CN103810732B - A kind of dynamic texture generation method - Google Patents

Abstract

The invention provides a kind of dynamic texture generation method, the method comprises the steps: that S1. definition generates size x*y and picture number n of texture image；S2. several obediences of stochastic generation alpha Stable distritation S_αThe stationary increment X of (σ, β, μ)；S3. choose from the stationary increment X of stochastic generation in the 2 dimension matrixes that x*y variable write pre-sets, utilize random midpoint displacement method that this matrix is processed, obtain the isotropism texture image that size is x*y；S4. step S2 S3 is repeated, until generating n frame isotropism texture image；S5. n two field picture write video file step S4 obtained, generates isotropism dynamic texture.Isotropism texture image input structure wave filter can also be obtained anisotropic texture image, and then generates anisotropy dynamic texture in the present invention.The method of the invention can generate more rich dynamic texture.

Description

A kind of dynamic texture generation method

Technical field

The invention belongs to field of virtual reality, be specifically related to a kind of dynamic texture generation method.

Background technology

Texture is the natural surface that a class is complicated, and what it was expressed it is critical only that the mathematics to texture is built Mould.At present, the texture that manually generated texture predominantly produces based on FBM model, FBM mould Type assumes that texture has preferable self-similarity, its increment Gaussian distributed, and has each To the feature of the same sex.

But, during actual natural texture is analyzed, find the most eurypalynous texture not There is the statistical property of FBM model, such as rock surface, carry alveolate sea water background etc., only Minority such as hydrothermal solution texture is had to meet FBM model.The FBM model that has its source in of problem is assumed: The increment Gaussian distributed of texture, this makes it that description of texture is confined to certain scope In, it is impossible to widely texture is described.

Summary of the invention

For the deficiencies in the prior art, the present invention provides a kind of dynamic texture generation method, it is possible to Generate abundant dynamic texture.

For achieving the above object, the present invention is achieved by the following technical programs:

A kind of dynamic texture generation method, the method comprises the steps:

S1. definition generates size x*y and picture number n of texture image；

S2. several obediences of stochastic generation alpha Stable distritation S_αThe stationary increment X of (σ, β, μ), Wherein α is characterized the factor, σ is scale parameter, β is position skew centered by the deflection factor, μ Parameter；

S3. from the stationary increment X of stochastic generation, choose x*y variable write 2 pre-set In dimension matrix, utilizing random midpoint displacement method to process this matrix, obtaining size is x*y Isotropism texture image；

S4. step S2-S3 is repeated, until generating n frame isotropism texture image；

S5. n two field picture write video file step S4 obtained, generates the dynamic stricture of vagina of isotropism Reason.

Wherein, described utilizing random midpoint displacement method to process this matrix, obtaining size is The isotropism texture image of x*y is:

Take the meansigma methods of four end points pixel values in matrix and add the random file amount pre-set As central point pixel value, all points in matrix are processed in a manner described, i.e. obtains Size is the isotropism texture image of x*y, and the random file amount wherein pre-set is obeyed Alpha Stable distritation.

Wherein, include after step s4 n two field picture is done normalized so that it is pixel value For [0,255].

Wherein, before step S5, include that n two field picture does dependency to be processed.

Further, the method also includes:

S6. every frame isotropism texture image input structure wave filter step S4 generated, obtains N frame anisotropic texture image；

S7. n two field picture write video file step S6 obtained, generates the dynamic stricture of vagina of anisotropy Reason.

Wherein, described step S6 includes:

If the system function of Structure Filter is:

H_φ(ω_x,ω_y)=(1+ α-2 α cos²(ω_θ-θ₀))^-1,

Whereinθ₀For deflection parameter, α is intensive parameter；

The n frame isotropism texture image utilizing 2 dimension FFT step S4 to be generated becomes successively Change to frequency domain, and be designated as S_I(ω_x,ω_y)；

Every frame isotropism texture image is input to Structure Filter, obtains anisotropic texture Image S_A(ω_x,ω_y), wherein S_A(ω_x,ω_y)=S_I(ω_x,ω_y)H_φ(ω_x,ω_y)；

By S_A(ω_x,ω_y) converted by the IFFT of 2 dimensions, obtain the anisotropic texture image of time domain.

Wherein, include n two field picture is done normalized after step S6 so that it is pixel value For [0,255].

