CN103268595A - Infrared image enhancement processing method based on lateral inhibition coefficient - Google Patents

Abstract

The invention relates to a lateral inhabitation coefficient determining method based on unilateral attenuation exponential and belongs to the field of image enhancement processing. A lateral inhibition coefficient distribution function is based on the unilateral attenuation exponential, the size of the radius N of a lateral inhibition field range is used for determining a coefficient sigma, the lateral inhibition coefficient is distributed and concentrated in a small lateral inhibition field range through controlling the lateral inhibition coefficient attenuation speed rate to reduce calculation amount of a lateral inhibition network, and a coefficient A is determined based on the optimal lateral inhibition coefficient and the sigma to effectively enhance image visual effects. The method is simple in implementation process, small in calculation amount and obvious in image enhancement effect, and has application and popularization value.

Description

Infrared image based on the lateral inhibition coefficient strengthens disposal route

Technical field

The present invention relates to the image enhancement technique field, specifically belong to the infrared image processing based on the lateral inhibition coefficient in the lateral inhibition network model of vision bionics.

Background technology

The lateral inhibition effect is ubiquitous phenomenon in the animal optic nerve system, and it originates from the careful research in 40 years of the soft-shelled turtle compound eye of Hartline.The years of researches result shows from original arthropod and from the peripheral neverous system to the central nervous system, to have the lateral inhibition effect to the people from the sense of touch to the vision.

Lateral inhibition plays a part to strengthen contrast, outstanding contour of object, so lateral inhibition can specifically comprise: detect the edge of infrared image, strengthen the contrast of infrared image target and background for the infrared image enhancing processing of low signal-to-noise ratio; The low-frequency component that suppresses infrared image, the radio-frequency component of reservation infrared image; The caused image blurring information of the defective of dioptric system is recovered.

The lateral inhibition ability of lateral inhibition under the fixed network model depends on distribution and the lateral inhibition open country of lateral inhibition coefficient.The distribution of lateral inhibition coefficient is determining the filtering characteristic of lateral inhibition network; Suppressing the open country refers to and can produce inhibiting receptor unit scope to receptor unit, center.For the wild scope of identical lateral inhibition, the lateral inhibition coefficient of optimization distributes and can improve the visual effect of infrared image better.But the research report of offside rejection coefficient distribution both at home and abroad at present is less, and the report of offside rejection coefficient optimization research is few especially.

People such as Hartline find in the electric Physiological Experiment that the soft-shelled turtle compound eye is being done: the effect of lateral inhibition increases along with the distance of receptor unit and reduces, therefore can think that the lateral inhibition coefficient is the function of distance between the receptor unit, in two-dimentional lateral inhibition model, two receptor unit N _{I, j}And N _{P, q}Between distance definition be Euclidean distance

Lateral inhibition coefficient distribution function mainly contains three kinds at present: (1) two-dimentional hyperbola distribution: distribution function is k _{Ij, pq}=a/d _{Ij, pq}If d _{Ij, pq}〉=0; Wherein a is hyp focus.This its distribution character of lateral inhibition coefficient based on hyperbola distribution is too simple, only depends on an adjustable parameter, is unfavorable for that the optimization of lateral inhibition coefficient is chosen.(2) uni-modal Gaussian: distribution function is

The present usefulness of this lateral inhibition system's distribution form at most.(3) bimodal Gauss: it is Chen Hui that this lateral inhibition coefficient distributes, Ou Yangkai proposes " being had the receptor of a segment distance also to be eager to excel than the most contiguous ommatidium effect from trying receptor " at " emulation that is used for the lateral inhibition network model of image enhancement is compared " article, its model is $k_{\mathrm{ij},\mathrm{pq}}=\frac{1}{\mathrm{\β}}*\{\frac{1}{{\mathrm{\β}}_1}*\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_1}\exp (-\frac{{(d_{\mathrm{ij},\mathrm{pq}}-{\mathrm{\μ}}_1)}^{\text{2}}}{{{2\mathrm{\σ}}_1}^2})-\frac{1}{{\mathrm{\β}}_2}*\frac{1}{\sqrt{2\mathrm{\π}}{\mathrm{\σ}}_2}\exp (-\frac{{(d_{\mathrm{ij},\mathrm{pq}}-{\mathrm{\μ}}_2)}^2}{{{2\mathrm{\σ}}_2}^2})\};$ People such as Du Hongwei compare bimodal Gauss and ordinary numbers formula lateral inhibition weights in " biological lateral inhibition mechanism and applied research ", though this distribution function is when the wild scope of lateral inhibition is big, figure image intensifying effect is better, but the lateral inhibition coefficient of bimodal Gaussian distribution effect when strengthening edge of image is relatively poor in the wild scope of lateral inhibition hour.

