CN109086734A

CN109086734A - The method and device that pupil image is positioned in a kind of pair of eye image

Info

Publication number: CN109086734A
Application number: CN201810934979.8A
Authority: CN
Inventors: 谢波
Original assignee: Ennew Digital Technology Co Ltd
Current assignee: Ennew Digital Technology Co Ltd
Priority date: 2018-08-16
Filing date: 2018-08-16
Publication date: 2018-12-25
Anticipated expiration: 2038-08-16
Also published as: CN109086734B

Abstract

The invention discloses the method and device positioned to pupil image, method includes: S0: determining that initial pixel point is origin in eye image；S1: several rays are determined according to origin, calculate the pixel gradient amplitude for constituting each pixel of ray, the corresponding pixel of pixel gradient amplitude that will be greater than preset threshold is determined as reference image vegetarian refreshments；S2: according to the spacing distance between reference image vegetarian refreshments and origin, several suspicious marginal points are determined, and determine geometric center pixel corresponding to suspicious marginal point；S3: detect this determination geometric center pixel between the history geometric center pixel of previous determination at a distance from whether be less than setting numerical value, be to execute S5；Otherwise S4 is executed；S4: geometric center pixel is determined as new origin and executes S1；S5: the position of pupil image is marked according to suspicious marginal point fit object model of ellipse.Technical solution of the present invention more accurate can position pupil image.

Description

The method and device that pupil image is positioned in a kind of pair of eye image

Technical field

The present invention relates to the sides that pupil image in field of computer technology more particularly to a kind of pair of eye image is positioned Method and device.

Background technique

When realizing recognition of face or iris recognition, it usually needs the pupil image carried to eye image positions, i.e., The position of emergent pupil hole image is marked in eye image.

Currently, generalling use vertical direction shade of gray integral when the pupil image carried to eye image positions Projection pattern and horizontal direction shade of gray integral projection mode determine several edges of pupil image in eye image Geometric center corresponding to each marginal point is determined as pupil center, and is marked according to pupil center and each marginal point by point The position of pupil image in eye image.

In above-mentioned technical proposal, there may be the point of much noise caused by eyelashes, eyelid in each marginal point for determining, It is excessive to will lead to real center gap of the determining pupil center apart from pupil image, according to determining pupil center and each side When edge point marks the position of pupil image, then the pupil image that can not be carried to eye image is accurately positioned.

Summary of the invention

The present invention provides the method and device that pupil image is positioned in a kind of pair of eye image, pair that can be more accurate The pupil image carried in eye image is positioned.

In a first aspect, the present invention provides the methods that pupil image in a kind of pair of eye image is positioned, comprising:

S0: initial pixel point is determined in eye image to be processed, and using the initial pixel point as origin；

S1: several rays are determined in the eye image to be processed according to the origin, calculates and constitutes each The pixel gradient amplitude of each pixel of the ray, and each pixel gradient amplitude that will be greater than preset threshold is divided Not corresponding pixel is determined as reference image vegetarian refreshments；

S2: according to each reference image vegetarian refreshments spacing distance between the origin respectively, from each reference Several suspicious marginal points are determined in pixel, and determine geometric center pixel corresponding to each suspicious marginal point Point；

S3: it detects between the geometric center pixel of this determination and the history geometric center pixel of previous determination Distance value whether be less than setting numerical value, if so, execute S5；Otherwise, S4 is executed；

S4: being determined as history geometric center pixel for the geometric center pixel, by the geometric center pixel It is determined as new origin, and executes S1；

S5: according to each suspicious marginal point fit object model of ellipse, institute is marked by the Target ellipse model State the position of pupil image in eye image.

Preferably,

It is described according to each suspicious marginal point fit object model of ellipse, pass through the Target ellipse model and mark institute State the position of pupil image in eye image, comprising:

A0: initial model of ellipse is fitted according to each suspicious marginal point；

A1: 1 sampled edge points are randomly choosed from each suspicious marginal point, are determined described at least three Sampled edge point respectively corresponds the sampling tangent line in the initial model of ellipse, and according to each sampled edge point and respectively Sampling tangent line described in item determines sampling pupil center；

A2: for each the non-sampled marginal point for being not selected for sampled edge point in each suspicious marginal point, It should be in the tangent line, each described in the initial model of ellipse according to the non-sampled marginal point, the non-sampled marginal point pair Sampling tangent line described in sampled edge point and each item determines calibration pupil center；

A3: being directed to each described calibration pupil center, when the calibration pupil center and the sampling pupil center When spacing distance is not more than set distance, non-sampled marginal point corresponding to the calibration pupil center is determined as credible edge Point；

A4: it determines between the first total amount of each credible marginal point and the second total amount of each suspicious marginal point Ratio, and detect whether the ratio is less than given threshold, if so, executing A1；Otherwise, A5 is executed；

A5: according to each credible marginal point fit object model of ellipse, institute is marked by the Target ellipse model State the position of pupil image in eye image.

Preferably,

The sampling tangent line according to each sampled edge point and each item determines sampling pupil center, comprising:

According to the initial model of ellipse, the midpoint between the every two adjacent sampled edge point is determined；

It is directed to each described midpoint, determines that two sampled edge points for corresponding to the midpoint are respectively corresponded The sampling tangent line intersection point, and determine the straight line where the midpoint and the intersection point；

According to least square method calculate each item of distance in the eye image described in the nearest apsis of linear distance, and The apsis is determined as to sample pupil center.

Preferably,

It is described according to each credible marginal point fit object model of ellipse, comprising:

B0: credible set is formed using each credible marginal point；

B1: transition model of ellipse is fitted according to each credible marginal point for including in the credible combination；

B2: calculate each the described credible marginal point for including in the credible set respectively with the transition model of ellipse Between algebraic distance；

B3: according to the credible total amount meter for the credible marginal point for including in each algebraic distance and the credible set Average fit deviation is calculated, according to the average fit deviation determination deviation threshold value；

B4: being directed to each described suspicious marginal point, detect the suspicious marginal point and the transition model of ellipse it Between algebraic distance, when the algebraic distance between the suspicious marginal point and the transition model of ellipse be greater than the deviation threshold When, the suspicious marginal point is determined as noise spot, and formed newly using each marginal point for being not determined to noise spot Credible set；

B5: whether credible set and the credible set of previous formation for detecting this formation are identical, if so, executing B6；Otherwise, B1 is executed；

B6: corresponding transition model of ellipse is determined as Target ellipse model when this is formed credible set.

