CN104156723B - A kind of extracting method with the most stable extremal region of scale invariability - Google Patents

Abstract

The invention discloses a kind of extracting method with the most stable extremal region of scale invariability, first use most stable extremal region algorithm in original image, detect initial most stable extremal region, the yardstick pyramid of initial most stable extremal region is set up again by M dyadic wavelet transform, and wherein according to the energy operator of M dyadic wavelet transform coefficient, determine the unique point with scale invariability, to each unique point, the extremal region corresponding with it is obtained from the pyramidal each tomographic image of yardstick of most stable extremal region, and pass through the stability indicator of extremal region in multiscale space, extract the most stable extremal region with scale invariability, finally the most stable extremal region with scale invariability is adjusted to elliptical shape, obtain the final most stable extremal region with unchangeability.Scale invariability combines with most stable extremal region by the present invention, is extracted most stable extremal region, has affine-invariant features completely.

Description

A kind of extracting method with the most stable extremal region of scale invariability

Technical field:

The present invention relates to the extraction in digital picture invariant features region, especially there is the extracting method of the most stable extremal region of scale invariability.

Background technology:

From image, detect characteristic area is one of primary study problem of computer vision field always.In recent years, the image affine invariants region based on regional area is subject to paying close attention to more and more widely, and is successfully applied to target identification, image retrieval, robot autonomous navigation and scene and understands and many fields such as subject image classification.

Typical local characteristic region detection algorithm comprises: scale invariant feature point SIFT (scaleinvariantfeaturetransform) that Lowe proposes, the affine invariants point based on multiple dimensioned Harris angle point that Mikolajczyk and Schmid proposes, the most stable extremal region MSER (maximumstableextremeregion) of people's propositions such as the affine constant marking area (AISR) based on information entropy that the people such as Kadir proposes and Matas.In the Local affine invariant feature Region detection algorithms proposed, MSER algorithm has superior performance in many aspects, all has good unchangeability to the rotation change of image, visual angle change, brightness change.MSER algorithm basic thought is, for any width gray scale picture, choose all possible threshold value from small to large, obtain the many connected regions this width image, finally obtain the region being less than certain value in these regions with its area change of the change of threshold value, namely most stable extremal region.Because MSER extracts from single scalogram picture, when larger change occurs graphical rule, the fuzzy meeting of image makes the border of most stable extremal region change, thus affects the stability of Affinely invariant region.

Summary of the invention

For MSER algorithm Problems existing, the present invention is by the stability indicator of extremal region in definition multiscale space, a kind of extracting method with the most stable extremal region of scale invariability is proposed, first key problem in affine invariants extraction can be solved: the location of characteristic area, namely where extract feature; Meanwhile, the basis that the Affinely invariant region that this extracting method detects can be laid for follow-up image procossing.

The technical scheme that the present invention takes is: a kind of extracting method with the most stable extremal region of scale invariability, and its concrete steps are:

1) first use most stable extremal region algorithm in original image, detect initial most stable extremal region;

2) set up the yardstick pyramid of initial most stable extremal region by M dyadic wavelet transform, according to the energy operator of M dyadic wavelet transform coefficient in the yardstick pyramid of initial most stable extremal region, determine the unique point with scale invariability:

Initial most stable extremal region is designated as f (x, y), increases progressively yardstick M ^jlower to x and y direction, do M dyadic wavelet transform, wherein M is greater than 1 real number being less than 2, and j is the positive integer increased progressively, and wavelet function ψ (x, y) is wherein defined as smooth function along the derivative in x and y direction, thus obtain by increasing progressively yardstick M ^junder level and smooth most stable extremal region and wavelet coefficient the MSER yardstick pyramid of composition; The wavelet transform result of different scale when being even number to j, in conjunction with the information on its adjacent position, adopts Teager energy operator to obtain the energy of M system wavelet conversion coefficient and the mould of correspondence carry out local non-maxima suppression according to this modulus value, namely compared with around 8 points, if the modulus value of current point is maximum, namely this point is confirmed as the unique point with scale invariability, is designated as X=(x, y, M ^j), the wherein position of (x, y) representation feature point, M ^jrepresentation feature point place yardstick;

3) to each unique point determined by above-mentioned steps, the extremal region corresponding with it is obtained from the pyramidal each tomographic image of yardstick of most stable extremal region, and by the stability indicator of extremal region in multiscale space, extract the most stable extremal region with scale invariability:

