CN103810677B - Enhancement Methods about Satellite Images based on super empirical mode decomposition - Google Patents

Abstract

The present invention relates to the Enhancement Methods about Satellite Images based on super empirical mode decomposition, remote sensing images to be reinforced are carried out the multiple dimensioned multi-direction decomposition to low frequency of the image high frequency, it is thus achieved that in the high fdrequency component of yardsticks at different levels, accumulate mode component and low frequency residual components；Adopt and accumulate mode component in the non-linear enhancing function high frequency to yardsticks at different levels and low frequency residual components carries out enhancement process；Mode component will be accumulate and low frequency residual components reversely reconstructs and obtains strengthening image in the high frequency of enhanced yardsticks at different levels；This method can the detailed information of extraction source image well, it is achieved remote sensing images strengthen based on brand-new multi-resolution decomposition structure, have the adaptivity that complete data drives, have higher detailed information acquisition capability；Use non-linear enhancing function to accumulateing mode component in high frequency and low frequency residual components carries out enhancement process so that after enhancing, the quality of image is improved, remote sensing images are strengthened to the application in field is significant and practical value.

Description

Enhancement Methods about Satellite Images based on super empirical mode decomposition

Technical field

The invention belongs to image enhancement technique field, relate to the Enhancement Methods about Satellite Images based on super empirical mode decomposition.

Background technology

Image enhaucament be intended to improve the visual effect of image, application scenario for given image, on purpose emphasize entirety or the local characteristics of image, original unsharp image is apparent from or is emphasized some feature interested, difference between different objects feature in expanded view picture, suppress uninterested feature, so as to improve picture quality, abundant information amount, strengthen image interpretation and recognition effect, meet the needs of some special analysis, be suitable for the demand of the understanding of computer, analysis and subsequent treatment.

Remote sensing images enhancing is an important application in image enhaucament.Remote sensing images are due to spatial noise, the impact of equipment self causes obtained remote sensing images resolution and contrast high, when analyzing remote sensing images, in order to make analyst easily can identify picture material definitely, according to analysis purpose, view data must be processed, it is therefore an objective to improve the interpretability of image.

At present, based on the blending algorithm of multiresolution, multi-resolution decomposition, it is used widely in image co-registration.Various countries' researcher proposes the processing method of the multiple small echos such as Wavelet conversion, Ridgelet conversion, Curvelet conversion, contourlet transformation and Bandelet conversion and extra small wave conversion, it is simply that the important research achievement of this respect.But no matter it is based on which kind of small echo, in image co-registration, all there is a problem in that enhanced image distortion can occur at local location.Therefore, engineering circles and mathematical region never stopped exploring better decomposition algorithm.

1999, astable nonlinear properties can be done adaptive decomposition by frequency by the NordenE.Huang teaching inventive of NASA Empirical Mode Decomposition Algorithm (EmpiricalModeDecomposition, EMD).Bidimensional Empirical Mode Decomposition is the popularization on two dimensional surface of the one-dimensional EMD decomposition algorithm, can be used for the analysis of view data and process, by by adaptive for the original image subimage being decomposed into effective quantity, can decomposing out by accumulateing mode component in image arrowband, local detailed information from high frequency to low frequency, residual components represents the trend of image.Accumulate mode component in decomposition out and there is the texture information of present image.But conventional two-dimensional empirical mode decomposition is defective: decomposing to accumulate in mode component image in obtaining has skin dark stain.Therefore the application in image processing field of the conventional two-dimensional empirical mode decomposition has been had a strong impact on.The window empirical mode decomposition WEMD occurred later solves the defect of conventional two-dimensional empirical mode decomposition preferably, remain again conventional two-dimensional empirical mode decomposition adaptive decomposition characteristic, and be applied in image enhaucament, but only in medical image enhancement is applied, effect is better, in remote sensing images strengthen, details reinforced effects is poor.

To sum up, current existing enhancement techniques is applied to remote sensing images and still has several drawbacks: the enhancing image based on small echo, extra small ripple there will be local distortion, Conventional wisdom Mode Decomposition method decompose obtain in accumulate mode component image skin dark stain and all enhancing had a great impact, the image that strengthens of window empirical mode decomposition there will be the problem that a small amount of information dropout, details reinforced effects are poor.