Preferably,

It is described according to each reference image vegetarian refreshments spacing distance between the origin respectively, from each reference Several suspicious marginal points are determined in pixel, comprising:

Detect each reference image vegetarian refreshments spacing distance between the origin respectively；

Calculate the expected value and standard deviation of each spacing distance；

Several target interval distances are extracted according in each spacing distance of the desired value and standard deviation institute；

Each target interval is determined as suspicious marginal point apart from the corresponding reference image vegetarian refreshments of institute.

Preferably,

Before the initial pixel point determining in eye image to be processed, further comprise:

Acquire video image；

Each frame original image of the video image is filtered respectively to obtain original graph described in each frame As a corresponding pretreatment image；

Extract the human eye figure to be processed that presently described pretreatment image carries respectively from each Zhang Suoshu pretreatment image Picture.

Preferably,

Before the initial pixel point determining in eye image to be processed, further includes:

A non-selected eye image to be processed is successively selected, and determines the eye image to be processed of selection in video Corresponding frame number in image；

It is then, described that initial pixel point is determined in eye image to be processed, comprising:

When the frame number is 1, the geometric center of the eye image to be processed of selection is determined as initial pixel Point；Or, when the frame number is greater than 1, according to the pupil of pupil image entrained by the eye image to be processed of previous selection Initial pixel point is determined at center in the eye image to be processed.

Preferably,

The S5 further comprises, the geometric center pixel is determined as carrying in the eye image to be processed The pupil center of pupil image.

Second aspect, the present invention provides the devices that pupil image in a kind of pair of eye image is positioned, comprising:

Initial point determining module, for determining initial pixel point in eye image to be processed, and by the initial pixel Point is used as origin；

Endpoint detections module, for determining that several are penetrated in the eye image to be processed according to the origin Line calculates the pixel gradient amplitude for constituting each pixel of ray described in each, and will be greater than each institute of preset threshold It states the corresponding pixel of pixel gradient amplitude institute and is determined as reference image vegetarian refreshments；

Noise spot filtering module, for according to each reference image vegetarian refreshments interval distance between the origin respectively From determining several suspicious marginal points from each reference image vegetarian refreshments, and determine that each suspicious marginal point institute is right The geometric center pixel answered；

Spot detection module, for detecting the geometric center pixel of this determination and the history geometry of previous determination Whether the distance between central pixel point value is less than setting numerical value, if so, trigger model fitting module；Otherwise, flip-flop transition Processing module；

The transition processing module will for the geometric center pixel to be determined as history geometric center pixel The geometric center pixel is determined as new origin, and triggers the endpoint detections module；

The models fitting module is used for according to each suspicious marginal point fit object model of ellipse, by described Target ellipse model marks the position of pupil image in the eye image.

Preferably,

The models fitting module, comprising: pretreatment unit, sample processing unit, calibration process unit, trusted processes list Member, detection processing unit and label processing unit；Wherein,

The pretreatment unit, for being fitted initial model of ellipse according to each suspicious marginal point；

The sample processing unit, for randomly choosing at least three sampled edges from each suspicious marginal point Point determines that the 1 sampled edge points respectively correspond the sampling tangent line in the initial model of ellipse, and according to each Sampling tangent line described in a sampled edge point and each item determines sampling pupil center；

The calibration process unit is not selected for the every of sampled edge point for being directed in each suspicious marginal point One non-sampled marginal point, should be in the initial model of ellipse according to the non-sampled marginal point, the non-sampled marginal point pair On tangent line, sampling tangent line determines calibration pupil center described in each sampled edge point and each item；

The trusted processes unit, for being directed to each described calibration pupil center, when the calibration pupil center and When the spacing distance of the sampling pupil center is not more than set distance, by non-sampled side corresponding to the calibration pupil center Edge point is determined as credible marginal point；

The detection processing unit, for determine each credible marginal point the first total amount and each suspicious side Ratio between second total amount of edge point, and detect whether the ratio is less than given threshold, if so, triggering at the sampling Manage unit；Otherwise, the meter processing unit is triggered；

The meter processing unit is used for according to each credible marginal point fit object model of ellipse, by described Target ellipse model marks the position of pupil image in the eye image.

Preferably,

Further include: image capture module, filtering processing module and image zooming-out module；Wherein,

Described image acquisition module, for acquiring video image；

The filtering processing module, for each frame original image to the video image be filtered respectively with Obtain the corresponding pretreatment image of the institute of original image described in each frame；

Described image extraction module, for extracting presently described pretreatment figure respectively from each Zhang Suoshu pretreatment image As the eye image to be processed carried.

The present invention provides the method and device that pupil image in a kind of pair of eye image is positioned, in this method, First stage, first using an initial pixel point in eye image to be processed as origin, then according to origin to be processed A plurality of ray is determined in eye image, the pixel gradient amplitude for constituting each pixel of each ray is calculated, due to people Pixel gradient amplitude in eye image between the marginal point and other pixels of pupil image is relatively large, therefore can will be greater than pre- If the corresponding pixel of each pixel gradient amplitude institute of threshold value is determined as reference image vegetarian refreshments, that is, realize from eye image The reference image vegetarian refreshments that a part may be pupil image marginal point is filtered out, it is subsequent, due to noise spot in eye image and original Spacing distance between point is larger relative to the distance between pupil image true edge point and origin gap, therefore can be according to each A reference image vegetarian refreshments spacing distance between origin respectively, determines several suspicious edges from each reference image vegetarian refreshments Point, realize go be not in reference pixel point unless each pupil image marginal point a part of noise spot, determining removal one Point noise spot and after the geometric center pixel of remaining each suspicious marginal point, then can be using the geometric center point as one New origin repeats the similar method of aforementioned each step, when this and previous determining geometric center pixel respectively The distance between value be less than setting numerical value, that is, illustrate it is adjacent twice determine geometric center pixel when, two groups of suspicious marginal point institutes Corresponding pupil center's (geometric center pixel) extremely approaches or coincides in eye image in the pupil of pupil image The heart also illustrates in each suspicious marginal point corresponding when this determines geometric center pixel without apparent noise Point, realization substantially reduce the noise spot being mixed into each suspicious marginal point；It is quasi- according to each suspicious marginal point in second stage Close Target ellipse model, by Target ellipse model mark eye image in pupil image position, be no longer dependent on it is each can Geometric center pixel corresponding to marginal point is doubted, residue (and a small amount of) noise spot being mixed into each suspicious marginal point is reduced To influence caused by positioning result；In conclusion largely reducing the noise spot being mixed into suspicious marginal point by the first stage Quantity, while a small amount of noise spot for being mixed into suspicious marginal point is reduced to the influence degree of positioning result by second stage, More accurate the pupil image carried in eye image can be positioned.