First to each unique point extracted in the yardstick pyramid from most stable extremal region, most stable extremal region level and smooth under finding out its place yardstick the connected region that the gray-scale value of middle pixel is all large or all less than the gray-scale value of its boundary pixel point, as the extremal region of its correspondence; Then judge spatial stability and the dimensional stability of this extremal region, if this extremal region is all stable at metric space and pixel space, so this extremal region is exactly the most stable extremal region with scale invariability;

In described step 3, stability indicator comprises the stability indicator ε (ER) of metric space and the stability indicator of pixel space

The stability indicator ε (ER) of metric space is defined as:

$\mathrm{\ϵ}\left(\mathrm{ER}\right)=\frac{{\mathrm{ER}}_{j-1}\∩{\mathrm{ER}}_j}{{\mathrm{ER}}_{j-1}\∪{\mathrm{ER}}_j}+\frac{{\mathrm{ER}}_{j+1}\∩{\mathrm{ER}}_j}{{\mathrm{ER}}_{j+1}\∪{\mathrm{ER}}_j}$

Wherein j represents the level of extremal region ER in yardstick pyramid;

The stability indicator of pixel space be defined as:

Wherein Δ is an input parameter of algorithm, || represent the gesture of set, ER ^{+ Δ}, ER ^-Δbelong to limit area, if the ε of extremal region ER (ER) exists local minimum, then claim extremal region ER to be stable at metric space, if extremal region ER be local minimum, so region ER is stable in pixel space, if extremal region ER is stable at metric space and pixel space, so extremal region ER is exactly the most stable extremal region with scale invariability.

4) finally the most stable extremal region with scale invariability is adjusted to elliptical shape, so that subsequent treatment:

First the region of repetition or poor stability is deleted according to size and area change rate size; Then according to covariance matrix, irregularly shaped region is adjusted to the elliptical region of rule; This elliptical region is amplified some ratios, obtains the final most stable extremal region with unchangeability.

According to the preferred version of embodiment, in the described M dyadic wavelet transform in step 2 metric space number, namely the maximal value J of j selects 21.

According to the preferred version of embodiment, in step 4, have according to the region that area change rate size carries out deleting: area change rate be less than 0.5 or be greater than 1 region.

According to the preferred version of embodiment, in step 4, according to the region that size carries out deleting be: the half of pixel threshold values as the number of pixels of selected test pattern of setting surface area, when the number of pixels of this surface area exceedes threshold values, be then large area region, need to delete.

According to the preferred version of embodiment, in step 4, elliptical region amplification factor is 2.5.

Design concept of the present invention is:

In existing Local affine invariant feature Region detection algorithms, the MSER algorithm that the people such as Matas proposes, all have good robustness, but it is inadequate to the vary stable of graphical rule to the rotation change of image, visual angle change, brightness change.Meanwhile, existing Multiscale Theory can be analyzed image under different resolution and multiple yardstick, with human visual system from coarse to fine to experience image information be consistent.The present invention is by the stability indicator of extremal region in definition multiscale space, and propose a kind of multiple dimensioned most stable extremal region detection method, it detects MSER first on the original image.Then set up the yardstick pyramid of the MSER of original image by M dyadic wavelet transform, in MSER yardstick pyramid, detect the local modulus maximum value of wavelet coefficient, determine the unique point with scale invariability.Then to each unique point determined, existing region growing algorithm is utilized to obtain the extremal region corresponding with it from the pyramidal each tomographic image of MSER yardstick, and by the stability indicator of extremal region in multiscale space, extract the most stable extremal region with scale invariability.Finally the most stable extremal region with scale invariability is adjusted to elliptical shape, so that subsequent treatment.

In sum, scale invariability combines with most stable extremal region by the present invention, have found the image-region that can extract invariant features, i.e. pixel coordinate figure.By the most stable extremal region that the present invention detects, all can maintain the invariance to the rotation change, dimensional variation, brightness change, visual angle change, noise etc. of image, namely there is affine-invariant features completely.

