CN105913393A - Self-adaptive wavelet threshold image de-noising algorithm and device - Google Patents

Abstract

The invention brings forward a self-adaptive wavelet threshold image de-noising algorithm and device. The image de-noising algorithm comprises the following steps: a noised image is subjected to wavelet transformation operation, and wavelet coefficients of all layers can be obtained; with signal correlation considered, coefficients in an area adjacent to each coefficient are averaged in wavelet coefficients of each layer; threshold is determined based on a wavelet coefficient which is obtained via an absolute mean value estimation method, and a self-adaptive threshold method is adopted for determining thresholds suitable for all different scales; as for the wavelet coefficients and thresholds, self-adaptive threshold functions for all directions at all layers are constructed, wavelet inverse transformation and reconstruction are performed, and a de-noised image can be obtained. According to the image de-noising algorithm, the self-adaptive threshold method is adopted for determining the thresholds, an overall uniform threshold is replaced with thresholds for different scales, wavelet threshold de-noising operation is performed via use of the self-adaptive thresholds and the self-adaptive threshold functions, and detailed information of the image can be protected; the self-adaptive wavelet threshold image de-noising algorithm is better than a conventional wavelet threshold de-noising algorithm in terms of peak signal to noise ratio and visual perception.

Description

A kind of adaptive wavelet threshold Image denoising algorithm and device

Technical field

The present invention relates to the Image Denoising Technology field in Digital Image Processing, particularly to a kind of adaptive wavelet threshold Image denoising algorithm and device.

Background technology

The impact of the reasons such as digital picture can be subject to such as sensor oscillation during producing, electronic device interference, leads The digital picture quality obtained after causing conversion declines, and have impact on the understanding to picture material.In order to ensure the correct of subsequent treatment Property, need image is carried out denoising.The application of Image Denoising Technology expands to biomedicine, Information Center from aeronautical field Every field and the industries such as, Resources and environmental sciences, astronomy, physics, industry, agricultural, national defence, education, art, to warp Ji, military, culture and daily life produce significant impact.Therefore, the research of Image Denoising Technology has of crucial importance Using value.

The method of image denoising has a variety of, is broadly divided into spatial domain and frequency domain two kinds.Traditional image denoising mainly exists Spatial domain realizes, and main method has mean filter, medium filtering and Wiener filtering.But the denoising effect of these methods is not Preferable, although noise can be removed, but the image after denoising there will be the phenomenon of fuzzy distortion.Frequency domain denoising method is to scheme As by space field transformation to frequency domain, the conversion coefficient in frequency domain is carried out denoising, then the coefficient after denoising is carried out contravariant gain The purpose of denoising is reached to spatial domain.Conventional method has Fourier transformation, wavelet transformation etc..Wavelet analysis is international in recent years On start a modish research frontier, be the breakthrough of Fourier analysis of continuing.Small echo with its good time Frequently characteristic and multi-resolution characteristics so that it is be widely used in denoising field.

Existing Wavelet noise-eliminating method is broadly divided into three classes: being correlated with denoising in (1) spatial domain, utilizes signal wavelet coefficient at each chi There is between degree dependency denoising；(2) denoising based on Singularity Detection, utilizes signal and noise to have different singularitys and goes Make an uproar；(3) wavelet threshold de-noise, is produced this hypothesis denoising according to the coefficient that amplitude is bigger by signal of interest.Due to first two Method calculates complexity, and the uncertainty affecting denoising effect is more, and therefore practicality is the strongest.And realization is simple and effect is preferable Wavelet threshold denoising algorithm, is most commonly used method of studying at present.

The principle of wavelet threshold denoising is that image carries out wavelet transformation, and bigger wavelet coefficient is typically all actual signal Be main, less coefficient be the most largely noise, therefore can by set suitable threshold value, the coefficient less than threshold value is put Zero, and remain larger than the wavelet coefficient of threshold value, being then passed through threshold function table mapping obtains estimation coefficient, finally enters estimation coefficient Row inverse transformation, it is achieved image denoising and reconstruction.Two fundamentals in wavelet threshold denoising are threshold value and threshold function table, they Selection be the key of denoising.

