CN109284744A - A method of iris image is encoded from eye gray level image likelihood figure and is retrieved - Google Patents
A method of iris image is encoded from eye gray level image likelihood figure and is retrieved Download PDFInfo
- Publication number
- CN109284744A CN109284744A CN201811306066.8A CN201811306066A CN109284744A CN 109284744 A CN109284744 A CN 109284744A CN 201811306066 A CN201811306066 A CN 201811306066A CN 109284744 A CN109284744 A CN 109284744A
- Authority
- CN
- China
- Prior art keywords
- image
- feature
- likelihood
- value
- iris
- 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
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/18—Eye characteristics, e.g. of the iris
- G06V40/197—Matching; Classification
-
- 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
-
- 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
- G06V10/467—Encoded features or binary features, e.g. local binary patterns [LBP]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Ophthalmology & Optometry (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of from eye gray level image likelihood figure to the method for iris image coding retrieval, image is zoomed in and out first and is used to generate likelihood figure, four different local features are generated using likelihood figure, that is mean value, standard deviation, the degree of bias and horizontal edge, for each pixel reduced in figure RD, these features are all derived from the neighborhood centered on the pixel, mean value, standard deviation and the degree of bias are calculated on the local rectangular window that a size is 7 pixels, and serves as invariable rotary sparse edge filter；In addition, these filters are also used as invariable rotary sparse edge filter, also deal with feedback to the edge of bristle part covering；Iris image is normalized on the basis of likelihood figure and calculates three feature vectors, and coding retrieval is carried out to textural characteristics finally by feature vector.Image down likelihood figure can reduce the amount of calculation of image, effectively raise the retrieval rate of iris in the case where guaranteeing that feature will not lose in the present invention.
Description
Technical field
The present invention relates to a kind of methods of iris image coding retrieval, more particularly to one kind from eye gray level image likelihood
Method of the figure to iris image coding retrieval.
Background technique
In recent years, a large amount of iris texture extraction algorithm is suggested.Forefront is studied and is developed by Daugman at present.
This is extracted iris using a kind of 2D Gabor filter of complexity and realizes vector product.Now, this method has become many
The basic skills of business system.
However, most of method requires complicated mathematical computations, and it may may require that and the knot to be calculated such as carry out for a long time
Fruit.In order to accelerate recognition speed and keep higher discrimination, it is proposed that a kind of diminution based on from eye gray level image is seemingly
So scheme calculated three characteristic quantities and come the coding to iris image feature and the new method of retrieval, the diminution likelihood figure of image can
To reduce the amount of calculation of image guaranteeing in the case where will not losing of feature, by average LPB, average gray value and
Three characteristic quantities of entropy of image can extract characteristics of image faster, and dimensional space is less, effectively raises rainbow in this way
The retrieval rate of film.
Summary of the invention
For overcome the deficiencies in the prior art, the present invention provides one kind and compiles from eye gray level image likelihood figure to iris image
The method of code retrieval can solve at present in the universal not high problem of iris recognition efficiency.