Wherein, before step S7, include that n two field picture does dependency to be processed.

The present invention at least has a following beneficial effect:

The dynamic texture generation method that the present invention provides, it is possible to generate more horn of plenty and fine and smooth from So texture, such as rock surface or carry isotropism or the anisotropy such as alveolate sea water background Natural texture, for traditional texture that can not generate based on FBM model, the present invention provides A kind of new mode carries out the generation of complex texture, meets the needs of people.

Accompanying drawing explanation

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below The accompanying drawing used required in embodiment or description of the prior art will be briefly described, aobvious and Easily insight, the accompanying drawing in describing below is some embodiments of the present invention, common for this area From the point of view of technical staff, on the premise of not paying creative work, it is also possible to according to these accompanying drawings Obtain other accompanying drawing.

Fig. 1 is the flow chart of dynamic texture generation method in the embodiment of the present invention；

Fig. 2 is the isotropism texture that dynamic texture generation method described in the embodiment of the present invention generates Schematic diagram；

Fig. 3 is the anisotropic texture that dynamic texture generation method described in the embodiment of the present invention generates Schematic diagram.

Detailed description of the invention

For making the purpose of the embodiment of the present invention, technical scheme and advantage clearer, below will knot Close the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, Complete description, it is clear that described embodiment be a part of embodiment of the present invention rather than Whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not having Make the every other embodiment obtained under creative work premise, broadly fall into present invention protection Scope.

The embodiment of the present invention proposes a kind of dynamic texture generation method, sees Fig. 1, including following Step:

Step 101: definition generates size x*y and picture number n of texture image.

In this step, definition generates size x*y and picture number n of texture image, such as, define The texture image size generated is 128*128, altogether 129 frame.

Step 102: stochastic generation several obediences alpha Stable distritation S_αThe steady increasing of (σ, β, μ) Amount X.

Usually, the probability density function of alpha Stable distritation, in addition to minority special case, does not exist and closes Formula is expressed, and typically describes its distribution character with characteristic function.It is steady that stochastic variable X obeys alpha Fixed distribution, then its characteristic function form is:

$\mathrm{Eexpi\θX}=\left\{\begin{array}{c}\exp \{-{\mathrm{\σ}}^{\mathrm{\α}}{\left|\mathrm{\θ}\right|}^{\mathrm{\α}}(1-\mathrm{i\β}\left(\mathrm{sign\θ}\right)\tan \frac{\mathrm{\π\α}}{2})+\mathrm{i\μ\θ}\},\mathrm{\α}\≠1\\ \exp \{-\mathrm{\σ}|\mathrm{\θ}|(1+\mathrm{i\β}\frac{\mathrm{\π}}{2}\left(\mathrm{sign\θ}\right)\ln |\mathrm{\θ}\left|\right)+\mathrm{i\μ\θ}\},\mathrm{\α}\≠1\end{array}\right.$

Wherein, sign θ is sign function, 0 ＜ α≤2, σ >=0 ,-1≤β≤1 and real number μ. Meet four parameters that this characteristic function is expressed, the canonical parameter of referred to as alpha Stable distritation System.Wherein, four parameters are respectively: characterization factor α, scale parameter σ, deflection factor-beta, Centre position deviation parameter μ, alpha Stable distritation is abbreviated as S_α(σ,β,μ)。

In this step, several obediences of stochastic generation alpha Stable distritation S_α(σ, β, μ's) is steady Increment X.Here with average (scale parameter) for 0, dispersion degree (Centre position deviation parameter) As a example by being the alpha Stable distritation of 1, it also has two parameters, is characterization factor α respectively, Deflection factor-beta, is designated as S_α(1, β, 0), concrete generating mode is as follows:

Making γ is the equally distributed stochastic variable on (-pi/2, pi/2), W be an average be 1 Exponential random variable, and γ and W is separate, when α ≠ 1, obeys alpha stable Distribution S_αThe stationary increment X of (0, β, 1) can be described as:

$X=\frac{\sin \mathrm{\α}(\mathrm{\γ}-{\mathrm{\γ}}_0)}{{\left(\cos \mathrm{\γ}\right)}^{1/\mathrm{\α}}}{\left(\frac{\cos (\mathrm{\γ}-\mathrm{\α}(\mathrm{\γ}-{\mathrm{\γ}}_0))}{W}\right)}^{(1-\mathrm{\α})/\mathrm{\α}}$

X obeys S_α(1, β, 0) is distributed, whereinK (α) is sign function, As α ＞ 1, K (α)=α；As α ＜ 1, K (α)=α-2.