Summary of the invention

It is an object of the invention to provide a kind of Enhancement Methods about Satellite Images based on super empirical mode decomposition, both occurred without small echo, there is distortion in the enhanced image of extra small ripple blending algorithm, occur without again the window empirical mode decomposition problem that effect is poor in remote sensing images strengthen, to strengthen the detailed information of image well.

For achieving the above object, the step based on the Enhancement Methods about Satellite Images of super empirical mode decomposition of the present invention is as follows:

(1) remote sensing images to be reinforced are carried out the multiple dimensioned multi-direction decomposition to low frequency of the image high frequency, it is thus achieved that in the high fdrequency component of yardsticks at different levels, accumulate mode component and low frequency residual components；

(2) adopt and accumulate mode component in the non-linear enhancing function high frequency to yardsticks at different levels and low frequency residual components carries out enhancement process；

(3) mode component will be accumulate in the high frequency of enhanced yardsticks at different levels and low frequency residual components reversely reconstructs and obtains strengthening image.

Described step (1) is to adopt super empirical mode decomposition that remote sensing images are decomposed, and concrete catabolic process is as follows:

(11) remote sensing images to be reinforced are carried out Radon and converts Raf (b, θ), b, θ respectively position and rotation angle parameter；

(12) use one-dimensional empirical mode decomposition that the radial section that each θ angle is corresponding is decomposed, obtain accumulateing mode component Rimf under Radon territory_jWith residual components Rr；

(13) to Rimf_jWith Rr carry out Radon inverse transformation obtain final in accumulate mode component imf_jWith residual components r.

In described step (12), the detailed process of one-dimensional empirical mode decomposition is as follows:

(121) definition initializing variable r_j-1, and make r_j-1=I, I are the radial section data that in step (12), each θ angle is corresponding, j=1, and j represents that decomposition accumulates mode component quantity in obtaining；

(122) mode component quantity is accumulate according in decomposing,

(a) definition intermediate variable h_i-1, make h_i-1=r_j-1, i=1, i represents screening number of times；

B () uses partial estimation to calculate h respectively_i-1Coenvelope line u_i-1With lower envelope line l_i-1；

C () calculates average envelope line m_i-1, m_i-1=(u_i-1+l_i-1)/2；

D () updates h_i=h_i-1-m_i-1, i=i+1；

E () repeats (b) to (d) until i < AI, AI is the screening number of times specified, then, imf_j=h_i, j=j+1；

(123) r is updated_j=r_j-1–imf_j；

(124) step (122), (123) are repeated, until r_jMaximum and minimum number and less than 3.

The concrete calculating process of (b) in described step (122) is as follows:

(b1) h is extracted_i-1All Local Extremum, calculate the distance of all adjacent Local Extremum, obtain the p that is sized to of maximum local window, and set initial window as w=3；

(b2) if local maximum is counted out and counted out equal to local minimum in the window w of point centered by current data, then the current point using the local maximum in window w as local maximum coenvelope line umax, and step (b4) is gone to；Otherwise go to step (b3)；

(b3) w window value increases by 2, if w < p, goes to step (b2)；If w > p, with the current point as local maximum coenvelope umax of the local maximum in current window w；And go to step (b4)；

(b4) go to next data point, and make w=3, go to step (b2)；Travel through all data points, obtain local maximum coenvelope line umax；

(b5) if in the window w of point, local maximum is counted out and counted out equal to local minimum centered by umax current data, then using the average in window w as the current point of coenvelope u, and go to step (b7)；

(b6) w window value increases by 2, if w < p, goes to step (b5), otherwise using the average in window w as the current point of coenvelope u；

(b7) go to umax next one data point, and make w=3, go to step (b5)；Travel through all data points, obtain coenvelope u；

(b8) with reference to step (b2) to (b7), obtaining lower envelope l, in calculating process, the word " maximum " occurred become " minimum " in step, parameter " umax " becomes " lmin ", and parameter " u " becomes " l ", other parameter constant.

In described step (b1), the Local modulus maxima of image is the point that gray value is all higher than 2, surrounding 3 region neighbor pixel gray value, and the minimum point of image is the point that gray value is all lower than 2 neighbor pixel gray values of surrounding.