In order to solve the above technical problems, the invention provides the following technical scheme:
A method of iris image is encoded from eye gray level image likelihood figure and is retrieved, specific steps are as follows:
Step 1: figure is zoomed in and out be used to generate likelihood figure first:
Allow RF to become a NO emissions reduction factor, the value RD of the image of the diminution from pixel Pix-i ∈ Img (Pix-I=Pi),
The two sides stride RF are calculated based on rectangular sliding window Width (W) and L=2*RF+1 pixel and pixel, for each position of Width
It sets, the mean intensity u in window is calculated as
Calculation window intensity histogram Hist, and the corresponding pixel value d calculation formula in image is
Step 2: likelihood figure is generated:
Four different local features, i.e. mean value, standard deviation, the degree of bias and horizontal edge are generated using likelihood figure, for contracting
Each pixel in small figure RD, these features are all derived from the neighborhood centered on the pixel, are 7 pictures in a size
Mean value, standard deviation and the degree of bias are calculated on the local rectangular window of element, and serve as invariable rotary sparse edge filter；In addition, this
A little filters are also used as invariable rotary sparse edge filter, also deal with feedback to the edge of bristle part covering；
Calculating average value is first moment (M1), and (is weighted close to the higher small yin of eyelid using the complement code of downscaled images as input
Shadow)；Calculating mean standard deviation is second order moment (M2)；Calculating semiconductor body is third moment (M3)；Use Prwitt operator
Detection level edge removes part pseudo-edge, is smoothed to noise, and the reinforcing as horizontal edge in likelihood figure；
It is worth noting that, even if eyes are misaligned with camera horizon, certain parts of eyelid also due to eyelid arc property
And causing biological respinse, each feature generates a relevant figure: mean value (M1), standard deviation (M2), gradient (M3) and water
Pingbian edge (E) figure, these combine with carrying out multiplication and division point by point generates likelihood figure L and is
The high-pass filter of E, M1 and M2 and the low-pass filter of M3 are had effectively achieved, a height is used on L
For 3 pixels, the mean filter of L width 30% is covered, with the horizontal disjoint high likelihood region of connection, and is increased straight
The response of line horizontal edge portions, this operation introduce some noises, by being set as 0 for insignificant, can partially disappear
Except these noises (for example value can be changed to 0 less than maximum value 0.1% to 1%), in addition, we also will be outside likelihood eyelid region
Likelihood mapping value be set as zero；
Step 3: iris image is normalized on the basis of likelihood figure and calculates three feature vectors
Normalization iris image is divided into N number of image block first；
Then one group of Statistic Texture is extracted in N, and the feature of each image block is together in series indicates the image；
Secondly, compressing using PCA to high dimensional feature, its main component is excavated；
Finally, carrying out coding retrieval to textural characteristics by ITQ.
As a preferred technical solution of the present invention, detailed coding searching step in step 3 are as follows:
3.1. static texture feature extraction:
Using LBP histogram as a vector characteristic, use average LBP, average gray value and image entropy as three scalars
Feature.The texture block o of n non-overlap is divided into normalized iris imageiMark, i=1,2 ..., n, oiIn image block
Pixel pi,j, j=1,2 ..., m are indicated, choose n=96, m=64 by test；Extract oiTextural characteristics in block:
3.1.1. vector characteristic: it is directed to oiP in picture point blocki,jDecimal system LBP calculate it is as follows:
It is related to pi,jEight neighborhood pixels near point, i.e. G=8, by o in patchiAll the points be mapped to LBP histogram
In the correspondence binary system of decimal value, usually by oiHistogram defined with following formula:
Wherein, B is the binary quantity of histogram, and g is two adjacent binary system space-numbers；
B=16, g=256/16 are set；Note that the decimal value of eight bit is differed from 0 to 255, LBP histogram
The difference of adjacent local gray level pixel value is reflected；
3.1.2. average LBP, average gray value and image entropy are calculated:
Firstly, average LBP is measured in an image block, the average gray of all the points and its consecutive points is poor, can use following public affairs
Formula calculates
Second feature is the average gray value of all the points in an image block, it reflects the gray level of image, it can
To be calculated with following formula:
Third image entropy measures the cumulative of the gray value of an image block, each gray value 0-255 occurs in image block
Probability, image entropy QvV=0,1 is identified ... ..255；
nv: it is the number of gray value V in image block
It does not include the image block of iris texture for one, gray value is concentrated to a very small extent, and this feature
Value is very low, on the contrary, gray value disperses in the larger context, and this feature for the patch for being filled with iris texture simultaneously
Value with higher；
In conjunction with the above feature；LBP is indicated as iris texture characteristic, is achieved good recognition performance, is shown it
It is a kind of preferable iris texture iamge description；Simultaneously, it is contemplated that LBP histogram only extracts the local feature of image patch, adopts
Supplement auxiliary is carried out with three global scalar characterizations；In addition, used scalar characterization is mutual；For example, two image blocks
Gray level having the same, but the grey scale change between consecutive points may be different；Alternatively, two spots with similar image entropy
Block may have different grey scale changes etc.；