Accompanying drawing explanation

Fig. 1 is the FB(flow block) of method described in embodiment;

Fig. 2 is the adjustment schematic diagram of covariance elliptical region;

Embodiment:

Below in conjunction with drawings and Examples data, said method is described in detail:

As shown in Figure 1, a kind of extracting method with the most stable extremal region of scale invariability, comprises the steps:

The first step, according to the MSER implementation algorithm that the people such as Murphy propose, detects MSER in original image; (this step utilizes existing algorithm, and detailed process repeats no more, and is only defined as follows MSER :)

Gray level image I can be defined as a mapping from 2 dimension pixel coordinates to gray-scale value S, and the limit area ER in image I is a connected region, and meets the following conditions:

$\∀p\∈\mathrm{ER},\∀q\∈C\left(\mathrm{ER}\right)\→I\left(p\right)\mathrm{RI}\left(q\right)$

Wherein relation R is greater than relation or is less than relation, and C (ER) represents the border of region ER, q and p represents pixel.If R is greater than relation, then claims region ER to be maximum value limit area, be expressed as ER+; If R is less than, then claims region ER to be minimal value limit area, be expressed as ER-.According to above definition, limit area is the connected region that the gray-scale value of pixel in image is all large or all less than the gray-scale value of its boundary pixel point.In order to search out most stability limit region from limit area, need to ask for the stability index of limit area in pixel space

Wherein Δ is an input parameter of algorithm, || represent the gesture of set, ER ^{+ Δ}, ER ^-Δbelong to limit area.If be local minimum, so region ER is exactly initial most stable extremal region MSER.

Second step, sets up the yardstick pyramid of original image MSER by M dyadic wavelet transform, in MSER yardstick pyramid, determine the unique point with scale invariability:

A) from j=1, multiple dimensioned M system wavelet transformation is carried out to original image MSER (being designated as f (x, y)), obtain the M system wavelet coefficient under J different scale.In order to substantially be consistent with the number of plies of graphical rule considered in SIFT method, the present embodiment selects j=21.The wavelet function adopted during wavelet transformation then for second order splines, this is because second order spline wavelets has compact schemes, frequency domain rapid decay, and the feature such as analytical expression is relatively simple.

The wavelet transform result of different scale when being b) even number to j, in conjunction with the information on its adjacent position, adopts Teager energy operator, obtains the energy of M system wavelet conversion coefficient with and the mould of correspondence wherein represent that logarithm value m rounds.

C) basis value carry out local non-maxima suppression, namely compared with around 8 points, only have current point maximum, then this point is candidate feature point, otherwise is just rejected from feature point detection algorithm by this point.

M dyadic wavelet transform is wherein defined as:

Suppose that function # (x) meets and time, claim θ (x) to be smooth function.When θ (x, y) is for two-dimentional smooth function, the smooth function θ on image f (x, y) and different scale a _a(x, y) convolution, will make image f (x, y) smooth, wherein

Definition 2-d wavelet function is respectively:

${\mathrm{\ψ}}^{\left(1\right)}(x,y)=\frac{\mathrm{d\θ}(x,y)}{\mathrm{dx}}$

${\mathrm{\ψ}}^{\left(2\right)}(x,y)=\frac{\mathrm{d\θ}(x,y)}{\mathrm{dy}}$

Work as ψ ⁽¹⁾(x, y) and ψ ⁽²⁾(x, y) meets completeness and the stability condition of 2-d wavelet, can as the wavelet mother function of two-dimensional wavelet transformation.Note:

${\mathrm{\ψ}}_a^{\left(1\right)}(x,y)=\frac{1}{a^2}{\mathrm{\ψ}}^{\left(1\right)}(\frac{x}{a},\frac{y}{a})$

${\mathrm{\ψ}}_a^{\left(2\right)}(x,y)=\frac{1}{a^2}{\mathrm{\ψ}}^{\left(2\right)}(\frac{x}{a},\frac{y}{a})$

Then the wavelet transformation of function f (x, y) is:

$W_a^{\left(1\right)}f(x,y)=f(x,y)*{\mathrm{\ψ}}_a^{\left(1\right)}(x,y)$

$W_a^{\left(2\right)}f(x,y)=f(x,y)*{\mathrm{\ψ}}_a^{\left(2\right)}(x,y)$

Realize for the ease of computing machine, the method for general discretize yardstick a progressively strengthens by the form of power series a, even a=M ^j, M>0, j ∈ Z ⁺, j<J.If get M=2, then the wavelet transformation deserving to be called formula representative is Dyadic Wavelet Transform.And in order to do finer division to yardstick a, also a can be taken as mark or irrational number, the wavelet transformation of now above formula representative is referred to as the wavelet transformation of M system, and its calculating often realizes by the method for frequency domain.