Detailed description of the invention

It in order to illustrate the embodiments of the present invention more clearly or existing technical solution, below will be to embodiment or the prior art Attached drawing needed in description is briefly described, it should be apparent that, the accompanying drawings in the following description is only in the present invention The some embodiments recorded without any creative labor, may be used also for those of ordinary skill in the art To obtain other drawings based on these drawings.

The process for the method that pupil image is positioned in a kind of pair of eye image that Fig. 1 provides for one embodiment of the invention Schematic diagram；

It is a kind of according to the sampled edge point of selection and the initial elliptical modes of fitting that Fig. 2 is that one embodiment of the invention provides Type determines the schematic diagram of sampling pupil center；

The structure for the device that pupil image is positioned in a kind of pair of eye image that Fig. 3 provides for one embodiment of the invention Schematic diagram；

Fig. 4 is the apparatus structure that the another kind that one embodiment of the invention provides positions pupil image in eye image Schematic diagram；

Fig. 5 is the structural schematic diagram of the electronic equipment provided in one embodiment of the invention.

Specific embodiment

To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with specific embodiment and accordingly Technical solution of the present invention is clearly and completely described in attached drawing.Obviously, described embodiment is only a part of the invention Embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making wound Every other embodiment obtained under the premise of the property made labour, shall fall within the protection scope of the present invention.

As shown in Figure 1, the embodiment of the invention provides the method that pupil image in a kind of pair of eye image is positioned, packet It includes:

Embodiment as shown in Figure 1 in the first stage first makees an initial pixel point in eye image to be processed For origin, a plurality of ray is then determined in eye image to be processed according to origin, calculates and constitutes each of each ray The pixel gradient amplitude of pixel, due to the pixel gradient in eye image between the marginal point of pupil image and other pixels Amplitude is relatively large, therefore the corresponding pixel of each pixel gradient amplitude institute that can will be greater than preset threshold is determined as joining Pixel is examined, that is, realizes the reference image vegetarian refreshments for filtering out that a part may be pupil image marginal point from eye image, it is subsequent , since the spacing distance in eye image between noise spot and origin is relative between pupil image true edge point and origin Range difference away from larger, therefore can be according to each reference image vegetarian refreshments spacing distance between origin respectively, from each reference image It determines several suspicious marginal points in vegetarian refreshments, realizes that going is not one of pupil image marginal point in reference pixel point unless each Point noise spot, determine to remove a part of noise spot and after the geometric center pixel of remaining each suspicious marginal point, Origin that then can be new as one using the geometric center point, repeats the similar method of aforementioned each step, when this and it is preceding The distance between the secondary geometric center pixel determined respectively value is less than setting numerical value, that is, illustrates adjacent geometric center determining twice When pixel, the corresponding pupil center's (geometric center pixel) of two groups of suspicious marginal point institutes extremely approaches or coincides with people The pupil center of pupil image, also illustrates each suspicious edge corresponding when this determines geometric center pixel in eye image Without apparent noise spot in point, realization substantially reduces the noise spot being mixed into each suspicious marginal point；In second-order Section marks pupil image in eye image by Target ellipse model according to each suspicious marginal point fit object model of ellipse Position, be no longer dependent on geometric center pixel corresponding to each suspicious marginal point, reduce and be mixed into each suspicious edge Residue (and a small amount of) noise spot in point is to influence caused by positioning result；In conclusion being largely reduced by the first stage It is mixed into the quantity of the noise spot in suspicious marginal point, while a small amount of noise being mixed into suspicious marginal point is reduced by second stage Point more accurate can position the influence degree of positioning result to the pupil image carried in eye image.

In above-described embodiment, gradient magnitude corresponding to pixel is referred specifically to along the ray direction of propagation, the latter picture Absolute difference between the pixel value of vegetarian refreshments and the pixel value of previous pixel；For example, with the eye image of 9*9 pixel For, if origin is (5,5), it is directed to the ray propagated horizontally to the right determined according to the origin, if in eye image The pixel for constituting the ray is successively (5,5), (5,6), (5,7), (5,8), (5,9), is directed to a pixel (5,6), Its pixel gradient amplitude is exhausted between the pixel value of pixel in eye image (5,6) and the pixel value of pixel (5,5) To difference.

It will be apparent that history geometric center pixel determined by previous cycle is this when circulation executes S1 to S4 Origin described in step S1 when secondary circulation is to determine geometric center pixel.