By connecting all pieces of feature in one image, it will be the one 1536 decimal feature tieed up, i.e. 96_16 dimension
Vector characteristics, by={ Hi | i=1,2 ..., n } is indicated and 96_3 ties up scalar characterization, by=MLBPi, MGi, IEi | i=
1,2 ..., n }, therefore, in order to improve retrieval rate, it is necessary to be compressed to these features and be converted further into binary system mould
Formula；
3.2. compressive features:
Vector characteristics and scalar characterization are compressed respectively using PCA；By taking the compression of vector characteristics as an example；Assuming that there is N number of processing
Image block, their vector characteristic vi=1,2 ..., N indicate that the covariance matrix of these iris feature images can calculate
It is as follows
The top eigenvectors k of Matrix C is used as PCA base W.Therefore, the vector characteristic of each image can be by following
Mode is compressed: KCW
CVi=W*Vi
K=64 is provided with to vector characteristics and scalar characterization；
3.3. Compression Vector feature and scalar characterization are separately encoded as binary system mould by iterative quantization method (ITQ)
Formula:
The similitude between decimal system feature and coding binary mode is kept by hyperspin feature, the rotation of matrix can
To be realized by the multiplication between matrix and orthogonal matrix；
By taking vector characteristic encodes as an example, it is assumed that the binary mode of orthogonal matrix and coding indicates by R and B respectively, we
Binary vector feature can be obtained by minimizing following quantization loss:
Wherein CV=[CV1, CV2 ..., CVN], is solved using alternating minimization algorithm, similarly, available two
System scalar characterization, and be combined using two kinds of binary features as retrieval coding；By compressing and encoding, ten are tieed up by 1536
System Feature Conversion is 128 dimension binary codes, this will greatly accelerate image retrieval speed.
Compared with prior art, the attainable beneficial effect of the present invention is:
The diminution likelihood figure of image can reduce the meter of image in the case where will not losing of feature of guarantee in the present invention
Operator workload can extract characteristics of image by three characteristic quantities of entropy of average LPB, average gray value and image faster, and
And dimensional space is less, effectively raises the retrieval rate of iris in this way.
Detailed description of the invention
Fig. 1 is that likelihood figure of the present invention generates simulation schematic diagram.
Specific embodiment
Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings, it should be understood that preferred reality described herein
Apply example only for the purpose of illustrating and explaining the present invention and is not intended to limit the present invention.
Please refer to shown in Fig. 1, the present invention provide it is a kind of from eye gray level image likelihood figure to iris image coding retrieval
Method, specific steps are as follows:
Step 1: figure is zoomed in and out be used to generate likelihood figure first:
Allow RF to become a NO emissions reduction factor, the value RD of the image of the diminution from pixel Pix-i ∈ Img (Pix-I=Pi),
The two sides stride RF are calculated based on rectangular sliding window Width (W) and L=2*RF+1 pixel and pixel, for each position of Width
It sets, the mean intensity u in window is calculated as
Calculation window intensity histogram Hist, and the corresponding pixel value d calculation formula in image is
Step 2: likelihood figure is generated:
Four different local features, i.e. mean value, standard deviation, the degree of bias and horizontal edge are generated using likelihood figure, for contracting
Each pixel in small figure RD, these features are all derived from the neighborhood centered on the pixel, are 7 pictures in a size
Mean value, standard deviation and the degree of bias are calculated on the local rectangular window of element, and serve as invariable rotary sparse edge filter；In addition, this
A little filters are also used as invariable rotary sparse edge filter, also deal with feedback to the edge of bristle part covering；
Calculating average value is first moment (M1), and (is weighted close to the higher small yin of eyelid using the complement code of downscaled images as input
Shadow)；Calculating mean standard deviation is second order moment (M2)；Calculating semiconductor body is third moment (M3)；Use Prwitt operator
Detection level edge removes part pseudo-edge, is smoothed to noise, and the reinforcing as horizontal edge in likelihood figure；
It is worth noting that, even if eyes are misaligned with camera horizon, certain parts of eyelid also due to eyelid arc property
And causing biological respinse, each feature generates a relevant figure: mean value (M1), standard deviation (M2), gradient (M3) and water
Pingbian edge (E) figure, these combine with carrying out multiplication and division point by point generates likelihood figure L and is
The high-pass filter of E, M1 and M2 and the low-pass filter of M3 are had effectively achieved, a height is used on L
For 3 pixels, the mean filter of L width 30% is covered, with the horizontal disjoint high likelihood region of connection, and is increased straight
The response of line horizontal edge portions, this operation introduce some noises, by being set as 0 for insignificant, can partially disappear
Except these noises (for example value can be changed to 0 less than maximum value 0.1% to 1%), in addition, we also will be outside likelihood eyelid region
Likelihood mapping value be set as zero；
Step 3: iris image is normalized on the basis of likelihood figure and calculates three feature vectors
Normalization iris image is divided into N number of image block first；
Then one group of Statistic Texture is extracted in N, and the feature of each image block is together in series indicates the image；
Secondly, compressing using PCA to high dimensional feature, its main component is excavated；
Finally, carrying out coding retrieval to textural characteristics by ITQ.