Can prove:

$\left[\begin{array}{c}{W}_{M^j}^{\left(1\right)}f(x,y)\\ W_{M^j}^{\left(2\right)}f(x,y)\end{array}\right]=M^j\left[\begin{array}{c}\frac{d}{\mathrm{dx}}(f(x,y)*{\mathrm{\&theta;}}_{M^j}(x,y))\\ \frac{d}{\mathrm{dy}}(f(x,y)*{\mathrm{\&theta;}}_{M^j}(x,y))\end{array}\right]=M^j\&dtri;(f(x,y)*{\mathrm{\&theta;}}_{M^j}(x,y))$

Can be found out by above formula, M system wavelet coefficient be proportional to yardstick M respectively ^jlower quilt institute smoothed image f (x, y) in the horizontal direction with the partial derivative of vertical direction.

3rd step, to each unique point determined, obtains the extremal region corresponding with it from the pyramidal each tomographic image of MSER yardstick, and by the stability indicator of extremal region in multiscale space, extracts the most stable extremal region with scale invariability;

The most stable extremal region feature of the Scale invariant that the present invention detects is the most stable extremal region extracted from yardstick pyramid, not by means of only the stability of extremal region in pixel space is investigated, and the stability of metric space is investigated.Extremal region at the definition of stability of metric space is:

$\mathrm{\&epsiv;}\left(\mathrm{ER}\right)=\frac{{\mathrm{ER}}_{j-1}\&cap;{\mathrm{ER}}_j}{{\mathrm{ER}}_{j-1}\&cup;{\mathrm{ER}}_j}+\frac{{\mathrm{ER}}_{j+1}\&cap;{\mathrm{ER}}_j}{{\mathrm{ER}}_{j+1}\&cup;{\mathrm{ER}}_j}$

Wherein j represents the level of extremal region in yardstick pyramid.If the ε of extremal region ER (ER) exists local minimum, then extremal region ER is claimed to be stable at metric space.If be local minimum, so region ER is stable in pixel space.If extremal region ER is stable at metric space and pixel space, so extremal region ER is exactly the most stable extremal region with scale invariability.

4th step, is finally adjusted to elliptical shape, so that the subsequent treatment such as feature interpretation, characteristic matching by the most stable extremal region with scale invariability.

Before carrying out elliptical region adjustment, delete some regions according to size and area change rate size.Because the region that area is larger or less has poor uniqueness, directly cause lower can matching.The Domain Stability that area change rate is larger is poor, and the less region of area change rate mostly is repeat region.The existence in these regions can increase the time complexity of algorithm.For the region that those area change rates are excessive and too small, delete according to area change rate size, namely remove area change rate be less than 0.5 or be greater than 1 region, the only region of Retention area rate of change between 0.5 to l:

For the region that area is larger, according to the number of pixels of selected test pattern, if the pixel threshold of large area region is the half of full figure number of pixels, if area pixel number is greater than this threshold value namely delete this region.And for the less region of area, because mainly caused by little vibration or noise, be called noise extremal region, what be specifically set smaller than structural element is area comparatively zonule, the closure operation fb=of existing gray scale morphology (f ⊕ b) Θ b is used to remove these noise extremal regions, wherein f is most stable extremal region image, ⊕ and Θ is respectively dilation and erosion computing, b is structural element, the size of this element can be determined according to the mean size of noise extremal region, generally should be greater than the mean size of noise extremal region, described in the present embodiment, structural element b selects the structural element of 5*5 pixel.

The characteristic area now obtained is irregularly shaped region, and irregularly shaped region is difficult to carry out feature interpretation.In order to the convenience of subsequent treatment, adopt and with the following method irregularly shaped region is adjusted to ellipse.

First the Pixel Information of utilization composition irregular area constructs the second-order central distance based on vector representation.

$\mathrm{\&Sigma;}\left(R\right)=\frac{1}{\left|R\right|}\underset{X\&Element;R}{\mathrm{\&Sigma;}}{(X-E\left(X\right))(X-E\left(X\right))}^T$

Wherein: X=[x, y] is the coordinate of pixel, and R represents irregular area, | R| is the gesture of the set be made up of pixels all in irregular area, i.e. the number of pixel in this region, and E represents expectation value, and T represents transposition.