Frequently with threshold function table have hard-threshold and soft-threshold function.Their basic thought is all to remove little wavelet systems Number, shrinks big wavelet coefficient and retains.Hard threshold function is defined as:

${\hat{W}}_{j,k}=\left\{\begin{array}{cc}{W}_{j,k}& \left|W_{j,k}\right|\&GreaterEqual;\mathrm{\&lambda;}\\ 0& \left|W_{j,k}\right|<\mathrm{\&lambda;}\end{array}\right.$

It is exactly the absolute value of signal wavelet coefficient to be compared with threshold value, the wavelet systems less than threshold value that hard-threshold processes Number is set to zero, keeps constant more than the wavelet coefficient of threshold value.

Soft thresholding formula is expressed as follows:

${\hat{W}}_{j,k}=\left\{\begin{array}{cc}sign\left(W_{j,k}\right)\&CenterDot;\left(\right|W_{j,k}|-\mathrm{\&lambda;})& \left|W_{j,k}\right|\&GreaterEqual;\mathrm{\&lambda;}\\ 0& \left|W_{j,k}\right|<\mathrm{\&lambda;}\end{array}\right.$

Soft-threshold processes the difference becoming this point and threshold value exactly wavelet coefficient being more than threshold value.

In formula, λ is Donoho wavelet threshold,W_j,kFor wavelet coefficient,Estimate for wavelet coefficient Value.Although both approaches is used widely in practice, also achieves better effects, but method itself has some potential Shortcoming.As in hard thresholding method,It is discontinuous at λ, utilizesReconstruct gained signal may produce Vibration；Estimated by Soft thresholdingAlthough overall seriality is good, but works as | W_j,k| during ＞ λ,With W_j,kAlways deposit In constant deviation, directly affect the approximation ratio of reconstruction signal and actual signal.

The determination of optimal threshold λ is a key issue in wavelet threshold denoising method.Conventional Research on threshold selection has Donoho Global thresholding and Birge-Massart threshold strategies:

(1) global threshold, i.e. Visushrink generic threshold value:Wherein σ is noise variance, and N is letter Number length.This threshold value draws for multidimensional independent normal variate Joint Distribution under Gauss model；

(2) Birge-Massart strategy, i.e. local threshold: a given Decomposition order j specified, to j+1 and higher All coefficients retain；I-th (i=1～j) layer is retained to the M/ (j+2-i) of maximum absolute value^αIndividual wavelet coefficient, wherein M is Empirical coefficient, remaining sets to 0.

Tradition major part method is all the global threshold used.Although global threshold calculates simple, but can cause not Gibbs' effect at continuity point, and the mistake of high energy signal is strangled, thus affect noise reduction.

Summary of the invention

It is an object of the invention to the shortcoming overcoming prior art with not enough, it is provided that a kind of adaptive wavelet threshold image goes Make an uproar algorithm and device.

According to disclosed embodiment, a first aspect of the present invention proposes a kind of adaptive wavelet threshold image denoising and calculates Method, described algorithm comprises the following steps:

S1, picture signal is carried out wavelet decomposition step, select suitable small echo and set highest level N of decomposition, meter (i, j) at the wavelet coefficient of each layer to calculate noisy picture signal s；

S2, the thresholding step of wavelet coefficient, determine suitable threshold value, to respectively in each layer all directions of wavelet decomposition Layer detail wavelet coefficients uses suitable threshold function table to process, and is former to retain the wavelet coefficient of picture signal as far as possible Then, the wavelet coefficient making noise is zero；

S3, reconstruct picture signal step, utilize the approximate part wavelet coefficient of n-th layer and at the 1st layer to n-th layer process Each detail wavelet coefficients reconstruct picture signal of reason.

Further, described S1, picture signal is carried out wavelet decomposition step particularly as follows:

S11, to noisy picture signal s, (i, j) carries out 3 layers of wavelet transformation, produces low frequency coefficient and direction respectively after decomposition It is level, vertical, the high frequency coefficient at diagonal angle；

S12, according to above layers wavelet coefficient, it is considered to its signal correlation, the coefficient of each coefficient adjacent area is asked Average:

S_i,j,k=S/R²

$S=\underset{m=-\frac{R-1}{2}}{\overset{m=\frac{R-1}{2}}{\&Sigma;}}\underset{n=-\frac{R-1}{2}}{\overset{n=\frac{R-1}{2}}{\&Sigma;}}{\left(W_{i,j+m,k+n}\right)}^2,i=1:N$

Wherein, R represents the field averaged, W_i,j,kBeing the wavelet coefficient of i-th layer, j, k represent pixel in image Position, S_i,j,kFor the wavelet coefficient using absolute average to retrieve after estimating.