Specific coding searching step are as follows:
Static texture feature extraction: using LBP histogram as a vector characteristic, with average LBP, average gray value and figure
As entropy is as three scalar characterizations.The texture block o of n non-overlap is divided into normalized iris imageiMark, i=1,
2 ..., n, oiPixel p in image blocki,j, j=1,2 ..., m are indicated, choose n=96, m=64 by test；Extract oi
Textural characteristics in block:
Vector characteristic: it is directed to oiP in picture point blocki,jDecimal system LBP calculate it is as follows:
It is related to pi,jEight neighborhood pixels near point, i.e. G=8, by o in patchiAll the points be mapped to LBP histogram
In the correspondence binary system of decimal value, usually by oiHistogram defined with following formula:
Wherein, B is the binary quantity of histogram, and g is two adjacent binary system space-numbers；
B=16, g=256/16 are set；Note that the decimal value of eight bit is differed from 0 to 255, LBP histogram
The difference of adjacent local gray level pixel value is reflected；
Calculate average LBP, average gray value and image entropy: firstly, average LBP measure in an image block all the points and its
The average gray of consecutive points is poor, can be calculated with following formula
Second feature is the average gray value of all the points in an image block, it reflects the gray level of image, it can
To be calculated with following formula:
Third image entropy measures the cumulative of the gray value of an image block, each gray value 0-255 occurs in image block
Probability, image entropy QvV=0,1 is identified ... ..255；
nv: it is the number of gray value V in image block
It does not include the image block of iris texture for one, gray value is concentrated to a very small extent, and this feature
Value is very low, on the contrary, gray value disperses in the larger context, and this feature for the patch for being filled with iris texture simultaneously
Value with higher.
In conjunction with the above feature；LBP is indicated as iris texture characteristic, is achieved good recognition performance, is shown it
It is a kind of preferable iris texture iamge description；Simultaneously, it is contemplated that LBP histogram only extracts the local feature of image patch, adopts
Supplement auxiliary is carried out with three global scalar characterizations；In addition, used scalar characterization is mutual；For example, two image blocks
Gray level having the same, but the grey scale change between consecutive points may be different；Alternatively, two spots with similar image entropy
Block may have different grey scale changes etc..
By connecting all pieces of feature in one image, it will be the one 1536 decimal feature tieed up, i.e. 96_16 dimension
Vector characteristics, by={ Hi | i=1,2 ..., n } is indicated and 96_3 ties up scalar characterization, by=MLBPi, MGi, IEi | i=
1,2 ..., n }, therefore, in order to improve retrieval rate, it is necessary to be compressed to these features and be converted further into binary system mould
Formula.
Compressive features: vector characteristics and scalar characterization are compressed respectively using PCA；By taking the compression of vector characteristics as an example；Assuming that
There is N number of processing image block, their vector characteristic vi=1,2 ..., N expression, the covariance matrix of these iris feature images
It can calculate as follows
The top eigenvectors k of Matrix C is used as PCA base W.Therefore, the vector characteristic of each image can be by following
Mode is compressed: KCW
CVi=W*Vi
K=64 is provided with to vector characteristics and scalar characterization.