Above several method is used for image when handling, and the choosing of its lateral inhibition coefficient usually relies on experience and decide, and has randomness.Wherein hyperbola distribution parameter adjustability is poor, and the lateral inhibition ability that occurs different lateral inhibition open country under the same distribution occasion is easily disperseed; Though Gaussian distribution meets biological vision mechanism, wild hour in lateral inhibition, the enhancing that is not suitable for image detail information is handled.Based on above reason, we have proposed monolateral damped expoential lateral inhibition coefficient distribution function.

Summary of the invention

The present invention is directed to prior art when carrying out the image processing, the lateral inhibition ability of different lateral inhibition open country is disperseed, and it is wild hour in lateral inhibition, Gaussian distribution is not suitable for the enhancing of image detail information and handles, propose a kind of lateral inhibition coefficient distribution function based on monolateral damped expoential, this function utilizes two parameters to control its distribution character, makes to be distributed in the lateral inhibition coefficient set in the wild scope of less lateral inhibition, to reduce the calculated amount that image is handled, reach effective enhancing image effect.The method that the present invention proposes is simple, is easy to realize having application value in image enhancement processing.

The technical scheme that the present invention solves the problems of the technologies described above is: each pixel in the image is as receptor unit, center, be subjected in lateral inhibition matrix template that the receptor unit is to its inhibiting effect on every side, its lateral inhibition weight distributes by monolateral decaying exponential function.The lateral inhibition matrix that generates based on monolateral decaying exponential function makes image in gray scale sudden change place because having quick attenuation characteristic, and high-frequency signal is enhanced, low frequency signal is inhibited, thereby the gray scale contrast is increased, be equivalent to the Hi-pass filter of image, have preferably " crisperding " effect.The application of lateral inhibition network in infrared image strengthens that reaches based on monolateral damped expoential of determining of the foundation of following offside rejection coefficient model, lateral inhibition coefficient is described in detail:

Choose the wild radius N of lateral inhibition, in the lateral inhibition open country, obtain maximum attenuation, according to formula

Determine parameter σ, with the contrast after the figure image intensifying, Y-PSNR, after the relation curve of information entropy and lateral inhibition coefficient and ∑ carries out normalization according to the linear transformation function, get the corresponding lateral inhibition coefficient of flex point that information entropy, contrast sharply rise and Y-PSNR sharply descends and be optimum lateral inhibition summation ∑, set up the lateral inhibition coefficient distributed model k based on monolateral damped expoential _{Ij, pq}=A*exp (σ * d _{Ij, pq}), calculate corresponding coefficient A according to lateral inhibition coefficient and ∑; Wherein, lateral inhibition coefficient k _{Ij, pq}The expression position is that (p, periphery q) experience cell pairs heart receptor unit, and (A and σ are coefficient, d for i, lateral inhibition weight j) _{Ij, pq}The expression (p, q) unit to the unit (i, distance j).Generation lateral inhibition matrix of coefficients H (p, q); According to receptor unit, center (i, input e j) _{I, j}, peripheral receptor unit (p, input e q) _{P, q}, call formula: r _{I, j}=e _{I, j}-H (p, q) * e _{P, q}Determine that the center experiences unit (i, output r j) _{I, j}

Calculating the wild radius of lateral inhibition is lateral inhibition coefficient and the ∑ of N.Get ∑ from 0.2 to 1, stepping △=0.01.Just can try to achieve different coefficient A according to different lateral inhibition coefficients with ∑.