Further, the choosing method of described threshold values is specific as follows:

Using Adaptive Thresholding to determine described threshold value, the threshold value being adapted to each layer is:

λ_i=(N+1-i) 2 β σ²Log (R) i=1:N, β ∈ (0,1)

Wherein, σ is noise variance, is estimated as follows shown in formula to it:

σ=median (| W_HH1|)/0.6745

W_HH1Represent ground floor HH sub-band coefficients.

Further, the building method of described threshold function table is specific as follows:

According to the wavelet coefficient obtained, it is constructed as follows the adaptive thresholding value function of wavelet coefficient in each layer all directions:

${\hat{W}}_{i,j,k}=\left\{\begin{array}{cc}{W}_{i,j,k}(1-\frac{{\mathrm{\&alpha;\&lambda;}}_i}{(N+1-i)S_{i,j,k}})& S_{i,j,k}\&GreaterEqual;{\mathrm{\&lambda;}}_i\\ 0& S_{i,j,k}<{\mathrm{\&lambda;}}_i\end{array}\right.,\mathrm{\&alpha;}\&Element;(0,1)$

Wherein α is attenuation quotient, and i is the wavelet decomposition number of plies, λ_iIt is the threshold value of i-th layer, W_i,j,kIt is that i-th layer of decomposition obtains Original wavelet coefficients,For the estimated value of purified signal wavelet coefficient, j, k represent the position of pixel in image.

Further, described S3, reconstruct picture signal step particularly as follows:

Described purified signal wavelet coefficient after utilizing threshold function table to processCarry out inverse wavelet transform and reconstruct, obtaining The denoising picture signal recovered.

Further, the field R=5 averaged described in.

Further, described regulatory factor β=0.3, described attenuation quotient α=0.1.

According to disclosed embodiment, a second aspect of the present invention proposes a kind of adaptive wavelet threshold image denoising dress Putting, described device includes following modules:

Wavelet decomposition module, for picture signal is carried out wavelet decomposition, selects suitable small echo and sets decomposition High-level N, (i, j) at the wavelet coefficient of each layer to calculate noisy picture signal s；

Wavelet coefficient module, for the threshold process of wavelet coefficient, it is suitable to determine in each layer all directions of wavelet decomposition Threshold value, to each layer detail wavelet coefficients use suitable threshold function table process, to retain the little of picture signal as far as possible Wave system number is principle, and the wavelet coefficient making noise is zero；

Image reconstruction module, is used for reconstructing picture signal, utilize n-th layer approximate part wavelet coefficient and from the 1st layer to N-th layer treated each detail wavelet coefficients reconstruct picture signal.

Further, described wavelet decomposition module includes:

Wavelet coefficient resolving cell, for noisy picture signal s, (i j) carries out 3 layers of wavelet transformation, produces low after decomposition Frequently coefficient and direction are level, vertical, the high frequency coefficient at diagonal angle respectively；

Wavelet coefficient averaging unit, for according to above layers wavelet coefficient, it is considered to its signal correlation, to each coefficient The coefficient of adjacent area is averaging:

S_i,j,k=S/R²

$S=\underset{m=-\frac{R-1}{2}}{\overset{m=\frac{R-1}{2}}{\&Sigma;}}\underset{n=-\frac{R-1}{2}}{\overset{n=\frac{R-1}{2}}{\&Sigma;}}{\left(W_{i,j+m,k+n}\right)}^2,i=1:N$

Wherein, R represents the field averaged, W_i,j,kBeing the wavelet coefficient of i-th layer, j, k represent pixel in image Position, S_i,j,kFor the wavelet coefficient using absolute average to retrieve after estimating.