Above formula is write as the covariance matrix form of dimensionality reduction, had

$D=\left[\begin{array}{cc}D\left(x\right)& \mathrm{COV}(x,y)\\ \mathrm{COV}(x,y)& D\left(y\right)\end{array}\right]$

Wherein D (x), D (y), COV (x, y) to be respectively in this region the covariance of horizontal ordinate variance, ordinate variance and transverse and longitudinal coordinate a little:

$\left\{\begin{array}{c}D\left(x\right)=E\left(x^2\right)-{\left[E\left(x\right)\right]}^2\\ D\left(y\right)=E\left(y^2\right)-{\left[E\left(y\right)\right]}^2\\ \mathrm{COV}(x,y)=E\left(\mathrm{xy}\right)-E\left(x\right)E\left(y\right))\end{array}\right.$

And:

$\begin{array}{cc}E\left(x\right)=\frac{1}{\left|R\right|}\underset{x\&Element;R}{\mathrm{\&Sigma;}}x& E\left(y\right)=\frac{1}{\left|R\right|}\underset{y\&Element;R}{\mathrm{\&Sigma;}}y\end{array}$

$\begin{array}{ccc}E\left(x^2\right)=\frac{1}{\left|R\right|}\underset{x\&Element;R}{\mathrm{\&Sigma;}}{x}^2& E\left(y^2\right)=\frac{1}{\left|R\right|}\underset{y\&Element;R}{\mathrm{\&Sigma;}}{y}^2& E\left(\mathrm{xy}\right)=\frac{1}{\left|R\right|}\underset{y\&Element;R}{\mathrm{\&Sigma;}}\mathrm{xy}\end{array}$

Wherein, | R| is the gesture of the set be made up of pixels all in irregular area, i.e. pixel number in this region.

Covariance matrix U is symmetrical matrix, according to its proper vector and characteristic root, can be adjusted major and minor axis direction and the amplitude of rear ellipse, as Fig. 2 shows, and α in figure ₁and α ₂be major and minor axis amplitude, θ represents the direction of major axis, and E (x) and E (y) is the center point coordinate of elliptical region after adjustment.

Each most stable extremal region has a gauge region corresponding with it, i.e. covariance elliptical region, and the area of this elliptical region is corresponding with the size of most stable extremal region.If this oval area is less, so its conspicuousness is just poor.In order to improve the conspicuousness in region, covariance ellipse is amplified, some background content of non-targeted in image is likely included by even now, but according to pertinent literature, suitably increase the area of gauge region, greatly can improve the conspicuousness of gauge region, and Comparatively speaking the impact of background is acceptable.Given this, the present invention amplifies covariance ellipse, and amplification factor is decided to be 2.5 by reference experiment result.Because the parameter of ellipse is only determined by the space distribution of corresponding region pixel, therefore good robustness can be kept by the method to affine change.

The present embodiment is at 8 groups of cycle testss such as the Graffiti keeping the consistent basis of experiment condition uses robot research group of Cambridge University to provide as far as possible, according to method of evaluating performance, local characteristic region repetition rate is adopted to investigate the performance of the inventive method and MSER algorithm.Definition local characteristic region repetition rate is as follows:

$\mathrm{repeatability}=\frac{\mathrm{correctMathes}}{\min \mathrm{Number}}$

Wherein correctMathes to represent between two width images the most stable extremal region number of correct coupling, and minNumber represents the smaller value of the most stable extremal region number of two width images.

No matter there occurs what kind of affine change between image, as long as local characteristic region corresponding in two width images also exists, they just should detect in the position of correspondence by desirable Local affine invariant feature Region detection algorithms.Therefore under identical experiment condition, the performance that the method that repetition rate is larger has had.

Respectively with the 1st width image of 8 groups of cycle testss for reference picture, obtain image to be matched with reference to image scaling some scale.The zoom factor of the 1st image relative reference image to be matched is the zoom factor of the 1.5,2nd image relative reference image to be matched is 2.0, the like, the zoom factor of last image relative reference image to be matched is 5.5.These 9 images to be matched are respectively Z1, Z2, Z3 ... and Z9.Adopt two kinds of methods to detect the local characteristic region (namely most stable extremal region) of 9 width images, and calculate the characteristic area repetition rate of 9 width image relative reference images respectively.For 8 groups of cycle testss, the characteristic area repetition rate that two kinds of methods obtain is as shown in the table

	Z1	Z2	Z3	Z4	Z5	Z6	Z7	Z8	Z9
										The present invention	0.86	0.81	0.77	0.70	0.65	0.63	0.51	0.42	0.35
MSER	0.85	0.78	0.71	0.61	0.44	0.21	0.17	0.09	0.03

As can be seen from the above table, when image generation dimensional variation, the characteristic area repetition rate that the present invention obtains all is greater than MSER method in any case.The dimensional variation of image is larger, and the advantage of the method for the invention is more obvious.