It is described according to each suspicious edge in a preferred embodiment of the invention based on embodiment as shown in Figure 1 Point fit object model of ellipse, the position of pupil image in the eye image is marked by the Target ellipse model, comprising:

In the embodiment, initial model of ellipse is fitted according to each suspicious marginal point first, it then then can be from each suspicious 1 sampled edge points are randomly selected in marginal point, and determine that each sampled edge point is respectively corresponded initial oval Sampling tangent line on model, and then determine that each sampled edge point institute is right according to each sampled edge point and each item sampling tangent line The sampling pupil center answered；In subsequent process, it is directed in each suspicious marginal point and is not selected for each of sampled edge point A non-sampled marginal point, then can according to the non-sampled marginal point, non-sampled marginal point pair should in initial model of ellipse tangent line, Each sampled edge point and each item sampling tangent line determine calibration pupil center；It calibrates between pupil center and sampling pupil center Spacing distance is smaller, then explanation when each sampled edge point selected is the true edge point of pupil image in eye image, Non-sampled marginal point corresponding to the calibration pupil center is that the probability of the true edge point of pupil image in eye image is higher, Conversely, then illustrating that non-sampled marginal point corresponding to the calibration pupil center is the true edge point of pupil image in eye image Probability it is lower, therefore, for each determining calibration pupil center, when between calibration pupil center and sampling pupil center When gauge is from set distance is not more than, then non-sampled marginal point corresponding to calibration pupil center can be determined as credible edge Point, it is clear that, be not determined to the non-sampled marginal point of credible marginal point relative to selection each sampled edge point i.e. For noise spot；Since noise spot has largely been removed in the first phase in determining each suspicious marginal point, it is mixed into each The noise spot quantity of suspicious marginal point should be far fewer than the true edge point quantity of pupil image in eye image, therefore, true It makes between the first total amount of the credible marginal point of whole and the second total amount of each marginal point relative to each sampled edge point Ratio when, the ratio is bigger, then illustrates each sampled edge point selected for the true edge of pupil image in eye image The probability of point is higher, conversely, then illustrating each sampled edge point selected for the true edge point of pupil image in eye image Probability it is lower, correspondingly, when determining ratio be less than given threshold when, then sampled edge point can be reselected and be held The aforementioned similar method of row is further to remove mixed noise spot in suspicious marginal point unless each, until determining ratio is not less than When given threshold, just according to remaining each credible marginal point fit object model of ellipse after removing noise spot again to mark people The position of pupil image in eye image, to reduce influence journey of the noise spot to positioning result being mixed into each suspicious marginal point Degree, realizes that the more accurate pupil image in eye image positions.

It should be noted that randomly choosing 1 sampled edge points from determining each suspicious marginal point and carrying out When subsequent business processing, the quantity of the sampled edge point of selection is more, then the calculation amount being related to is bigger, therefore, in order to reduce Calculation amount accurately positions the pupil image carried in eye image so as to more quick, every time selection sampling side The quantity of edge point can be 3.

It should be noted that the corresponding sampling in initial model of ellipse of the sampled edge point as described in the examples is cut Specifically there is the following two kinds situation A and B in line:

A, sampled edge point are located in the initial model of ellipse of fitting, at this point, sampled edge point is corresponding in initial elliptical modes Sampling tangent line in type refers to the tangent line of the sampled edge point in initial model of ellipse.

B, sampled edge point is not in the initial model of ellipse of fitting, at this time, it may be necessary to determine the initial ellipse of fitting A nearest low coverage point of the algebraic distance of the distance sampled edge point, sampled edge point are corresponding in initial elliptical modes on circle model Sampling tangent line in type refers to tangent line of the low coverage point corresponding with the sampled edge point in initial model of ellipse.

It is described according to each sampled edge point and each item institute in one embodiment of the invention based on previous embodiment It states sampling tangent line and determines sampling pupil center, comprising:

It is implemented in step A1 by the method that the embodiment provides and is cut according to each sampled edge point and the sampling of each item When line determines sampling pupil center, if each sampled edge point of selection is the true of the pupil image carried in eye image Marginal point, the sampling pupil center determined then extremely should approach or coincide with the true pupil center of pupil image, so as to subsequent Mixed noise spot in suspicious marginal point unless each is removed in the process and extracts the true edge point that greater probability is pupil image Credible marginal point.Specifically, it is determined based on each sampled edge point of selection by similar method in subsequent process When each non-sampled corresponding calibration pupil center of marginal point institute, calibration pupil center corresponding to non-sampled marginal point The distance between the sampling pupil center is closer, then illustrates that the sampled edge point is the probability of the true edge point of pupil image It is higher, conversely, then illustrate non-sampled marginal point be noise spot probability it is higher.

For example, referring to FIG. 2, by for extracting three sampled edge points M1, M2, M3 in each each marginal point, Determining that M1, M2, M3 respectively correspond after the sampling tangent line in initial model of ellipse, it may be determined that go out M1 and M2 it is right respectively Answer sampling tangent line intersection point P2, M1 and M3 institute it is corresponding sampling tangent line intersection point P1 and M2 and M3 it is corresponding Sample the intersection point P3 of tangent line；It can determine that M1 adjacent with M2, M2 according to initial model of ellipse and M3 be adjacent, M1 and M3 are adjacent, it can Further determine that out the midpoint X point between adjacent sampled edge point M1 and M2, between adjacent sampled edge point M1 and M3 Midpoint Z point between midpoint Y point and adjacent sampled edge point M2 and M3；Correspondingly, being directed to midpoint X, it may be determined that corresponding In midpoint X two sampled edge points M1, M2 corresponding sampling tangent line intersection point P2, and then determine midpoint X and friendship Straight line L2 where point P2 can determine that straight line straight line L1 and midpoint Z where midpoint Y and intersection point P1 by similar method And straight line L3 where intersection point P3；In subsequent process, then can be calculated by least square method or other algorithms in eye image away from The apsis nearest from straight line L1, L2 and L, the apsis are sampling pupil center corresponding to sampled edge point M1, M2, M3 O。

It should be noted that Tu2Zhong pupil center is the intersection point of L1, L2, L3, and in practical business scene, L1, L2, L3 Common intersection point may and be not present, at this point, needing to calculate distance in eye image by least square method or other algorithms Straight line L1, L2 and L nearest apsis.