If matrix size is 3 * 3, then $H(p,q)=\left[\begin{array}{ccc}k(-1,-1)& k(-\mathrm{1,0})& k(-\mathrm{1,1})\\ k(0,-1)& k\left(\mathrm{0,0}\right)& k\left(\mathrm{0,1}\right)\\ k(1,-1)& k\left(\mathrm{1,0}\right)& k\left(\mathrm{1,1}\right)\end{array}\right];$

$e(p,q)=\left[\begin{array}{ccc}e(i-1,j-1)& e(i-1,j)& e(i-1,j+1)\\ e(i,j-1)& e(i,j)& e(i,j+1)\\ e(i+1,j-1)& e(i+1,j)& e(i+1,j+1)\end{array}\right].$

The present invention proposes a kind of lateral inhibition coefficient based on monolateral decaying exponential function and distributes, and provide definite method of optimal coefficient, namely utilize the size of lateral inhibition open country to adjust the rate of decay that the lateral inhibition coefficient distributes, the lateral inhibition coefficient is sharply descended in the lateral inhibition open country, in the wild scope of less lateral inhibition, just can reach best lateral inhibition effect.The inventive method realizes simple, and operand is little, has to use and promotional value.

Description of drawings

Fig. 1 determines method implementing procedure figure for lateral inhibition coefficient of the present invention;

Fig. 2 is lateral inhibition coefficient distribution plan of the present invention;

Fig. 3 determines the method synoptic diagram for the parameter σ that the lateral inhibition coefficient distributes;

Fig. 4 is lateral inhibition matrix template synoptic diagram;

Fig. 5 handles index and lateral inhibition coefficient and graph of relation for the lateral inhibition after the normalization;

Fig. 6 is that the enhancing effect that several lateral inhibition coefficients distribute compares synoptic diagram.

Embodiment

Below at accompanying drawing and example implementation process of the present invention is specifically described, Fig. 1 is the implementing procedure figure of the inventive method, specifically may further comprise the steps: the lateral inhibition coefficient distributed model of setting up monolateral damped expoential, generate the lateral inhibition matrix, utilize the lateral inhibition matrix to carry out infrared image and strengthen processing.Following mask body is set forth the implementation process of each step:

1 sets up the lateral inhibition coefficient distributed model of monolateral damped expoential

Monolateral decaying exponential function f (t)=e ^{σ t}, σ＜0; Have quick attenuation characteristic, to remaining exponential function form behind time variable differential and the integration, this makes lateral inhibition network suppress low frequency, unique advantage on the reserved high-frequency composition.Therefore we propose to set up the lateral inhibition coefficient distributed model k based on monolateral damped expoential _{Ij, pq}=A*exp (σ * d _{Ij, pq}), k wherein _{Ij, pq}The expression position is that (p, (A and σ are coefficient for i, lateral inhibition effect j), represent product calculation, d * number in the formula in the cell pairs heart receptor unit of experiencing q) _{Ij, pq}The expression (p, q) unit to the unit (i, distance j).Idiographic flow is:

1. the horizontal ordinate with monolateral decaying exponential function changes receptor unit N into by time t _{I, j}And N _{P, q}Between apart from d _{Ij, pq}

2. the ordinate of the monolateral decaying exponential function function f (t) by time t is changed into about receptor cell distance d _{Ij, pq}The lateral inhibition coefficient k _{Ij, pq}

Rejection coefficient distributes as shown in Figure 2, wherein N _{I, j}The expression position is (i, receptor unit, center j), N _{P, q}The expression position is (p, receptor unit q), distance

Be receptor unit N _{P, q}To N _{I, j}Distance.

2 generate the lateral inhibition matrix

Choose the wild radius N of lateral inhibition, determine σ and A in the coefficient distribution, generate lateral inhibition matrix of coefficients H.

Determine factor sigma: σ determines the rate of decay of monolateral decaying exponential function.If the wild scope of lateral inhibition is more little, the lateral inhibition ability is more strong, operand is more little, then the lateral inhibition network performance is just more good.Choose the wild radius N of lateral inhibition, obtain maximum attenuation in the lateral inhibition open country, namely distance is

The time, then the decay of lateral inhibition coefficient is approximately zero.By formula But just calculating parameter σ.

Determine that coefficient A:A determines the amplitude of monolateral decaying exponential function.At first with the above-mentioned σ substitution lateral inhibition coefficient distribution formula k that calculates _{Ij, pq}=A*exp (σ * d _{Ij, pq}), namely the available expression formula that contains coefficient A is represented the lateral inhibition coefficient k _{Ij, pq}Calculating the wild radius of lateral inhibition is lateral inhibition coefficient and the ∑ of N.Get ∑ from 0.2 to 1, stepping △=0.01.Just can try to achieve different coefficient A according to different lateral inhibition coefficients with ∑.