Compression Vector feature and scalar characterization are separately encoded as binary mode by iterative quantization method (ITQ): logical
Hyperspin feature is crossed to keep the similitude between decimal system feature and coding binary mode, the rotation of matrix can pass through matrix
Multiplication between orthogonal matrix is realized；By taking vector characteristic encodes as an example, it is assumed that the binary mode of orthogonal matrix and coding
It is indicated respectively by R and B, we can obtain binary vector feature by minimizing following quantization loss:
Wherein CV=[CV1, CV2 ..., CVN], is solved using alternating minimization algorithm, similarly, available two
System scalar characterization, and be combined using two kinds of binary features as retrieval coding；By compressing and encoding, ten are tieed up by 1536
System Feature Conversion is 128 dimension binary codes, this will greatly accelerate image retrieval speed.
The implementation of the present invention is not limited to this, and above-described embodiment content according to the invention utilizes the routine of this field
Technological know-how and customary means, under the premise of not departing from above-mentioned basic fundamental thought of the invention, preferred embodiment above can be with
The modification, replacement or combination of other diversified forms are made, other embodiments obtained all fall within rights protection scope of the present invention
Within.
Claims (2)
1. a kind of method retrieved from eye gray level image likelihood figure to iris image coding, which is characterized in that specific steps are as follows:
Step 1: figure is zoomed in and out be used to generate likelihood figure first:
RF is allowed to become a NO emissions reduction factor, the value RD of the image of the diminution is based on from pixel Pix-i ∈ Img (Pix-I=Pi)
Rectangular sliding window Width (W) and L=2*RF+1 pixel and pixel calculate the two sides stride RF, for each position of Width,
Mean intensity u in window is calculated as
Calculation window intensity histogram Hist, and the corresponding pixel value d calculation formula in image is
Step 2: likelihood figure is generated:
Four different local features, i.e. mean value, standard deviation, the degree of bias and horizontal edge are generated using likelihood figure, are schemed for reducing
Each pixel in RD, these features are all derived from the neighborhood centered on the pixel, are 7 pixels in a size
Mean value, standard deviation and the degree of bias are calculated on local rectangular window, and serve as invariable rotary sparse edge filter；In addition, these are filtered
Wave device is also used as invariable rotary sparse edge filter, also deals with feedback to the edge of bristle part covering；It calculates
Average value is first moment (M1), and (is weighted close to the higher small opacities of eyelid) using the complement code of downscaled images as input；Meter
Calculation mean standard deviation is second order moment (M2)；Calculating semiconductor body is third moment (M3)；Water is detected using Prwitt operator
Pingbian edge removes part pseudo-edge, is smoothed to noise, and the reinforcing as horizontal edge in likelihood figure；It is worth note
Meaning, even if eyes are misaligned with camera horizon, certain parts of eyelid also due to the arc property of eyelid and cause
Biological respinse, each feature generate a relevant figure: mean value (M1), standard deviation (M2), gradient (M3) and horizontal edge
(E) figure, these combine with carrying out multiplication and division point by point generates likelihood figure L and is
The high-pass filter of E, M1 and M2 and the low-pass filter of M3 are had effectively achieved, uses a height on L as 3
Pixel, the mean filter for covering L width 30% with the horizontal disjoint high likelihood region of connection, and increase straight line water
The response of flat marginal portion, this operation introduce some noises, by being set as 0 for insignificant, can partially remove this
A little noises (for example value can be changed to 0 less than maximum value 0.1% to 1%), in addition, we also by outside likelihood eyelid region seemingly
Right mapping value is set as zero；
Step 3: iris image is normalized on the basis of likelihood figure and calculates three feature vectors:
Normalization iris image is divided into N number of image block first；
Then one group of Statistic Texture is extracted in N, and the feature of each image block is together in series indicates the image；
Secondly, compressing using PCA to high dimensional feature, its main component is excavated；
Finally, carrying out coding retrieval to textural characteristics by ITQ.
2. a kind of method that iris image coding is retrieved from eye gray level image likelihood figure according to claim 1,
It is characterized in that, detailed coding searching step in step 3 are as follows:
3.1. static texture feature extraction:
Using LBP histogram as a vector characteristic, use average LBP, average gray value and image entropy as three scalar characterizations.