Specifically, it is based on previous embodiment, it is described according to each credible edge in a kind of mode in the cards Point fit object model of ellipse, comprising:

B0: credible set is formed using each credible marginal point；

In the implementation, when circulation executes B1 to B5, this recycles average fit deviation calculated can be with this Secondary circulation is formed by transition model of ellipse and combines independently to measure each marginal point relative to the transition model of ellipse and be No is noise spot；It specifically, can be according to average fit deviation determination deviation threshold value, when a suspicious marginal point and transition elliptical modes When algebraic distance between type is greater than the deviation threshold, then illustrate the transition elliptical modes that the suspicious marginal point is formed relative to this Type is noise spot, conversely, can then determine it as credible marginal point (determines it as the true of pupil image in eye image Marginal point), so, it can be achieved that by a part in each suspicious marginal point by the suspicious marginal point weight for being defined as noise spot of mistake Newly it is determined as credible marginal point, and determines by the noise spot for being determined as credible marginal point of mistake, thus using not true The each suspicious marginal point for being set to noise spot forms new credible set；New credible set circulation execution for formation is aforementioned Method then illustrates that this is recycled in obtained credible set when the adjacent credible set formed twice is identical The noise spot in each suspicious marginal point is accurately eliminated through complete, each credible side in this credible set formed Edge point should all be the true edge point of pupil image in eye image, correspondingly, corresponding when by this credible set of formation Transition model of ellipse be determined as Target ellipse model, and the pupil carried in eye image is marked by the Target ellipse model It is when image, then more efficiently to avoid noise spot influence caused by positioning result, it can be more accurately in eye image The pupil image of carrying is positioned.

In the embodiment, average fit deviation refers specifically to each credible marginal point in the credible set of this formation The average value of the algebraic distance between transition model of ellipse formed respectively with this.

In the embodiment, when forming deviation threshold each time, this deviation threshold formed is usually the flat of this calculating 1 to 2 times of equal fitness bias, specific multiple is adjusted in combination with practical business scene.

In a preferred embodiment of the invention, being mixed into order to prevent can be still excessive with the noise spot in marginal point, nothing Method quickly select it is multiple with high probability be pupil image in eye image true edge point sampled edge point, thus Lead to not quickly position the pupil image carried in eye image, in a preferred embodiment of the invention, further includes:

The ratio that record determines each time, and each ratio that record continuously determines continuously are less than described set Determine the cycle-index of threshold value；

When the cycle-index reaches setting numerical value, the maximum mesh of numerical value is selected from each ratio of record Mark ratio；

Corresponding each credible marginal point fit object model of ellipse, passes through when the target ratio according to determination The Target ellipse model marks the position of pupil image in the eye image.

In the embodiment, by the determining each ratio of record, and records each ratio continuously determined and be continuously less than and set The cycle-index for determining threshold value is the true side of pupil image in eye image when continuously failing to select multiple with high probability When the number of the sampled edge point of edge point reaches setting numerical value, that is, each ratio continuously determined is continuously less than following for given threshold When ring number reaches setting numerical value, due to ratio size determining each time, directly reacts this and determine corresponding choosing when ratio The each sampled edge point selected is the probability height of the true edge point of pupil image in eye image, therefore, can be from record The maximum target ratio of numerical value is selected in each ratio, corresponding each credible marginal point when according to the determining target ratio Fit object model of ellipse, and then by the position of pupil image in Target ellipse model label eye image, it avoids longer Time in can not select suitable multiple sampled edge points and lead to not to the pupil image carried in eye image into Row quickly positioning.

It is described according to each reference pixel in a preferred embodiment of the invention based on embodiment as shown in Figure 1 The spacing distance between the origin respectively is put, several suspicious marginal points are determined from each reference image vegetarian refreshments, Include:

Calculate the expected value and standard deviation of each spacing distance；

In the embodiment, due in practical business scene, the quantity meeting of noise spot in each reference image vegetarian refreshments of detection Far below the quantity of the true edge point of pupil image, and the spacing distance between most noise spots and origin is relative to pupillogram Spacing distance between each true edge point and origin of picture differs greatly, and the different true edge points minute of pupil image Spacing distance difference not between origin is smaller, by detecting each reference image vegetarian refreshments interval distance between origin respectively From, and the expected value and standard deviation of each spacing distance is calculated, in the data set be made of each spacing distance, if one is worked as Space before distance is bigger on the influence of the dispersion degree of the data set, then illustrates the present interval apart from corresponding reference image vegetarian refreshments It is higher for the probability of noise spot, therefore, it can be determined in each spacing distance according to the expected value and standard deviation of calculating to discrete More noisy interval distance extracts each target interval distance for being not determined to noisy interval distance, and will be each The corresponding reference image vegetarian refreshments of a target interval distance institute as suspicious marginal point, go mixed in reference pixel point unless each by realization The much noise point entered.

For example, by circulation execute a S1 to S4 for, calculate reference image vegetarian refreshments A, B, C, D, E respectively with origin Spacing distance between O is successively to calculate the desired value and standard deviation of these spacing distances for 19,20,21,20,35 Later, the standard deviation obtained is greater than corresponding reference value, then illustrates that the dispersion degree of each spacing distance is excessively high, that is, illustrate each There may be noise spots in reference image vegetarian refreshments；At this point, according to the difference between the desired value of calculating and each spacing distance, it can be true Making to the spacing distance that the dispersion degree of data set is affected includes " 35 ", i.e. reference image corresponding to spacing distance " 35 " Vegetarian refreshments is that the probability of noise spot is higher, can remove noisy interval distance " 35 ", extracts its for being not determined to noisy interval distance His four target interval distances, and this corresponding reference image vegetarian refreshments of four target intervals distance institutes is determined as suspicious edge Point.

In general, by taking the desired value u and standard deviation δ that calculate these distance values as an example, can specifically remove with origin it Between spacing distance be greater than u+1.5 δ or the reference image vegetarian refreshments less than u-1.5 δ, each reference image vegetarian refreshments not being removed is Suspicious marginal point.

Based on aforementioned any embodiment, in a preferred embodiment of the invention, it is described in eye image to be processed really Before determining initial pixel point, further comprise:

Acquire video image；

By in the embodiment, it can be achieved that being taken respectively to each frame original image in the video image image of continuous acquisition The pupil image of band is positioned.

In the embodiment, the sample image for Face datection and human eye detection can be acquired in advance, in conjunction with sample image Haar-Like rectangular characteristic and Adaboost algorithm train Face datection model and human eye detection model；Collecting view After frequency image, due in the image capture devices such as video camera in the sole mass problem and shooting environmental of each electronic component The influence of intensity of illumination, there may be a degree of external noises in each frame original image of video image, therefore, It can be pre-processed first against each frame original image of video image, original image can be specifically filtered, It realizes the external noises such as removal additivity, the spiced salt, Gauss, avoids these noises from covering the raw information of image, to enhance original The signal-to-noise ratio of image obtains the corresponding pretreatment image of each frame original image institute；It further, then can be for obtaining Each pretreatment image orients face location (i.e. in current original image using trained Face datection model first Realization marks position of the facial image in current pretreatment image), then determined using trained human eye detection model Human eye portion is located at the position on the facial image of label, to extract eye image entrained by current pretreatment image (i.e. Eye image to be processed).