When α=1, obey alpha Stable distritation S_αThe stationary increment X of (0, β, 1) can be described as:

$X=(\frac{\mathrm{\π}}{2}+\mathrm{\β\γ})\tan \mathrm{\γ}-\mathrm{\β}\log \left(\frac{W\cos \mathrm{\γ}}{\frac{\mathrm{\π}}{2}+\mathrm{\β\γ}}\right).$

In the present embodiment, the parameter taking alpha Stable distritation is α=1.7, and therefore β=0 gives birth to Become to obey alpha Stable distritation S_1.7The stationary increment of (1,0,0), due to when α ≠ 1,

$X=\frac{\sin \mathrm{\α}(\mathrm{\γ}-{\mathrm{\γ}}_0)}{{\left(\cos \mathrm{\γ}\right)}^{1/\mathrm{\α}}}{\left(\frac{\cos (\mathrm{\γ}-\mathrm{\α}(\mathrm{\γ}-{\mathrm{\γ}}_0))}{W}\right)}^{(1-\mathrm{\α})/\mathrm{\α}}$

WhereinK (α) is sign function, as α ＞ 1, and K (α)=α, Several stationary increments are calculated, the stationary increment calculated here according to the expression formula of above-mentioned X Number is greater than x*y.

Step 103: choose x*y variable from the stationary increment X of stochastic generation and write in advance In the 2 dimension matrixes arranged, utilize random midpoint displacement method that this matrix is processed, obtain big The little isotropism texture image for x*y.

In this step, from the stationary increment X of stochastic generation, x*y variable write is chosen pre- In the 2 dimension matrixes first arranged, then take the meansigma methods of four end points pixel values in matrix and add All in matrix as central point pixel value, are pressed above-mentioned by the random file amount pre-set Mode processes, and i.e. obtains the isotropism texture image that size is x*y, sets the most in advance The random file amount put obeys alpha Stable distritation.Finally the data texturing generated is carried out yardstick All pixel values are changed to [0,255] by change.

Step 104: judge whether to need to generate anisotropic texture, if desired, perform step 105, otherwise perform step 106.

Step 105: isotropism texture image is generated anisotropic texture by Structure Filter Image.

In this step, described Structure Filter is for characterizing anisotropic Structure Filter, complete Entirely feature the anisotropic character of texture, be mapped to anisotropic texture from isotropism texture, Can be equivalent to isotropism texture by a Structure Filter meeting ad hoc structure function. Anisotropic texture F_A(x y) can be considered isotropism texture F_I(x, y) by wave filter h, (u v) obtains Arrive, it may be assumed that

F_A(x, y)=∫ F_I(x-u,y-v)h(u,v)d(u,v)；

Corresponding frequency spectrum designation is:

S_A(ω_x,ω_y)=S_I(ω_x,ω_y)H_φ(ω_x,ω_y),

Wherein, H_φ(ω_x,ω_y) it is the frequency domain representation of wave filter, Structure Filter concrete form is: H_φ(ω_x,ω_y)=(1+ α-2 α cos²(ω_θ-θ₀))^-1, wherein,Parameter alpha ∈ [0,1] is The intensity of anisotropic texture, parameter θ₀∈ [0, π] is the direction of anisotropic texture.

Structure Filter does not change the autocorrelation of texture, simply changes its Direction Distribution Characteristics, The deflection parameter of setting structure wave filter in the present embodimentIntensive parameter α=0.7.

First with 2 dimension FFT, the isotropism texture image obtained is transformed to frequency domain, And it is designated as S_I(ω_x,ω_y).Anisotropic texture can be considered each tropism texture sound by Structure Filter Should, every frame isotropism texture image is input to Structure Filter, obtains anisotropic texture Image S_A(ω_x,ω_y), wherein S_A(ω_x,ω_y)=S_I(ω_x,ω_y)H_φ(ω_x,ω_y)；

By S_A(ω_x,ω_y) converted by the IFFT of 2 dimensions, obtain the anisotropic texture image of time domain；

Finally, the data texturing generated is carried out dimensional variation, all pixel values are changed to [0,255]。

Step 106: judge that the number of the texture image generated, whether equal to n, if being equal to, performs Step 107, otherwise performs step 102.