The texture block o of n non-overlap is divided into normalized iris imageiMark, i=1,2 ..., n, oiPixel in image block
Use pi,j, j=1,2 ..., m are indicated, choose n=96, m=64 by test；Extract oiTextural characteristics in block:
3.1.1. vector characteristic: it is directed to oiP in picture point blocki,jDecimal system LBP calculate it is as follows:
It is related to pi,jEight neighborhood pixels near point, i.e. G=8, by O in patchiAll the points be mapped to the LBP histogram decimal system
In the correspondence binary system of value, usually by OiHistogram defined with following formula:
Wherein, B is the binary quantity of histogram, and g is two adjacent binary system space-numbers；
B=16, g=256/16 are set；Note that the decimal value of eight bit is differed from 0 to 255, the reflection of LBP histogram
The difference of adjacent local gray level pixel value is gone out；
3.1.2. average LBP, average gray value and image entropy are calculated:
Firstly, average LBP is measured in an image block, the average gray of all the points and its consecutive points is poor, can use following formula meter
It calculates
Second feature is the average gray value of all the points in an image block, it reflects the gray level of image, it can be used
Following formula calculates:
Third image entropy measures the cumulative of the gray value of an image block, each gray value 0-255 occurs general in image block
Rate, image entropy QvV=0,1 is identified ... ..255；
nv: it is the number of gray value V in image block
It does not include the image block of iris texture for one, gray value is concentrated to a very small extent, and this feature value is very
It is low, on the contrary, gray value disperses in the larger context, and this feature has for the patch for being filled with iris texture simultaneously
Higher value；
In conjunction with the above feature；LBP is indicated as iris texture characteristic, is achieved good recognition performance, is shown that it is one
The preferable iris texture iamge description of kind；Simultaneously, it is contemplated that LBP histogram only extracts the local feature of image patch, using three
A overall situation scalar characterization carries out supplement auxiliary；In addition, used scalar characterization is mutual；For example, two image blocks have
Identical gray level, but the grey scale change between consecutive points may be different；Alternatively, two patches with similar image entropy can
There can be different grey scale changes etc.；
By connecting all pieces of feature in one image, it will be the one 1536 decimal feature tieed up, i.e. 96_16 dimensional vector
Feature, by={ Hi | i=1,2 ..., n } is indicated and 96_3 ties up scalar characterization, by=MLBPi, MGi, IEi | i=1,
2 ..., n }, therefore, in order to improve retrieval rate, it is necessary to be compressed to these features and be converted further into binary system mould
Formula；
3.2. compressive features:
Vector characteristics and scalar characterization are compressed respectively using PCA；By taking the compression of vector characteristics as an example；Assuming that there is N number of processing image
Block, their vector characteristic vi=1,2 ..., N indicate that the covariance matrix of these iris feature images can calculate as follows
The top eigenvectors k of Matrix C is used as PCA base W.Therefore, the vector characteristic of each image can be in the following manner
To compress: KCW
CVi=W*Vi
K=64 is provided with to vector characteristics and scalar characterization；
3.3. Compression Vector feature and scalar characterization are separately encoded as binary mode by iterative quantization method (ITQ):
The similitude between decimal system feature and coding binary mode is kept by hyperspin feature, the rotation of matrix can lead to
The multiplication crossed between matrix and orthogonal matrix is realized；
By taking vector characteristic encodes as an example, it is assumed that the binary mode of orthogonal matrix and coding indicates by R and B respectively, we can be with
Binary vector feature is obtained by minimizing following quantization loss:
Wherein CV=[CV1, CV2 ..., CVN], is solved using alternating minimization algorithm, similarly, available binary system
