction to original LBP patterns so that data volume is reduced
In the case of can best representative image information.
In order to solve the problems, such as that binary mode is excessive, statistics is improved, Ojala is proposed using a kind of " equivalent formulations "
(Uniform Pattern) to carry out dimensionality reduction to the schema category of LBP operators.In real image, most LBP patterns are most
More include the saltus step from 1 to 0 or from 0 to 1 twice.Therefore, " equivalent formulations " are defined as by Ojala:When corresponding to some LBP
Cycling binary number from 0 to 1 or from 1 to 0 be up to saltus step twice when, the binary system corresponding to the LBP is known as one etc.
Valency pattern class.Such as 00000000 (0 saltus step), 00000111 (containing only the saltus step once from 0 to 1), 10001111 (are first jumped by 1
1 is jumped to 0, then by 0, saltus step twice altogether) all it is equivalent formulations class.Pattern in addition to equivalent formulations class be all classified as it is another kind of,
Referred to as mixed mode class, such as 10010111 (totally four saltus steps).
Improvement in this way, the species of binary mode greatly reduce, without losing any information.Pattern quantity by
2P kinds originally are reduced to+2 kinds of P (P-1), and wherein P represents the sampling number in neighborhood collection.For 8 samplings in 3 × 3 neighborhoods
For point, binary mode is reduced to 58 kinds by original 256 kinds, this so that the dimension of feature vector is less, and can subtract
The influence that few high-frequency noise is brought.
By above-mentioned summary it can be seen that:The judgement that LBP after deformation is well suited for face and personage moves, but the meter of LBP
Calculation amount is huge, inefficent advantage, it is difficult to adapt to the similar judgement of internet mass picture.
Applicant before LBP methods, is first compressed at name of patent application " the similar determination method of picture based on LBP "
Picture adjusts pixel grey scale, does dct transform afterwards, obtain most simple information, then carry out LBP, abandoned traditional Nogata afterwards
Figure determination methods, but the method for using LBP averages hash, obtain more details;But due to being the full figure piece after compressing
LBP is together, and calculation amount still shows huge, in addition, robot needs manually accurate judgement after judging, therefore, early period only needs
One general judgement, it is desirable that be:Do not miss similar pictures, that is to say, that can find out more and be accused of similar figure
Piece, as to whether it is genuine similar, it can be with artificial judgment;Empirical evidence, within 100 pictures, artificial judgment efficiency is very high.
The content of the invention
In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a kind of similar determination methods of the improved pictures of LBP.
A kind of similar determination method of improved pictures of LBP provided by the invention, comprises the following steps:
The first step, compressed picture size
By picture compression to 64*64 pixels;
Second step, by compressed picture gray processing
Compressed picture is down to 64 grades of gray scales;
3rd step extracts diagonal gray scale
4th step, dct transform
Cornerwise each pixel is subjected to dct transform;
5th step calculates the LBP values of diagonal pixels point
For each diagonal pixels DCT values compared with the DCT values of 8 adjacent pixels, if surrounding pixel values
More than middle imago DCT values, then the position of the pixel is marked as 1, is otherwise 0;8 points in 3*3 neighborhoods can be produced through comparing
Give birth to LBP value of 8 bits to get imago DCT values in this;
6th step establishes the hash of LBP
The average of diagonal L BP, each LBP values and average comparison are calculated, more than or equal to being denoted as 1, less than being denoted as 0;It adopts
It is hashed with uniformity, obtains 15 hashed values based on LBP;
7th step, comparison judge
The hashed value of the LBP of different pictures is compared, less than or equal to 2, then judges approximation.
Advantageous effect:The LBP averages that the present invention is only compared on diagonal (only calculate diagonal and neighbor pixel
LBP), and 15 are formed as hashed value, 4 times of calculation amount is saved compared with 64 hash.
Description of the drawings
Fig. 1 is LBP method schematics.
Fig. 2 is circular LBP method schematics.
Fig. 3 is the process schematic for the LBP for asking for invariable rotary.
Fig. 4 is original figure spectrum and LBP collection of illustrative plates.
Specific embodiment
Embodiment:
The first step, compressed picture size
By picture compression to 64*64 pixels;
Second step, by compressed picture gray processing
Compressed picture is down to 64 grades of gray scales;
3rd step extracts diagonal gray scale
4th step, dct transform
Cornerwise each pixel is subjected to dct transform;
5th step calculates the LBP values of diagonal pixels point
For each diagonal pixels DCT values compared with the DCT values of 8 adjacent pixels, if surrounding pixel values
More than middle imago DCT values, then the position of the pixel is marked as 1, is otherwise 0;8 points in 3*3 neighborhoods can be produced through comparing
Give birth to LBP value of 8 bits to get imago DCT values in this;
6th step establishes the hash of LBP
The average of diagonal L BP, each LBP values and average comparison are calculated, more than or equal to being denoted as 1, less than being denoted as 0;It adopts
It is hashed with uniformity, obtains 15 hashed values based on LBP;
7th step, comparison judge
The hashed value of the LBP of different pictures is compared, less than or equal to 2, then judges approximation.