Further, when needing at the corresponding eye image of frame original image each in video image institute Reason to extract the suspicious marginal point of the pupil image carried respectively in multiple eye images, and is realized and is determined pupil image When position, in order to save the processing time, treatment effeciency is improved, in one embodiment of the invention, the S5 further comprises: will be described Geometric center pixel is determined as the pupil center of the pupil image carried in the eye image to be processed.

For example, when needing the pupil image carried to an eye image to be processed to position, if to be processed Original image corresponding to eye image is the first frame of the video image of acquisition, in step s 5 further by geometric center picture Vegetarian refreshments is determined as after the pupil center of the pupil image carried in eye image to be processed, with the pupil center in people to be processed For corresponding pixel coordinate is (x, y) in eye image, when needing to work as to corresponding to the second frame original image in video image Preceding eye image is handled, and when with the suspicious marginal point of the pupil image carried in the current eye image of determination, then can will be worked as Pixel (x, y) in preceding eye image is determined as initial pixel point, which levels off in current eye image The pupil center of the pupil image carried in current eye image regard the pixel (x, y) in current eye image as origin, It, can suspicious side that is more quick and accurately determining the pupil image that current eye image carries and when carrying out subsequent processing Edge point.

Correspondingly, determining initial pixel point in eye image to be processed described in a preferred embodiment of the invention Before, further includes:

It should be noted that in order to avoid what the eyelashes and eyelid of user were formed in eye image extremely levels off to pupil The noise spot of bore edges point by a large amount of and mistake is determined as reference pixel point, when frame number is greater than 1, in eye image to be processed In determine initial pixel point after, in a preferred embodiment of the invention, by the initial pixel in facial image to be processed Point is as origin and when determining several rays, the deviation angle position of each ray drawn by the origin in the horizontal direction In -30 degree between 30 degree and 150 degree to 210 degree.

Based on design identical with embodiment of the present invention method, referring to FIG. 3, the embodiment of the invention provides a kind of couple of people The device that pupil image is positioned in eye image, comprising:

Initial point determining module 301, for determining initial pixel point in eye image to be processed, and by the initial picture Vegetarian refreshments is as origin；

Endpoint detections module 302, for determining several in the eye image to be processed according to the origin Ray calculates the pixel gradient amplitude for constituting each pixel of ray described in each, and will be greater than each of preset threshold The corresponding pixel of pixel gradient amplitude institute is determined as reference image vegetarian refreshments；

Noise spot filtering module 303, for according to each reference image vegetarian refreshments interval between the origin respectively Distance determines several suspicious marginal points from each reference image vegetarian refreshments, and determines each suspicious marginal point institute Corresponding geometric center pixel；

Spot detection module 304, the history of the geometric center pixel and previous determination for detecting this determination Whether the distance between geometric center pixel value is less than setting numerical value, if so, trigger model fitting module 306；Otherwise, it touches Send out transition processing module 305；

The transition processing module 305, for the geometric center pixel to be determined as history geometric center pixel, The geometric center pixel is determined as new origin, and triggers the endpoint detections module；

The models fitting module 306, for passing through institute according to each suspicious marginal point fit object model of ellipse State the position that Target ellipse model marks pupil image in the eye image.

In a preferred embodiment of the invention, the models fitting module 306, comprising: pretreatment unit, sampling processing list Member, calibration process unit, trusted processes unit, detection processing unit and label processing unit；Wherein,

Referring to FIG. 4, in a preferred embodiment of the invention, described device further include: image capture module 401, filtering Processing module 402 and image zooming-out module 403；Wherein,

Described image acquisition module 401, for acquiring video image；

The filtering processing module 402 is filtered place for each frame original image to the video image respectively Reason is to obtain the corresponding pretreatment image of the institute of original image described in each frame；

Described image extraction module 403, for from extracted respectively in each Zhang Suoshu pretreatment image it is presently described it is pre- from Manage the eye image to be processed that image carries.

Fig. 5 is the structural schematic diagram of one embodiment of the present of invention electronic equipment.In hardware view, which includes Processor, optionally further comprising internal bus, network interface, memory.Wherein, memory may include memory, such as high speed Random access memory (Random-Access Memory, RAM), it is also possible to further include nonvolatile memory (non- Volatile memory), for example, at least 1 magnetic disk storage etc..Certainly, which is also possible that other business institutes The hardware needed.

Processor, network interface and memory can be connected with each other by internal bus, which can be ISA (Industry Standard Architecture, industry standard architecture) bus, PCI (Peripheral Component Interconnect, Peripheral Component Interconnect standard) bus or EISA (Extended Industry Standard Architecture, expanding the industrial standard structure) bus etc..The bus can be divided into address bus, data/address bus, control always Line etc..Only to be indicated with a four-headed arrow in Fig. 5, it is not intended that an only bus or a type of convenient for indicating Bus.

Memory, for storing program.Specifically, program may include program code, and said program code includes calculating Machine operational order.Memory may include memory and nonvolatile memory, and provide instruction and data to processor.

In a kind of mode in the cards, processor read from nonvolatile memory corresponding computer program to It is then run in memory, corresponding computer program can also be obtained from other equipment, to be formed on logic level to human eye The device that pupil image is positioned in image.Processor executes the program that memory is stored, with real by the program executed The device that pupil image in eye image is positioned provided in existing any embodiment of the present invention.