Claims (5)

1. there is an extracting method for the most stable extremal region of scale invariability, it is characterized in that, comprise the steps:

1) first use most stable extremal region algorithm in original image, detect initial most stable extremal region;

2) set up the yardstick pyramid of initial most stable extremal region by M dyadic wavelet transform, according to the energy operator of M dyadic wavelet transform coefficient in the yardstick pyramid of initial most stable extremal region, determine the unique point with scale invariability:

Initial most stable extremal region is designated as f (x, y), increases progressively yardstick M ^jlower to x and y direction, do M dyadic wavelet transform, wherein M is greater than 1 real number being less than 2, and j is the positive integer increased progressively, and wavelet function ψ (x, y) is wherein defined as smooth function along the derivative in x and y direction, thus obtain by increasing progressively yardstick M ^junder level and smooth most stable extremal region and wavelet coefficient the MSER yardstick pyramid of composition; The wavelet transform result of different scale when being even number to j, in conjunction with the information on its adjacent position, adopts Teager energy operator to obtain the energy of M system wavelet conversion coefficient and the mould of correspondence carry out local non-maxima suppression according to this modulus value, namely compared with around 8 points, if the modulus value of current point is maximum, namely this point is confirmed as the unique point with scale invariability, is designated as X=(x, y, M ^j), the wherein position of (x, y) representation feature point, M ^jrepresentation feature point place yardstick;

3) to each unique point determined by above-mentioned steps, the extremal region corresponding with it is obtained from the pyramidal each tomographic image of yardstick of most stable extremal region, and by the stability indicator of extremal region in multiscale space, extract the most stable extremal region with scale invariability:

First to each unique point extracted in the yardstick pyramid from most stable extremal region, most stable extremal region level and smooth under finding out its place yardstick the connected region that the gray-scale value of middle pixel is all large or all less than the gray-scale value of its boundary pixel point, as the extremal region of its correspondence; Then judge spatial stability and the dimensional stability of this extremal region, if this extremal region is all stable at metric space and pixel space, so this extremal region is exactly the most stable extremal region with scale invariability;

Aforementioned stable index comprises the stability indicator ε (ER) of metric space and the stability indicator of pixel space

The stability indicator ε (ER) of metric space is defined as:

Wherein j represents the level of extremal region ER in yardstick pyramid;

The stability indicator of pixel space be defined as:

Wherein Δ is an input parameter of algorithm, || represent the gesture of set, ER ^{+ Δ}, ER ^-Δbelong to limit area, if the ε of extremal region ER (ER) exists local minimum, then claim extremal region ER to be stable at metric space, if extremal region ER be local minimum, so region ER is stable in pixel space, if extremal region ER is stable at metric space and pixel space, so extremal region ER is exactly the most stable extremal region with scale invariability.

4) finally the most stable extremal region with scale invariability is adjusted to elliptical shape:

First the region of repetition or poor stability is deleted according to size and area change rate size; Then according to covariance matrix, irregularly shaped region is adjusted to the elliptical region of rule; This elliptical region is amplified some ratios, obtains the final most stable extremal region with unchangeability.

2. there is the extracting method of the most stable extremal region of scale invariability according to claim 1, it is characterized in that, in described step 2 in M dyadic wavelet transform

3. have the extracting method of the most stable extremal region of scale invariability according to claim 1, it is characterized in that, the region of carrying out deleting according to area change rate size in described step 4 has: area change rate be less than 0.5 or be greater than 1 region.

4. there is the extracting method of the most stable extremal region of scale invariability according to claim 1, it is characterized in that, in described step 4, according to the region that size carries out deleting be: the half of pixel threshold values as the number of pixels of selected test pattern of setting large area region, when the number of pixels of this surface area exceedes threshold values, be then large area region, need to delete.

5. according to one of claim 1-4, have the extracting method of the most stable extremal region of scale invariability, it is characterized in that, in described step 4, elliptical region amplification factor is 2.5.