Step 107: n frame texture image is done dependency and processes.

In this step, in order to make the texture more continuity of generation, need to increase image sequence Dependency each other, concrete operations be adjacent some two field pictures summation is averaged after make It is a two field picture (such as 5 adjacent two field picture summations being averaged), in this manner to all Image increases dependency successively.

Step 108: n frame texture image is write video file, generates dynamic texture.

In this step, the image write video file after dependency is processed, generate dynamic stricture of vagina Reason.

The dynamic texture generation method that the embodiment of the present invention provides, it is possible to generate more horn of plenty with thin Greasy natural texture, such as rock surface or carry the isotropism such as alveolate sea water background or each Heterotropic natural texture, for traditional texture that can not generate based on FBM model, this Bright provide a kind of new mode and carry out the generation of complex texture, meet the needs of people.

Above example is merely to illustrate technical scheme, is not intended to limit；Although With reference to previous embodiment, the present invention is described in detail, those of ordinary skill in the art It is understood that the technical scheme described in foregoing embodiments still can be modified by it, Or wherein portion of techniques feature is carried out equivalent；And these amendments or replacement, do not make The essence of appropriate technical solution departs from the spirit and scope of various embodiments of the present invention technical scheme.

Claims (8)

1. a dynamic texture generation method, it is characterised in that the method comprises the steps:

S1. definition generates size x*y and picture number n of texture image；

S2. several obediences of stochastic generation alpha Stable distritation S_αThe stationary increment X of (σ, β, μ), Wherein α is characterized the factor, σ is scale parameter, β is position skew centered by the deflection factor, μ Parameter；

S3. from the stationary increment X of stochastic generation, choose x*y variable write 2 pre-set In dimension matrix, utilizing random midpoint displacement method to process this matrix, obtaining size is x*y Isotropism texture image；

S4. step S2-S3 is repeated, until generating n frame isotropism texture image；

S5. n two field picture write video file step S4 obtained, generates the dynamic stricture of vagina of isotropism Reason.

Method the most according to claim 1, it is characterised in that the random midpoint of described utilization This matrix is processed by displacement method, obtains the isotropism texture image that size is x*y and is:

Take the meansigma methods of four end points pixel values in matrix and add the random file amount pre-set As central point pixel value, all points in matrix are processed in a manner described, i.e. obtains Size is the isotropism texture image of x*y, and the random file amount wherein pre-set is obeyed Alpha Stable distritation.

Method the most according to claim 2, it is characterised in that wrap after step s4 Include and n two field picture is done normalized so that it is pixel value is [0,255].

Method the most according to claim 3, it is characterised in that wrapped before step S5 Include and n two field picture is done dependency process.

Method the most according to claim 4, it is characterised in that the method also includes:

S6. every frame isotropism texture image input structure wave filter step S4 generated, obtains N frame anisotropic texture image；

S7. n two field picture write video file step S6 obtained, generates the dynamic stricture of vagina of anisotropy Reason.

Method the most according to claim 5, it is characterised in that described step S6 includes:

The system function of described Structure Filter is:

H_φ(ω_x,ω_y)=(1+ α-2 α cos²(ω_θ-θ₀))^-1,

Whereinθ₀For deflection parameter, α is intensive parameter, ω_x、ω_yPoint Do not represent two dimensional filter frequency in the two directions x and y；

The n frame isotropism texture image utilizing 2 dimension FFT step S4 to be generated becomes successively Change to frequency domain, and be designated as S_I(ω_x,ω_y)；

Every frame isotropism texture image is input to Structure Filter, obtains anisotropic texture Image S_A(ω_x,ω_y), wherein S_A(ω_x,ω_y)=S_I(ω_x,ω_y)H_φ(ω_x,ω_y)；

By S_A(ω_x,ω_y) converted by the IFFT of 2 dimensions, obtain the anisotropic texture image of time domain.

Method the most according to claim 6, it is characterised in that wrap after step S6 Include and n two field picture is done normalized so that it is pixel value is [0,255].

Method the most according to claim 7, it is characterised in that wrapped before step S7 Include and n two field picture is done dependency process.