The above-mentioned device positioned to pupil image in eye image provided such as embodiment illustrated in fig. 5 of the present invention executes Method can be applied in processor, or realized by processor.Processor may be a kind of IC chip, have letter Number processing capacity.During realization, each step of the above method can pass through the integration logic electricity of the hardware in processor The instruction of road or software form is completed.Above-mentioned processor can be general processor, including central processing unit (Central Processing Unit, CPU), network processing unit (Network Processor, NP) etc.；It can also be Digital Signal Processing Device (Digital Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array (Field-Programmable Gate Array, FPGA) or other can Programmed logic device, discrete gate or transistor logic, discrete hardware components.It may be implemented or execute present invention implementation Disclosed each method, step and logic diagram in example.General processor can be microprocessor or the processor can also be with It is any conventional processor etc..

The step of method in conjunction with disclosed in the embodiment of the present invention, can be embodied directly in hardware decoding processor and execute At, or in decoding processor hardware and software module combination execute completion.Software module can be located at random access memory, This fields such as flash memory, read-only memory, programmable read only memory or electrically erasable programmable memory, register maturation In storage medium.The storage medium is located at memory, and processor reads the information in memory, completes above-mentioned side in conjunction with its hardware The step of method.

The embodiment of the present invention also proposed a kind of computer readable storage medium, the computer-readable recording medium storage one A or multiple programs, the one or more program include instruction, which holds when by the electronic equipment including multiple application programs When row, it can make what is provided in electronic equipment execution any embodiment of the present invention to position to pupil image in eye image Method, and be specifically used for executing method as shown in Figure 1.

System, device, module or the unit that above-described embodiment illustrates can specifically realize by computer chip or entity, Or it is realized by the product with certain function.It is a kind of typically to realize that equipment is computer.Specifically, computer for example may be used Think personal computer, laptop computer, cellular phone, camera phone, smart phone, personal digital assistant, media play It is any in device, navigation equipment, electronic mail equipment, game console, tablet computer, wearable device or these equipment The combination of equipment.

For convenience of description, it describes to be divided into various units when apparatus above with function or module describes respectively.Certainly, exist Implement to realize the function of each unit or module in the same or multiple software and or hardware when the present invention.

It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more, The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces The form of product.

The present invention be referring to the method for the embodiment of the present invention, equipment (system) and computer program product flow chart and/ Or block diagram describes.It should be understood that each process that can be realized by computer program instructions in flowchart and/or the block diagram and/ Or the combination of the process and/or box in box and flowchart and/or the block diagram.It can provide these computer program instructions To general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices processor to generate one A machine so that by the instruction that the processor of computer or other programmable data processing devices executes generate for realizing The device for the function of being specified in one or more flows of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates, Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or The function of being specified in multiple boxes.

These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one The step of function of being specified in a box or multiple boxes.

In a typical configuration, calculating equipment includes one or more processors (CPU), input/output interface, net Network interface and memory.

Memory may include the non-volatile memory in computer-readable medium, random access memory (RAM) and/or The forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flash RAM).Memory is computer-readable medium Example.

Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method Or technology come realize information store.Information can be computer readable instructions, data structure, the module of program or other data. The example of the storage medium of computer includes, but are not limited to phase change memory (PRAM), static random access memory (SRAM), moves State random access memory (DRAM), other kinds of random access memory (RAM), read-only memory (ROM), electric erasable Programmable read only memory (EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-ROM), Digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or other magnetic storage devices Or any other non-transmission medium, can be used for storage can be accessed by a computing device information.As defined in this article, it calculates Machine readable medium does not include temporary computer readable media (transitory media), such as the data-signal and carrier wave of modulation.

It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described want There is also other identical elements in the process, method of element, commodity or equipment.

It will be understood by those skilled in the art that the embodiment of the present invention can provide as method, system or computer program product. Therefore, complete hardware embodiment, complete software embodiment or embodiment combining software and hardware aspects can be used in the present invention Form.It is deposited moreover, the present invention can be used to can be used in the computer that one or more wherein includes computer usable program code The shape for the computer program product implemented on storage media (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) Formula.

The present invention can describe in the general context of computer-executable instructions executed by a computer, such as program Module.Generally, program module includes routines performing specific tasks or implementing specific abstract data types, programs, objects, group Part, data structure etc..The present invention can also be practiced in a distributed computing environment, in these distributed computing environments, by Task is executed by the connected remote processing devices of communication network.In a distributed computing environment, program module can be with In the local and remote computer storage media including storage equipment.

Various embodiments are described in a progressive manner in the present invention, same and similar part between each embodiment It may refer to each other, each embodiment focuses on the differences from other embodiments.Implement especially for system For example, since it is substantially similar to the method embodiment, so being described relatively simple, related place is referring to embodiment of the method Part illustrates.

The above description is only an embodiment of the present invention, is not intended to restrict the invention.For those skilled in the art For, the invention may be variously modified and varied.All any modifications made within the spirit and principles of the present invention are equal Replacement, improvement etc., should be included within scope of the presently claimed invention.

Claims

1. the method that pupil image is positioned in a kind of pair of eye image characterized by comprising

S1: several rays are determined in the eye image to be processed according to the origin, calculates and constitutes described in each The pixel gradient amplitude of each pixel of ray, and each pixel gradient amplitude institute that will be greater than preset threshold is right respectively The pixel answered is determined as reference image vegetarian refreshments；

S2: according to each reference image vegetarian refreshments spacing distance between the origin respectively, from each reference pixel Several suspicious marginal points are determined in point, and determine geometric center pixel corresponding to each suspicious marginal point；

S3: detect this determination the geometric center pixel and previous determination history geometric center pixel between away from Whether it is less than setting numerical value from value, if so, executing S5；Otherwise, S4 is executed；

S4: being determined as history geometric center pixel for the geometric center pixel, and the geometric center pixel is determined For new origin, and execute S1；

S5: according to each suspicious marginal point fit object model of ellipse, the people is marked by the Target ellipse model The position of pupil image in eye image.