Further, described wavelet coefficient module includes:

Threshold values determines unit, is used for using Adaptive Thresholding to determine described threshold value, and the threshold value being adapted to each layer is:

λ_i=(N+1-i) 2 β σ²Log (R) i=1:N, β ∈ (0,1)

Wherein, σ is noise variance, is estimated as follows shown in formula to it:

σ=median (| W_HH1|)/0.6745

W_HH1Represent ground floor HH sub-band coefficients；

Threshold values construction of function unit, for according to the wavelet coefficient obtained, being constructed as follows wavelet coefficient in each layer all directions Adaptive thresholding value function:

${\hat{W}}_{i,j,k}=\left\{\begin{array}{cc}{W}_{i,j,k}(1-\frac{{\mathrm{\&alpha;\&lambda;}}_i}{(N+1-i)S_{i,j,k}})& S_{i,j,k}\&GreaterEqual;{\mathrm{\&lambda;}}_i\\ 0& S_{i,j,k}<{\mathrm{\&lambda;}}_i\end{array}\right.,\mathrm{\&alpha;}\&Element;(0,1)$

Wherein α is attenuation quotient, and i is the wavelet decomposition number of plies, λ_iIt is the threshold value of i-th layer, W_i,j,kIt is that i-th layer of decomposition obtains Original wavelet coefficients,For the estimated value of purified signal wavelet coefficient, j, k represent the position of pixel in image.

The present invention has such advantages as relative to prior art and effect:

1) the adaptive wavelet threshold Image denoising algorithm that the present invention proposes has taken into full account the field dependency of signal, will The coefficient of each coefficient adjacent area be applied to threshold value choose with in the structure of threshold function table；

2) use Adaptive Thresholding to determine threshold value, threshold values different for each yardstick level is replaced the uniform threshold of the overall situation；

3) shortcoming that the threshold function table constructed overcomes firmly, Soft thresholding exists, has adaptivity；

4) by utilization, there is the threshold value of adaptivity and threshold function table carries out wavelet threshold denoising, can not only effectively remove White Gaussian noise, moreover it is possible to retain edge and the detailed information of image well.

Accompanying drawing explanation

Fig. 1 is the denoising theory diagram of the adaptive wavelet threshold Image denoising algorithm disclosed in the present invention；

Fig. 2 is three layers of exploded view of small echo in the adaptive wavelet threshold Image denoising algorithm disclosed in the present invention；

Fig. 3 is the denoising result contrast of the adaptive wavelet threshold Image denoising algorithm disclosed in the present invention and other algorithms Figure；

Fig. 4 is the structure composition frame chart of the adaptive wavelet threshold image denoising device disclosed in the present invention.

Detailed description of the invention

For making the purpose of the present invention, technical scheme and advantage clearer, clear and definite, develop simultaneously embodiment pair referring to the drawings The present invention further describes.Should be appreciated that specific embodiment described herein, and need not only in order to explain the present invention In limiting the present invention.

Embodiment one

As it is shown in figure 1, present embodiment discloses a kind of adaptive wavelet threshold Image denoising algorithm, comprise the following steps:

S1, picture signal is carried out wavelet decomposition: select suitable small echo and set highest level N of decomposition, calculate figure (i, j) at the wavelet coefficient of each layer for image signal s；

S11, as in figure 2 it is shown, noisy image carries out 3 layers of wavelet transformation, produces low frequency coefficient and direction difference after decomposition Being level, vertical, the high frequency coefficient at diagonal angle, wherein the energy of signal, the key character information of image are concentrated mainly on low frequency and put down On sliding component LL.

S12, each layer wavelet coefficient obtained according to step S11, it is considered to its signal correlation, to each coefficient adjacent area Coefficient be averaging:

S_i,j,k=S/R²

$S=\underset{m=-\frac{R-1}{2}}{\overset{m=\frac{R-1}{2}}{\&Sigma;}}\underset{n=-\frac{R-1}{2}}{\overset{n=\frac{R-1}{2}}{\&Sigma;}}{\left(W_{i,j+m,k+n}\right)}^2,i=1:N$

Wherein, R represents the field averaged, and here takes R=5, W_i,j,kBeing the wavelet coefficient of i-th layer, j, k represent figure The position of pixel, S in Xiang_i,j,kFor the wavelet coefficient using absolute average to retrieve after estimating.