Also very sensitive to noise when lateral inhibition network strengthens the edge, it is particularly important to take into account figure image intensifying effect when reducing image processing operand.Select the lateral inhibition coefficient of figure image intensifying optimum for use and calculate A based on this reason.Optimum lateral inhibition coefficient and ∑ determine that method is: with the contrast after the figure image intensifying, Y-PSNR, after the relation curve of information entropy and lateral inhibition coefficient and ∑ carries out normalization according to the linear transformation function, get that information entropy, contrast sharply rise and the corresponding lateral inhibition coefficient of flex point that Y-PSNR sharply descends and, be optimum lateral inhibition summation ∑.

According to the wild radius of the lateral inhibition of having chosen and fixed σ and A, again in conjunction with lateral inhibition coefficient distribution k _{Ij, pq}=A*exp (σ * d _{Ij, pq}) just can generate lateral inhibition matrix of coefficients H.On engineering,

The time, k _{Ij, pq}=0.0138A just can think to decay to zero approx.

1. determine factor sigma: choose the wild radius N of lateral inhibition, determine parameter σ in the lateral inhibition coefficient distribution function according to formula (1):

$N=-\frac{4}{\mathrm{\σ}}---\left(1\right)$

N=1 in the formula, 2,3,4,5,6.Corresponding lateral inhibition matrix is 3 * 3,5 * 5,7 * 7,9 * 9,11 * 11, and the common maximum of 13 * 13, N gets 6, determines that method is as shown in Figure 3.

2. determine coefficient A: at first with the above-mentioned σ substitution formula (2) that calculates, represent the lateral inhibition coefficient k with the expression formula that contains A _{Ij, pq}:

k _ij,pq=A*exp(σ*d _ij,pq) (2)

Calculate the distance of each unit and receptor unit, center according to formula (3):

$d_{\mathrm{ij},\mathrm{pq}}=\sqrt{{(i-p)}^2+{(j-q)}^2}---\left(3\right)$

D in the formula _{Ij, pq}For the receptor unit (i, j) and (p, distance q), the line-spacing deviation of (i-p) two receptor unit for example, it is poor (j-q) to represent the column distance of two receptor unit.

If the wild radius N of lateral inhibition coefficient is 1, then the size of lateral inhibition matrix masterplate is 3 * 3, and formula (4) provides matrix form:

$H(p,q)=\left[\begin{array}{ccc}k(-1,-1)& k(-\mathrm{1,0})& k(-\mathrm{1,1})\\ k(0,-1)& k\left(\mathrm{0,0}\right)& k\left(\mathrm{0,1}\right)\\ k(1,-1)& k\left(\mathrm{1,0}\right)& k\left(\mathrm{1,1}\right)\end{array}\right]---\left(4\right)$

(p q) is the inboard rejection coefficient k of masterplate to H in the formula _{Ij, pq}Matrix, numerical value is that (i is j) with (for example k (1 ,-1) can be regarded as receptor (p is q) to receptor unit, center (i, lateral inhibition coefficient j), the p=i-1 of this moment, q=j-1 for p, the range difference of row, column q) in the receptor unit in the bracket.

The wild radius N=3 of the lateral inhibition that this paper experiment is chosen, then can obtain the lateral inhibition matrix of coefficients is 7 * 7, concrete matrix as shown in Figure 4, the lateral inhibition matrix of coefficients is according to the difference of the wild radius of lateral inhibition and difference.

Just can try to achieve lateral inhibition matrix of coefficients H (p, q) (being H) of only containing parameter A by formula (1) (2) (3) (4) associating.

Shown in formula (5), ask for lateral inhibition coefficient and ∑:

$\mathrm{\Σ}=\underset{p=-N}{\overset{N}{\mathrm{\Σ}}}\underset{q=-N}{\overset{N}{\mathrm{\Σ}}}H---\left(5\right)$

In the formula, H is the matrix of lateral inhibition coefficient, and N is the radius of lateral inhibition matrix of coefficients, i.e. the wild radius of lateral inhibition.Get ∑ and progressively be increased to 1, step-length △=0.01 from 0.2.Just can try to achieve corresponding parameters A according to different lateral inhibition coefficients and ∑.