2. the method according to claim 1, wherein according to each suspicious marginal point fit object elliptical modes Type marks the position of pupil image in the eye image by the Target ellipse model, comprising:

A1: 1 sampled edge points are randomly choosed from each suspicious marginal point, determine at least three samplings Marginal point respectively corresponds the sampling tangent line in the initial model of ellipse, and according to each sampled edge point and each item institute It states sampling tangent line and determines sampling pupil center；

A2: for each the non-sampled marginal point for being not selected for sampled edge point in each suspicious marginal point, according to The non-sampled marginal point, the non-sampled marginal point pair should be in the tangent lines in the initial model of ellipse, each sampling Sampling tangent line described in marginal point and each item determines calibration pupil center；

A3: being directed to each described calibration pupil center, when the interval of the calibration pupil center and the sampling pupil center When distance is not more than set distance, non-sampled marginal point corresponding to the calibration pupil center is determined as credible marginal point；

A4: the ratio between the first total amount of each credible marginal point and the second total amount of each suspicious marginal point is determined Value, and detect whether the ratio is less than given threshold, if so, executing A1；Otherwise, A5 is executed；

A5: according to each credible marginal point fit object model of ellipse, the people is marked by the Target ellipse model The position of pupil image in eye image.

3. according to the method described in claim 2, it is characterized in that, being sampled according to each sampled edge point and each item Tangent line determines sampling pupil center, comprising:

It is directed to each described midpoint, determines corresponding institute of two that correspond to the midpoint sampled edge points institutes The intersection point of sampling tangent line is stated, and determines the straight line where the midpoint and the intersection point；

According to least square method calculate each item of distance in the eye image described in the nearest apsis of linear distance, and by institute Apsis is stated to be determined as sampling pupil center.

4. according to the method described in claim 2, it is characterized in that,

B0: credible set is formed using each credible marginal point；

B2: each the described credible marginal point for including in the calculating credible set is respectively between the transition model of ellipse Algebraic distance；

B3: it is calculated according to the credible total amount for the credible marginal point for including in each algebraic distance and the credible set flat Equal fitness bias, according to the average fit deviation determination deviation threshold value；

B4: being directed to each described suspicious marginal point, detects between the suspicious marginal point and the transition model of ellipse Algebraic distance, when the algebraic distance between the suspicious marginal point and the transition model of ellipse is greater than the deviation threshold, The suspicious marginal point is determined as noise spot, and using each marginal point for being not determined to noise spot formed it is new can Letter set；

B5: whether credible set and the credible set of previous formation for detecting this formation are identical, if so, executing B6； Otherwise, B1 is executed；

5. the method according to claim 1, wherein

It is described according to each reference image vegetarian refreshments spacing distance between the origin respectively, from each reference pixel Several suspicious marginal points are determined in point, comprising:

Calculate the expected value and standard deviation of each spacing distance；

6. according to claim 1 to any method in 5, which is characterized in that

Acquire video image；

Each frame original image of the video image is filtered respectively to obtain the institute of original image described in each frame Corresponding pretreatment image；

Extract the eye image to be processed that presently described pretreatment image carries respectively from each Zhang Suoshu pretreatment image.

7. according to the method described in claim 6, it is characterized in that,

A non-selected eye image to be processed is successively selected, and determines the eye image to be processed of selection in video image In corresponding frame number；

When the frame number is 1, the geometric center of the eye image to be processed of selection is determined as initial pixel point； Or, when the frame number is greater than 1, according in the pupil of pupil image entrained by the eye image to be processed of previous selection The heart determines initial pixel point in the eye image to be processed；

And/or

The S5 further comprises that the geometric center pixel is determined as to the pupil carried in the eye image to be processed The pupil center of image.

8. the device that pupil image is positioned in a kind of pair of eye image characterized by comprising

Initial point determining module is made for determining initial pixel point in eye image to be processed, and by the initial pixel point For origin；

Endpoint detections module is counted for determining several rays in the eye image to be processed according to the origin The pixel gradient amplitude for constituting each pixel of ray described in each is calculated, and will be greater than each pixel of preset threshold The corresponding pixel of gradient magnitude institute is determined as reference image vegetarian refreshments；

Noise spot filtering module, for according to each reference image vegetarian refreshments spacing distance between the origin respectively, from Several suspicious marginal points are determined in each reference image vegetarian refreshments, and are determined several corresponding to each suspicious marginal point What central pixel point；

Spot detection module, for detecting the geometric center pixel of this determination and the history geometric center of previous determination Whether the distance between pixel value is less than setting numerical value, if so, trigger model fitting module；Otherwise, flip-flop transition is handled Module；

The transition processing module will be described for the geometric center pixel to be determined as history geometric center pixel Geometric center pixel is determined as new origin, and triggers the endpoint detections module；

The models fitting module, for passing through the target according to each suspicious marginal point fit object model of ellipse Model of ellipse marks the position of pupil image in the eye image.

9. device according to claim 8, which is characterized in that

The models fitting module, comprising: pretreatment unit, sample processing unit, calibration process unit, trusted processes unit, Detection processing unit and label processing unit；Wherein,

The sample processing unit, for randomly choosing 1 sampled edge points from each suspicious marginal point, really The fixed 1 sampled edge points respectively correspond the sampling tangent line in the initial model of ellipse, and according to each described Sampling tangent line described in sampled edge point and each item determines sampling pupil center；

The calibration process unit, for for each for being not selected for sampled edge point in each suspicious marginal point Non-sampled marginal point, should be in the initial model of ellipse according to the non-sampled marginal point, the non-sampled marginal point pair Sampling tangent line described in tangent line, each sampled edge point and each item determines calibration pupil center；

The trusted processes unit, for be directed to each described calibration pupil center, when the calibration pupil center with it is described When sampling the spacing distance of pupil center no more than set distance, by non-sampled marginal point corresponding to the calibration pupil center It is determined as credible marginal point；

The detection processing unit, for determine each credible marginal point the first total amount and each suspicious marginal point The second total amount between ratio, and detect whether the ratio is less than given threshold, if so, triggering the sampling processing list Member；Otherwise, the meter processing unit is triggered；

The meter processing unit, for passing through the target according to each credible marginal point fit object model of ellipse Model of ellipse marks the position of pupil image in the eye image.

10. device according to claim 8 or claim 9, which is characterized in that

Described image acquisition module, for acquiring video image；

The filtering processing module is filtered respectively for each frame original image to the video image to obtain The corresponding pretreatment image of the institute of original image described in each frame；

Described image extraction module is taken for extracting presently described pretreatment image respectively from each Zhang Suoshu pretreatment image The eye image to be processed of band.