S2, the threshold process of wavelet coefficient: in each layer all directions of wavelet decomposition, determine suitable threshold value, thin to each layer Joint wavelet coefficient uses suitable threshold function table to process, and makes the wavelet coefficient of signal retain as much as possible, makes noise Wavelet coefficient is zero；

The present invention proposes new Research on threshold selection, it may be assumed that

Using the threshold value that Adaptive Thresholding determines, the threshold value being adapted to each layer is:

λ_i=(N+1-i) 2 β σ²Log (R) i=1:N, β ∈ (0,1)

The present embodiment takes regulatory factor β=0.3.σ is noise variance, is estimated as follows shown in formula to it:

σ=median (| W_HH1|)/0.6745

W_HH1Represent ground floor HH sub-band coefficients.

The present invention proposes new threshold function table building method, it may be assumed that

According to the wavelet coefficient obtained, it is constructed as follows the adaptive thresholding value function of wavelet coefficient in each layer all directions:

${\hat{W}}_{i,j,k}=\left\{\begin{array}{cc}{W}_{i,j,k}(1-\frac{{\mathrm{\&alpha;\&lambda;}}_i}{(N+1-i)S_{i,j,k}})& S_{i,j,k}\&GreaterEqual;{\mathrm{\&lambda;}}_i\\ 0& S_{i,j,k}<{\mathrm{\&lambda;}}_i\end{array}\right.,\mathrm{\&alpha;}\&Element;(0,1)$

Wherein α is attenuation quotient, and i is the wavelet decomposition number of plies, λ_iIt is the threshold value of i-th layer, W_i,j,kIt is that i-th layer of decomposition obtains Original wavelet coefficients,For the estimated value of purified signal wavelet coefficient, j, k represent the position of pixel in image.This enforcement Example takes attenuation quotient α=0.1.

S3, reconstruct picture signal: utilize the approximate part wavelet coefficient of n-th layer and treated to n-th layer from the 1st layer Each detail wavelet coefficients reconstruct picture signal.

S31, utilize threshold function table to process after purified signal wavelet coefficient enter traveling wave coefficient and carry out inverse wavelet transform and lay equal stress on Structure, the picture signal being restored is final denoising image.

Calculate the value of Y-PSNR PSNR corresponding to this signal.

Calculating the PSNR value recovering signal corresponding in this example is:

$PSNR=10{\log }_{10}\left(\frac{{255}^2}{MSE}\right)$

Wherein MSE is mean square error, is defined as:

$MSE=\underset{m=0}{\overset{M-1}{\&Sigma;}}\underset{n=0}{\overset{N-1}{\&Sigma;}}{\&lsqb;f(m,n)-\hat{f}(m,n)\&rsqb;}^2/MN$

Wherein f (m, n) is original noise-free picture signal,Recovery signal after denoising, MN is that the pixel of image is big Little.

Table 1 compares for single image denoising effect, uses and not only comprises smooth but also have the Lena of abundant detail textures to scheme As test object, in test image, superposition average respectively is 0, and standard deviation (σ) is the white Gaussian of 10,15,20,25,30 Noise.In order to verify the superiority of Image denoising algorithm that this patent proposed, it is utilized respectively traditional hard-threshold and soft-threshold Denoising Algorithm and herein algorithm carry out denoising to adding image of making an uproar, and obtain by contrasting respective peak value to-noise ratio (PSNR) Comparison to denoising effect.Comparative result is as shown in table 1.

The each algorithm of table 1 adds the denoising result of image of making an uproar and compares (PSNR) Lena

After objectively comparing the denoising result of this algorithm and traditional algorithm, contrast several going from subjective evaluation below Make an uproar the actual effect figure of algorithm.One Lena image is added average be 0, standard deviation be the Gaussian noise of 20, each algorithm is to it Image after denoising is as shown in Figure 3.