Scalar characterization, and be combined using two kinds of binary features as retrieval coding；By compressing and encoding, the decimal system is tieed up by 1536
Feature Conversion is 128 dimension binary codes, this will greatly accelerate image retrieval speed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811306066.8A CN109284744A (en) | 2018-11-02 | 2018-11-02 | A method of iris image is encoded from eye gray level image likelihood figure and is retrieved |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811306066.8A CN109284744A (en) | 2018-11-02 | 2018-11-02 | A method of iris image is encoded from eye gray level image likelihood figure and is retrieved |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109284744A true CN109284744A (en) | 2019-01-29 |
Family
ID=65175049
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811306066.8A Pending CN109284744A (en) | 2018-11-02 | 2018-11-02 | A method of iris image is encoded from eye gray level image likelihood figure and is retrieved |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109284744A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612786A (en) * | 2020-06-19 | 2020-09-01 | 国网湖南省电力有限公司 | Concrete defect detection method and device based on full convolution neural network and storage medium |
| CN116052313A (en) * | 2023-02-10 | 2023-05-02 | 北京中超伟业信息安全技术股份有限公司 | Intelligent secret cabinet control method, device, equipment and storage medium |
| CN116385414A (en) * | 2023-04-10 | 2023-07-04 | 哈尔滨耐是智能科技有限公司 | Component multi-scale defect feature extraction method, computing device and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011037684A1 (en) * | 2009-09-25 | 2011-03-31 | International Business Machines Corporation | Iris recognition system and method |
| CN102542243A (en) * | 2010-12-17 | 2012-07-04 | 北京理工大学 | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method |
| CN104036296A (en) * | 2014-06-20 | 2014-09-10 | 深圳先进技术研究院 | Method and device for representing and processing image |
| CN104484649A (en) * | 2014-11-27 | 2015-04-01 | 北京天诚盛业科技有限公司 | Method and device for identifying irises |
| US20160014121A1 (en) * | 2013-10-08 | 2016-01-14 | Sri International | Authorization of a financial transaction |
| CN105260698A (en) * | 2015-09-08 | 2016-01-20 | 北京天诚盛业科技有限公司 | Method and device for positioning iris image |
| US20170017843A1 (en) * | 2009-09-25 | 2017-01-19 | International Business Machines Corporation | System and method for generating and employing short length iris codes |
| US20170154202A1 (en) * | 2013-09-16 | 2017-06-01 | EyeVerify Inc. | Feature extraction and matching for biometric authentication |
| CN106845497A (en) * | 2017-01-12 | 2017-06-13 | 天津大学 | Maize in Earlier Stage image damage caused by a drought recognition methods based on multi-feature fusion |
| CN107451575A (en) * | 2017-08-08 | 2017-12-08 | 济南大学 | A kind of face anti-fraud detection method in identity authorization system |
| CN108681698A (en) * | 2018-04-28 | 2018-10-19 | 武汉大学 | A kind of extensive iris identification method with privacy protection function |
-
2018
- 2018-11-02 CN CN201811306066.8A patent/CN109284744A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011037684A1 (en) * | 2009-09-25 | 2011-03-31 | International Business Machines Corporation | Iris recognition system and method |
| US20170017843A1 (en) * | 2009-09-25 | 2017-01-19 | International Business Machines Corporation | System and method for generating and employing short length iris codes |
| CN102542243A (en) * | 2010-12-17 | 2012-07-04 | 北京理工大学 | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method |
| US20170154202A1 (en) * | 2013-09-16 | 2017-06-01 | EyeVerify Inc. | Feature extraction and matching for biometric authentication |
| US20160014121A1 (en) * | 2013-10-08 | 2016-01-14 | Sri International | Authorization of a financial transaction |
| CN104036296A (en) * | 2014-06-20 | 2014-09-10 | 深圳先进技术研究院 | Method and device for representing and processing image |
| CN104484649A (en) * | 2014-11-27 | 2015-04-01 | 北京天诚盛业科技有限公司 | Method and device for identifying irises |
| CN105260698A (en) * | 2015-09-08 | 2016-01-20 | 北京天诚盛业科技有限公司 | Method and device for positioning iris image |