To bring the lateral inhibition coefficient k into by the parameter A that lateral inhibition coefficient and ∑ obtain _{Ij, pq}=A*exp (σ * d _{Ij, pq}) in, image is suppressed the enhancing of network and handle the image after being enhanced.

Get ∑ and progressively be increased to 1 from 0.2, step-length △=0.01, depict lateral inhibition coefficient and Σ and strengthen the relationship of contrast curve of back image through lateral inhibition network according to formula (6):

$C=\frac{g_{\max }-g_{\min }}{g_{\max }+g_{\min }}---\left(6\right)$

G wherein _Max, g _MinMaximal value and minimum value for gray-scale value in the image.

Depict the relation curve of the Y-PSNR of image after lateral inhibition coefficient and ∑ and the enhancing according to formula (6):

$\mathrm{PSNR}=10\log \left[\frac{{255}^2}{\frac{1}{M\×N}\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}{(x_{\mathrm{mn}}-x_{\mathrm{mn}}^{^})}^2}\right]---\left(7\right)$

X wherein _MnThe gray scale of back image is handled in expression,

Represent the gray scale of original image, the number of M * N presentation video pixel.

Depict the relation curve of the information entropy of image after lateral inhibition coefficient and ∑ and the enhancing according to formula (8) (9):

$H=-\underset{i=1}{\overset{M}{\mathrm{\Σ}}}p\left(k\right){\log }_{2}p\left(k\right)---\left(8\right)$

$p_k=\frac{A_k}{N\×N}(k=\mathrm{0,1},...N)---\left(9\right)$

The maximum gray scale number of M presentation video wherein, A _kThe pixel number of gray scale k in the expression entire image, N * N represents the pixel number of entire image, p _kProbability for the appearance of k gray level.

Carry out normalization according to formula (10), its linear function conversion formula is:

$Y=\frac{N-N_{\min }}{N_{\max }-N_{\min }}---\left(10\right)$

N wherein, Y is the value before and after the conversion, N _Max, N _MinMaximal value and minimum value for sample.

The contrast of back image will be strengthened, Y-PSNR, after the relation curve normalization of information entropy and lateral inhibition coefficient and ∑, get that information entropy, contrast sharply rise and the corresponding lateral inhibition coefficient of turning point that Y-PSNR sharply descends and, be optimum lateral inhibition summation ∑.Just can generate the lateral inhibition matrix H according to the wild radius N of the lateral inhibition of having chosen and formula (2).

Fig. 5 handles index and lateral inhibition coefficient and graph of relation for the lateral inhibition after the normalization.For the wild scope of different lateral inhibition, its lateral inhibition coefficient and matrix thereof just can be determined accordingly according to above method.

(3) utilize the lateral inhibition matrix of monolateral damped expoential to carry out the enhancing processing of infrared image

Each pixel in the image is as receptor unit, center, is subjected in lateral inhibition matrix template that the receptor unit is to its inhibiting effect on every side, and its lateral inhibition weight distributes by monolateral decaying exponential function.The lateral inhibition matrix that generates based on monolateral decaying exponential function makes image in gray scale sudden change place because having quick attenuation characteristic, and high-frequency signal is enhanced, low frequency signal is inhibited, thereby the gray scale contrast is increased, be equivalent to the Hi-pass filter of image, have preferably " crisperding " effect.

The lateral inhibition matrix of coefficients of above-mentioned generation is applied to the acyclic lateral inhibition network model of subtraction to carry out infrared image and strengthens and handle.

According to receptor unit, center (i, input e j) _{I, j}, (p is q) to receptor unit, center (i, the lateral inhibition coefficient k of lateral inhibition weight j) in expression peripheral receptor unit _{Ij, pq}, call formula: Computing center receptor unit (i, output r j) _{I, j},

That is: r _{I, j}=e _{I, j}-H (p, q) * e _{P, q}(11)

Suppose that lateral inhibition matrix of coefficients size is 3 * 3, then $H(p,q)=\left[\begin{array}{ccc}k(-1,-1)& k(-\mathrm{1,0})& k(-\mathrm{1,1})\\ k(0,-1)& k\left(\mathrm{0,0}\right)& k\left(\mathrm{0,1}\right)\\ k(1,-1)& k\left(\mathrm{1,0}\right)& k\left(\mathrm{1,1}\right)\end{array}\right];$ $e(p,q)=\left[\begin{array}{ccc}e(i-1,j-1)& e(i-1,j)& e(i-1,j+1)\\ e(i,j-1)& e(i,j)& e(i,j+1)\\ e(i+1,j-1)& e(i+1,j)& e(i+1,j+1)\end{array}\right].$ (p q) is the matrix of lateral inhibition coefficient to H in the formula, and it is described as defined above, and (p is the matrix form of adjacent receptor unit q) to e, and (p q) represents receptor unit e to e in the matrix _{P, q}, symbol * represents two matrix correspondence position elements and multiplies each other.