Table 2 is algorithm Common Testing, and the multiple image in standard picture storehouse is carried out denoising emulation experiment, chooses five Representative picture: Baboon, Barbara, Fishingboat, Goldhill, Peppers.In these images, have Detailed information abundanter, some marginal informations are more, also have some complex scenes and simple fine and smooth smoothed image, and total comes Say and can represent real-life major applications scene.Adding average in these images is 0, and standard deviation is the Gauss of 20 White noise, application traditional algorithm and this paper algorithm are tested respectively, and experimental result is as shown in table 2.

The each algorithm of table 2 adds the denoising result of image of making an uproar and compares (PSNR) difference

By the comparing result of Tables 1 and 2, it can be seen that the adaptive wavelet threshold Denoising Algorithm of the present embodiment is respectively Plant denoising effect under noise level and be all substantially better than additive method, and be applicable to different types of image.

Below from objective experimental data, not only demonstrate the superiority of the present invention, and by human eye subjective judgement denoising After image also can embody the advantage of algorithm denoising effect herein.This algorithm overcomes the shortcoming of conventional threshold values denoising method, Picture quality after denoising is high, and edge detail information also can be sufficiently reserved.By to add that the Baboon image denoising made an uproar processes can To find out, the situation that this algorithm is complicated and changeable for picture material, to comprise quantity of information bigger, have and have the denoising of uniqueness excellent Gesture.

Embodiment two

As shown in Figure 4, present embodiment discloses a kind of adaptive wavelet threshold image denoising device, under described device includes Row module:

Wavelet decomposition module, for picture signal is carried out wavelet decomposition, selects suitable small echo and sets decomposition High-level N, (i, j) at the wavelet coefficient of each layer to calculate noisy picture signal s；

Wavelet coefficient module, for the threshold process of wavelet coefficient, it is suitable to determine in each layer all directions of wavelet decomposition Threshold value, to each layer detail wavelet coefficients use suitable threshold function table process, to retain the little of picture signal as far as possible Wave system number is principle, and the wavelet coefficient making noise is zero；

Image reconstruction module, is used for reconstructing picture signal, utilize n-th layer approximate part wavelet coefficient and from the 1st layer to N-th layer treated each detail wavelet coefficients reconstruct picture signal.

Wherein, described wavelet decomposition module includes:

Wavelet coefficient resolving cell, for noisy picture signal s, (i j) carries out 3 layers of wavelet transformation, produces low after decomposition Frequently coefficient and direction are level, vertical, the high frequency coefficient at diagonal angle respectively；

Wavelet coefficient averaging unit, for according to above layers wavelet coefficient, it is considered to its signal correlation, to each coefficient The coefficient of adjacent area is averaging:

S_i,j,k=S/R²

$S=\underset{m=-\frac{R-1}{2}}{\overset{m=\frac{R-1}{2}}{\&Sigma;}}\underset{n=-\frac{R-1}{2}}{\overset{n=\frac{R-1}{2}}{\&Sigma;}}{\left(W_{i,j+m,k+n}\right)}^2,i=1:N$

Wherein, R represents the field averaged, W_i,j,kBeing the wavelet coefficient of i-th layer, j, k represent pixel in image Position, S_i,j,kFor the wavelet coefficient using absolute average to retrieve after estimating.

Wherein, described wavelet coefficient module includes:

Threshold values determines unit, is used for using Adaptive Thresholding to determine described threshold value, and the threshold value being adapted to each layer is:

λ_i=(N+1-i) 2 β σ²Log (R) i=1:N, β ∈ (0,1)

Wherein, σ is noise variance, is estimated as follows shown in formula to it:

σ=median (| W_HH1|)/0.6745

W_HH1Represent ground floor HH sub-band coefficients；

Threshold values construction of function unit, for according to the wavelet coefficient obtained, being constructed as follows wavelet coefficient in each layer all directions Adaptive thresholding value function:

${\hat{W}}_{i,j,k}=\left\{\begin{array}{cc}{W}_{i,j,k}(1-\frac{{\mathrm{\&alpha;\&lambda;}}_i}{(N+1-i)S_{i,j,k}})& S_{i,j,k}\&GreaterEqual;{\mathrm{\&lambda;}}_i\\ 0& S_{i,j,k}<{\mathrm{\&lambda;}}_i\end{array}\right.,\mathrm{\&alpha;}\&Element;(0,1)$

Wherein α is attenuation quotient, and i is the wavelet decomposition number of plies, λ_iIt is the threshold value of i-th layer, W_i,j,kIt is that i-th layer of decomposition obtains Original wavelet coefficients,For the estimated value of purified signal wavelet coefficient, j, k represent the position of pixel in image.