Claims (1)
1. the similar determination method of a kind of improved pictures of LBP, which is characterized in that comprise the following steps:
The first step, compressed picture size
By picture compression to 64*64 pixels;
Second step, by compressed picture gray processing
Compressed picture is down to 64 grades of gray scales;
3rd step extracts diagonal gray scale
4th step, dct transform
Cornerwise each pixel is subjected to dct transform;
5th step calculates the LBP values of diagonal pixels point
For each diagonal pixels DCT values compared with the DCT values of 8 adjacent pixels, if surrounding pixel values are more than
Middle imago DCT values, then the position of the pixel be marked as 1, be otherwise 0;8 points in 3*3 neighborhoods can generate 8 through comparing
Binary number to get imago DCT values in this LBP values;
6th step establishes the hash of LBP
The average of diagonal L BP, each LBP values and average comparison are calculated, more than or equal to being denoted as 1, less than being denoted as 0;Using one
Cause property hash, obtains 15 hashed values based on LBP;
7th step, comparison judge
The hashed value of the LBP of different pictures is compared, less than or equal to 2, then judges approximation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711305720.9A CN108052945A (en) | 2017-12-11 | 2017-12-11 | A kind of similar determination method of improved pictures of LBP |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711305720.9A CN108052945A (en) | 2017-12-11 | 2017-12-11 | A kind of similar determination method of improved pictures of LBP |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108052945A true CN108052945A (en) | 2018-05-18 |
Family
ID=62123743
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711305720.9A Withdrawn CN108052945A (en) | 2017-12-11 | 2017-12-11 | A kind of similar determination method of improved pictures of LBP |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108052945A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113962993A (en) * | 2021-12-21 | 2022-01-21 | 武汉霖杉工贸有限公司 | Paper cup raw material quality detection method based on computer vision |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104504120A (en) * | 2014-12-29 | 2015-04-08 | 北京奇艺世纪科技有限公司 | Detection method and device for similar pictures |
CN105681898A (en) * | 2015-12-31 | 2016-06-15 | 北京奇艺世纪科技有限公司 | Similar video and pirated video detection method and device |
EP3035209A1 (en) * | 2014-12-18 | 2016-06-22 | Thomson Licensing | Method and apparatus for deriving a perceptual hash value from an image |
CN105912739A (en) * | 2016-07-14 | 2016-08-31 | 湖南琴海数码股份有限公司 | Similar image retrieval system and method |
CN106528743A (en) * | 2016-11-01 | 2017-03-22 | 山东浪潮云服务信息科技有限公司 | High-efficiency similar picture identification method based on picture mining technology |
-
2017
- 2017-12-11 CN CN201711305720.9A patent/CN108052945A/en not_active Withdrawn
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3035209A1 (en) * | 2014-12-18 | 2016-06-22 | Thomson Licensing | Method and apparatus for deriving a perceptual hash value from an image |
CN104504120A (en) * | 2014-12-29 | 2015-04-08 | 北京奇艺世纪科技有限公司 | Detection method and device for similar pictures |
CN105681898A (en) * | 2015-12-31 | 2016-06-15 | 北京奇艺世纪科技有限公司 | Similar video and pirated video detection method and device |
CN105912739A (en) * | 2016-07-14 | 2016-08-31 | 湖南琴海数码股份有限公司 | Similar image retrieval system and method |
CN106528743A (en) * | 2016-11-01 | 2017-03-22 | 山东浪潮云服务信息科技有限公司 | High-efficiency similar picture identification method based on picture mining technology |
Non-Patent Citations (2)
Title |
---|
Q123456789098: "图像特征提取三大法宝:HOG特征、LBP特征、Haar-like特征", 《CSDN》 * |
ZOUXY09: "基于感知哈希算法的视觉目标跟踪", 《CSDN》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113962993A (en) * | 2021-12-21 | 2022-01-21 | 武汉霖杉工贸有限公司 | Paper cup raw material quality detection method based on computer vision |
CN113962993B (en) * | 2021-12-21 | 2022-03-15 | 武汉霖杉工贸有限公司 | Paper cup raw material quality detection method based on computer vision |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
TWI426774B (en) | A method for classifying an uncompressed image respective to jpeg compression history, an apparatus for classifying an image respective to whether the image has undergone jpeg compression and an image classification method | |
WO2020224458A1 (en) | Method for detecting corona discharge employing image processing | |
WO2017121018A1 (en) | Method and apparatus for processing two-dimensional code image, and terminal and storage medium | |
CN115861135A (en) | Image enhancement and identification method applied to box panoramic detection | |
CN110428450B (en) | Scale-adaptive target tracking method applied to mine tunnel mobile inspection image | |
CN108898132B (en) | Terahertz image dangerous article identification method based on shape context description | |
CN108829711B (en) | Image retrieval method based on multi-feature fusion | |
CN106780449A (en) | A kind of non-reference picture quality appraisement method based on textural characteristics | |
CN106203461B (en) | Image processing method and device | |
CN107610093B (en) | Full-reference image quality evaluation method based on similarity feature fusion | |
CN104504662A (en) | Homomorphic filtering based image processing method and system | |
CN110032946B (en) | Aluminum/aluminum blister packaging tablet identification and positioning method based on machine vision | |
CN115937216A (en) | Magnetic rotor appearance quality detection method for new energy automobile | |
CN101976340B (en) | License plate positioning method based on compressed domain | |
Lee et al. | Color image enhancement using histogram equalization method without changing hue and saturation | |
Kumar et al. | Near lossless image compression using parallel fractal texture identification | |
CN115331119A (en) | Solid waste identification method | |
CN110782409B (en) | Method for removing shadow of multiple moving objects | |
CN111738984B (en) | Skin image spot evaluation method and system based on watershed and seed filling | |
CN108052945A (en) | A kind of similar determination method of improved pictures of LBP | |
CN110766614B (en) | Image preprocessing method and system of wireless scanning pen | |
CN107909076A (en) | A kind of similar decision method of picture based on LBP | |
CN110223273B (en) | Image restoration evidence obtaining method combining discrete cosine transform and neural network | |
CN103796017B (en) | Image discriminating device and method | |
Mustaghfirin et al. | The comparison of iris detection using histogram equalization and adaptive histogram equalization methods |
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 | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20180518 |