In described step (b1), maximum local window p is that traversal extreme value point set s finds out the extreme point that next-door neighbour's extreme point is closest, and calculates the distance P of point-to-point transmission by all local maximums and minimum point Maximal plus algebra s_i(i=1,2,3 ...), with P_iMaximum one rounds as maximum local window p intermediate value；When p is even number, perform p=p+1 operation.

Described step (2) step is:

(21) intensification factor of each coefficient in imf or r is obtained according to following formula,

$\left\{\begin{array}{c}y(x,\mathrm{\&sigma;})=1;\left|x\right|<\mathrm{c\&sigma;}\\ y(x,\mathrm{\&sigma;})=\frac{\left|x\right|-\mathrm{c\&sigma;}}{\mathrm{c\&sigma;}}{\left(\frac{z}{\mathrm{c\&sigma;}}\right)}^q+\frac{2\mathrm{c\&sigma;}-\left|x\right|}{\mathrm{c\&sigma;}};\left|x\right|<2\mathrm{c\&sigma;}\\ y(x,\mathrm{\&sigma;})={\left(\frac{z}{\left|x\right|}\right)}^q;2\mathrm{c\&sigma;}<\left|x\right|<z\\ y(x,\mathrm{\&sigma;})=1;\left|x\right|>z\end{array}\right.$

Y (x, σ) is intensification factor, and x is the coefficient in imf or r, and σ is the noise variance of current imf or r；Q determines the degree of crook of power function curve；Q is between 0 and 1；σ is the noise variance of imf, incorporating parametric c and z, is divided into different grades to strengthen the coefficient in imf or r, c between 1 and 2, z=gM_imf, M_imfFor the maximum absolute value value in imf or r, g is between 0 and 1；

(22) following formula is utilized to obtain imf_jImf enhanced with r_j' and r ',

${\mathrm{imf}}_j{\left(s\right)}^{\&prime;}={\mathrm{imf}}_j\left(s\right)\&times;y({\mathrm{imf}}_j\left(s\right),{\mathrm{\&sigma;}}_{{\mathrm{imf}}_j}),$

r(s)′=r(s)×y(r(s),σ_r),

S is the position parameter of the coefficient in imf and r matrix,With y (r (s), σ_r) for imf_jIntensification factor with r each coefficient calculated.

The Enhancement Methods about Satellite Images based on super empirical mode decomposition of the present invention, the super empirical mode decomposition source images to collecting is adopted to carry out multiple dimensioned multi-direction decomposition, the advantage that catabolic process inherits Conventional wisdom Mode Decomposition: completely by data-driven, ask at different levels in accumulate the similar High frequency filter process of mode component, with small echo, extra small ripple is compared, the detailed information of more horn of plenty can be obtained, the quality strengthening image is improved, compared with window empirical mode decomposition, solve window empirical mode decomposition and strengthen the problem that a small amount of information dropout occurs in algorithm.Obtain the HFS of yardstick at different levels, according to non-linear enhancing function to accumulateing mode component at different levels and residual components carries out enhancement process, the enhancing detailed information of maximizing, reduce the impact of non-detailed information.This method had both occurred without small echo, distortion occurs in the enhanced image of extra small ripple blending algorithm, occur without again window empirical mode decomposition in remote sensing images strengthen, occur that details enhancing is poor, can the detailed information of extraction source image well, realize remote sensing images to strengthen based on brand-new multi-resolution decomposition structure, there is the adaptivity that complete data drives, there is higher detailed information acquisition capability；Use non-linear enhancing function to accumulateing mode component in high frequency and low frequency residual components carries out enhancement process so that after enhancing, the quality of image is improved, remote sensing images are strengthened to the application in field is significant and practical value.

Accompanying drawing explanation

Fig. 1 is the inventive method flow chart；

Fig. 2 is that Conventional wisdom Mode Decomposition and super empirical mode decomposition are to Lena image two-stage decomposition result comparison diagram；

Fig. 3 is the enhancing comparison diagram of the inventive method embodiment reinforced effects and small echo, extra small ripple and window empirical mode decomposition；

Mode component and residual components is accumulate in the 1st, 2 grades that in Fig. 2, the first behavior Conventional wisdom Mode Decomposition obtains, accumulate mode component and residual components in the 1st, 2 grades that second behavior window empirical mode decomposition obtains, in the 1st, 2 grades that the third line is that super empirical mode decomposition obtains, accumulate mode component and residual components.