It should be noted that in said apparatus embodiment, included modules and unit are according to function logic Carry out dividing, but be not limited to above-mentioned division, as long as being capable of corresponding function；It addition, each module and list The specific name of unit, also only to facilitate mutually distinguish, is not limited to protection scope of the present invention.

Above-described embodiment is the present invention preferably embodiment, but embodiments of the present invention are not by above-described embodiment Limit, the change made under other any spirit without departing from the present invention and principle, modify, substitute, combine, simplify, All should be the substitute mode of equivalence, within being included in protection scope of the present invention.

Claims (10)

1. an adaptive wavelet threshold Image denoising algorithm, it is characterised in that described algorithm comprises the following steps:

S1, picture signal being carried out wavelet decomposition step, select suitable small echo and set highest level N of decomposition, calculating contains (i, j) at the wavelet coefficient of each layer for picture signal s of making an uproar；

S2, the thresholding step of wavelet coefficient, determine suitable threshold value in each layer all directions of wavelet decomposition, thin to each layer Joint wavelet coefficient use suitable threshold function table process, with retain picture signal as far as possible wavelet coefficient as principle, make The wavelet coefficient of noise is zero；

S3, reconstruct picture signal step, utilize the approximate part wavelet coefficient of n-th layer and treated to n-th layer from the 1st layer Each detail wavelet coefficients reconstruct picture signal.

A kind of adaptive wavelet threshold Image denoising algorithm the most according to claim 1, it is characterised in that described S1, right Picture signal carry out wavelet decomposition step particularly as follows:

S11, to noisy picture signal s, (i, j) carries out 3 layers of wavelet transformation, and producing low frequency coefficient and direction after decomposition is water respectively Flat, vertical, the high frequency coefficient at diagonal angle；

S12, according to above layers wavelet coefficient, it is considered to its signal correlation, the coefficient of each coefficient adjacent area is averaging:

S_i,j,k=S/R²

$S=\underset{m=-\frac{R-1}{2}}{\overset{m=\frac{R-1}{2}}{\&Sigma;}}\underset{n=-\frac{R-1}{2}}{\overset{n=\frac{R-1}{2}}{\&Sigma;}}{\left(W_{i,j+m,k+n}\right)}^2,i=1:N$

Wherein, R represents the field averaged, W_i,j,kBeing the wavelet coefficient of i-th layer, j, k represent the position of pixel in image, S_i,j,kFor the wavelet coefficient using absolute average to retrieve after estimating.

3. according to a kind of adaptive wavelet threshold Image denoising algorithm according to claim 1, it is characterised in that described valve The choosing method of value is specific as follows:

Using Adaptive Thresholding to determine described threshold value, the threshold value being adapted to each layer is:

λ_i=(N+1-i) 2 β σ²Log (R) i=1:N, β ∈ (0,1)

Wherein, σ is noise variance, is estimated as follows shown in formula to it:

σ=median (| W_HH1|)/0.6745

W_HH1Represent ground floor HH sub-band coefficients.

4. according to a kind of adaptive wavelet threshold Image denoising algorithm according to claim 3, it is characterised in that described threshold The building method of value function is specific as follows:

According to the wavelet coefficient obtained, it is constructed as follows the adaptive thresholding value function of wavelet coefficient in each layer all directions:

${\hat{W}}_{i,j,k}=\left\{\begin{array}{cc}{W}_{i,j,k}(1-\frac{{\mathrm{\&alpha;\&lambda;}}_i}{(N+1-i)S_{i,j,k}})& S_{i,j,k}\&GreaterEqual;{\mathrm{\&lambda;}}_i\\ 0& S_{i,j,k}<{\mathrm{\&lambda;}}_i\end{array}\right.,\mathrm{\&alpha;}\&Element;(0,1)$

Wherein α is attenuation quotient, and i is the wavelet decomposition number of plies, λ_iIt is the threshold value of i-th layer, W_i,j,kBe i-th layer of decomposition obtain original Wavelet coefficient,For the estimated value of purified signal wavelet coefficient, j, k represent the position of pixel in image.