| CN106845497A (en) * | 2017-01-12 | 2017-06-13 | 天津大学 | Maize in Earlier Stage image damage caused by a drought recognition methods based on multi-feature fusion |
| CN107451575A (en) * | 2017-08-08 | 2017-12-08 | 济南大学 | A kind of face anti-fraud detection method in identity authorization system |
| CN108681698A (en) * | 2018-04-28 | 2018-10-19 | 武汉大学 | A kind of extensive iris identification method with privacy protection function |
Non-Patent Citations (5)
| Title |
|---|
| H. AHMADI 等: "An Efficient Iris Coding Based on Gauss-Laguerre Wavelets", 《ICB 2007》 * |
| HUGO PROENCA 等: "Iris Biometrics: Indexing and Retrieving Heavily Degraded Data", 《IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY》 * |
| HUGO PROENCA 等: "Iris Recognition: An Entropy-Based Coding Strategy Robust to Noisy Imaging Environments", 《INTERNATIONAL SYMPOSIUM ON VISUAL COMPUTING》 * |
| 张丽红 等: "结合图像熵和分块匹配的图像检索", 《微计算机信息》 * |
| 朱慧敏: "基于LBP的加权分布纹理熵的图像检索算法", 《微型机与应用》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612786A (en) * | 2020-06-19 | 2020-09-01 | 国网湖南省电力有限公司 | Concrete defect detection method and device based on full convolution neural network and storage medium |
| CN116052313A (en) * | 2023-02-10 | 2023-05-02 | 北京中超伟业信息安全技术股份有限公司 | Intelligent secret cabinet control method, device, equipment and storage medium |
| CN116052313B (en) * | 2023-02-10 | 2024-02-23 | 北京中超伟业信息安全技术股份有限公司 | Intelligent secret cabinet control method, device, equipment and storage medium |
| CN116385414A (en) * | 2023-04-10 | 2023-07-04 | 哈尔滨耐是智能科技有限公司 | Component multi-scale defect feature extraction method, computing device and storage medium |
| CN116385414B (en) * | 2023-04-10 | 2023-11-07 | 哈尔滨耐是智能科技有限公司 | Component multi-scale defect feature extraction method, computing device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9736455B2 (en) | Method and apparatus for downscaling depth data for view plus depth data compression | |
| US10789465B2 (en) | Feature extraction and matching for biometric authentication | |
| Zahran et al. | A MODIFIED LBP METHOD TO EXTRACT FEATURES FROM COLOR IMAGES. | |
| Kwon et al. | Data-driven depth map refinement via multi-scale sparse representation | |
| US8200021B2 (en) | Image signature matching device | |
| Rosenfeld | Image pattern recognition | |
| CN109284744A (en) | A method of iris image is encoded from eye gray level image likelihood figure and is retrieved | |
| Kat et al. | Matching pixels using co-occurrence statistics | |
| Minaee et al. | Screen content image segmentation using sparse-smooth decomposition | |
| CN108717690B (en) | Method for synthesizing high dynamic range picture | |
| Paschalakis et al. | Real-time face detection and tracking for mobile videoconferencing | |
| Naresh et al. | Breast cancer detection using local binary patterns | |
| US10133956B2 (en) | Image signature extraction device | |
| Senturk et al. | Seam carving based image retargeting: A survey | |
| Shivashankar et al. | Galois field-based approach for rotation and scale invariant texture classification | |
| Singh et al. | Multiscale reflection component based weakly illuminated nighttime image enhancement | |
| Liu et al. | Segmentation by weighted aggregation and perceptual hash for pedestrian detection | |
| Dhar et al. | Interval type-2 fuzzy set and human vision based multi-scale geometric analysis for text-graphics segmentation | |
| Lefevre et al. | An extended snake model for real-time multiple object tracking | |
| Freeman et al. | Shape recipes: scene representations that refer to the image | |
| Voronin et al. | Fast texture and structure image reconstruction using the perceptual hash | |
| Hubo et al. | Self-Similarity-Based Compression of Point Clouds, with Application to Ray Tracing. | |
| Niu et al. | An Adaptive Recovering Algorithm for The Color Bar Code Image Based on Gamma Transforms | |
| S Sathiya | Texture classification with modified rotation invariant local binary pattern and gradient boosting | |
| Zhang et al. | Image Compression with Laplacian Guided Scale Space Inpainting |
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: 20190129 |
|
| RJ01 | Rejection of invention patent application after publication |