Table 1 is lateral inhibition coefficient and the comparison sheets of several lateral inhibition coefficients under different distributions; As seen from the table, the lateral inhibition ability of monolateral decaying exponential function concentrates in the little template, and its lateral inhibition ability of 3 * 3 obviously lateral inhibition ability than other distribution is strong, and its lateral inhibition summation of 5 * 5 is approximately equal to bimodal Gauss's 7 * 7 lateral inhibition summation.

Table 1: lateral inhibition coefficient and comparison sheet under the different distributions

Fig. 6 is the infrared image enhancing design sketch that two kinds of lateral inhibition coefficients distribute corresponding; Corresponding monolateral exponential distribution 3 * 3 templates (upper left) in the left side wherein, the infrared image 5 * 5 templates (lower-left) under strengthens design sketch, corresponding bimodal Gaussian distribution 3 * 3 templates (upper right) in the right, the infrared image enhancing design sketch under 5 * 5 templates (bottom right).

The present invention proposes a kind of lateral inhibition coefficient based on monolateral decaying exponential function and distributes, utilize the wild radius of lateral inhibition and optimum lateral inhibition coefficient and its coefficient distribution character is controlled, its lateral inhibition ability is effectively sharply being descended in the wild scope of lateral inhibition, to reduce operand, strengthen the visual effect of image.

Claims (5)

1. the infrared image based on the lateral inhibition coefficient strengthens disposal route, it is characterized in that, each pixel in the image is chosen the wild radius N of lateral inhibition as the receptor unit, obtains maximum attenuation in the lateral inhibition open country, according to formula

Determining parameter σ, determine optimum lateral inhibition coefficient summation ∑, is lateral inhibition coefficient distributed model with monolateral damped expoential model conversion, sets up the lateral inhibition coefficient distributed model k based on monolateral damped expoential _{Ij, pq}=A*exp (σ * d _{Ij, pq}), calculate corresponding coefficient A according to lateral inhibition coefficient summation ∑; Generation lateral inhibition matrix of coefficients H (p, q); According to receptor unit, center N _{I, j}Input e _{I, j}, peripheral receptor unit N _{P, q}Input e _{P, q}, call formula: r _{I, j}=e _{I, j}-H (p, q) * e _{P, q}Determine that the center experiences unit N _{I, j}Output r _{I, j}, wherein, k _{Ij, pq}Be the lateral inhibition coefficient, A and σ are coefficient, d _{Ij, pq}Expression receptor unit N _{P, q}To receptor unit, center N _{I, j}Distance.

2. method according to claim 1, it is characterized in that, with the contrast after the figure image intensifying, Y-PSNR, information entropy and lateral inhibition coefficient and relation curve carry out normalization according to the linear transformation function after, get the corresponding lateral inhibition coefficient of flex point that information entropy, contrast rise and Y-PSNR descends and be optimum lateral inhibition summation ∑.

3. method according to claim 1 is characterized in that, is that lateral inhibition coefficient distributed model is specially with monolateral damped expoential model conversion, changes the horizontal ordinate of monolateral decaying exponential function into receptor unit N by time t _{I, j}And N _{P, q}Between apart from d _{Ij, pq}, N wherein _{I, j}The expression position is (i, receptor unit, center j), N _{P, q}The expression position is (p, receptor unit q); Ordinate is converted to about receptor cell distance d by the function f (t) about time t _{Ij, pq}The lateral inhibition coefficient k _{Ij, pq}

4. method according to claim 1 is characterized in that, gets ∑ from 0.2 to 1, adopts stepping △=0.01, determines corresponding coefficient A according to lateral inhibition coefficient summation ∑.

5. method according to claim 1 is characterized in that, generates the lateral inhibition matrix of coefficients and is specially: according to the lateral inhibition coefficient k _{Ij, pq}Call formula:

Determine matrix of coefficients.

Images