5. according to a kind of adaptive wavelet threshold Image denoising algorithm according to claim 4, it is characterised in that described S3, reconstruct picture signal step particularly as follows:

Described purified signal wavelet coefficient after utilizing threshold function table to processCarry out inverse wavelet transform and reconstruct, being restored Denoising picture signal.

6. according to a kind of adaptive wavelet threshold Image denoising algorithm according to claim 2, it is characterised in that described in ask The field R=5 of meansigma methods.

7. according to a kind of adaptive wavelet threshold Image denoising algorithm according to claim 4, it is characterised in that described tune Joint factor-beta=0.3, described attenuation quotient α=0.1.

8. an adaptive wavelet threshold image denoising device, it is characterised in that described device includes following modules:

Wavelet decomposition module, for picture signal is carried out wavelet decomposition, selects suitable small echo and sets the top of decomposition Secondary N, (i, j) at the wavelet coefficient of each layer to calculate noisy picture signal s；

Wavelet coefficient module, for the threshold process of wavelet coefficient, determines suitable threshold in each layer all directions of wavelet decomposition Value, uses suitable threshold function table to process, to retain the wavelet systems of picture signal as far as possible each layer detail wavelet coefficients Number is principle, and the wavelet coefficient making noise is zero；

Image reconstruction module, is used for reconstructing picture signal, utilizes the approximate part wavelet coefficient of n-th layer and from the 1st layer to n-th layer Treated each detail wavelet coefficients reconstruct picture signal.

A kind of adaptive wavelet threshold image denoising device the most according to claim 8, it is characterised in that described little wavelength-division Solution module includes:

Wavelet coefficient resolving cell, for noisy picture signal s, (i, j) carries out 3 layers of wavelet transformation, produces low frequency system after decomposition Number and direction are level, vertical, the high frequency coefficient at diagonal angle respectively；

Wavelet coefficient averaging unit is for according to above layers wavelet coefficient, it is considered to its signal correlation, adjacent to each coefficient The coefficient in region is averaging:

S_i,j,k=S/R²

$S=\underset{m=-\frac{R-1}{2}}{\overset{m=\frac{R-1}{2}}{\&Sigma;}}\underset{n=-\frac{R-1}{2}}{\overset{n=\frac{R-1}{2}}{\&Sigma;}}{\left(W_{i,j+m,k+n}\right)}^2,i=1:N$

Wherein, R represents the field averaged, W_i,j,kBeing the wavelet coefficient of i-th layer, j, k represent the position of pixel in image, S_i,j,kFor the wavelet coefficient using absolute average to retrieve after estimating.

A kind of adaptive wavelet threshold image denoising device the most according to claim 8, it is characterised in that described small echo Coefficient module includes:

Threshold values determines unit, is used for using Adaptive Thresholding to determine described threshold value, and the threshold value being adapted to each layer is:

λ_i=(N+1-i) 2 β σ²Log (R) i=1:N, β ∈ (0,1)

Wherein, σ is noise variance, is estimated as follows shown in formula to it:

σ=median (| W_HH1|)/0.6745

W_HH1Represent ground floor HH sub-band coefficients；

Threshold values construction of function unit, for according to the wavelet coefficient that obtains, be constructed as follows wavelet coefficient in each layer all directions from Adaptation threshold function table:

${\hat{W}}_{i,j,k}=\left\{\begin{array}{cc}{W}_{i,j,k}(1-\frac{{\mathrm{\&alpha;\&lambda;}}_i}{(N+1-i)S_{i,j,k}})& S_{i,j,k}\&GreaterEqual;{\mathrm{\&lambda;}}_i\\ 0& S_{i,j,k}<{\mathrm{\&lambda;}}_i\end{array}\right.,\mathrm{\&alpha;}\&Element;(0,1)$

Wherein α is attenuation quotient, and i is the wavelet decomposition number of plies, λ_iIt is the threshold value of i-th layer, W_i,j,kBe i-th layer of decomposition obtain original Wavelet coefficient,For the estimated value of purified signal wavelet coefficient, j, k represent the position of pixel in image.