In Fig. 3, the upper left corner is remote sensing images to be reinforced, and the upper right corner is curvelet reinforced effects figure, and the lower left corner is window empirical mode decomposition reinforced effects figure, and the lower right corner is the reinforced effects figure of the inventive method embodiment.

Detailed description of the invention

In order to be more fully understood that technical scheme, below in conjunction with accompanying drawing, the embodiment of the Enhancement Methods about Satellite Images based on super empirical mode decomposition is described in detail.

As shown in Figure 1, first source images is carried out super empirical mode decomposition based on the Enhancement Methods about Satellite Images of super empirical mode decomposition by this, obtain in the n level of source images, accumulateing mode component and a residual components, to accumulateing mode component at different levels and residual components adopts non-linear enhancing rule to carry out enhancement process；Last inverse transformation obtains and strengthens image.

Originally specifically comprising the following steps that based on the Enhancement Methods about Satellite Images of super empirical mode decomposition

1. use super empirical mode decomposition Ei algorithm to be decomposed by source images I, obtain intrinsic mode functions component imf_jWith residual components r, j=1,2,3 ..., n, n decomposes the progression of imf obtained；

The processing procedure of super empirical mode decomposition is as follows:

Step 1: the Radon calculating image x converts Raf (b, θ)；(b, θ) respectively position and rotation angle parameter；

Step 2: use partial estimation empirical mode decomposition that the section that each θ angle is corresponding is decomposed, obtain accumulateing mode component Rimf under Radon territory_jWith residual components Rr；

Step 3: to Rimf_jWith Rr carry out Radon inverse transformation obtain final in accumulate mode component imf_jWith residual components r.

Partial estimation empirical mode processing procedure comprises the steps:

Step 1: definition initializing variable r_j-1, and make r_j-1=I, I are the radial section data that in step (12), each θ angle is corresponding, j=1, and j represents that decomposition accumulates mode component quantity in obtaining；

Step 2: accumulate mode component quantity according in decomposing,

(a) definition intermediate variable h_i-1, make h_i-1=r_j-1, i=1, i represents screening number of times；

B () uses partial estimation to calculate h respectively_i-1Coenvelope line u_i-1With lower envelope line l_i-1；

C () calculates average envelope line m_i-1, m_i-1=(u_i-1+l_i-1)/2；

D () updates h_i=h_i-1-m_i-1, i=i+1；

E () repeats (b) to (d) until i < AI, AI is the screening number of times specified, then, imf_j=h_i, j=j+1；

Step 3: update r_j=r_j-1–imf_j；

Step 4: repeat step (2) to (3) until r_jMaximum and minimum number and less than 3.

Partial estimation calculates h respectively_i-1Coenvelope u_i-1With lower envelope l_i-1Step as follows:

Step 1: extract h_i-1All Local Extremum, calculate the distance of all adjacent Local Extremum, obtain the p that is sized to of maximum local window, initial window is w=3；

Local modulus maxima is the point that gray value is all higher than 2, surrounding 3 region neighbor pixel gray value, and the minimum point of image is the point that gray value is all lower than 2 neighbor pixel gray values of surrounding.

Maximum local window p is that traversal extreme value point set s finds out the extreme point that next-door neighbour's extreme point is closest, and calculates the distance P of point-to-point transmission by all local maximums and minimum point Maximal plus algebra s_i(i=1,2,3 ...), with P_iMaximum one rounds as maximum local window p intermediate value；When p is even number, perform p=p+1 operation.

Step 2: if local maximum is counted out and counted out equal to local minimum in the window w of point centered by current data, then the current point using the local maximum in window w as local maximum coenvelope line umax, and go to step (4)；Otherwise go to step (3)；

Step 3:w window value increases by 2, if w < p, goes to step (b2)；If w > p, with the current point as local maximum coenvelope umax of the local maximum in current window w；And go to step (4)；

Step 4: go to next data point, and make w=3, go to step (2).Travel through all data points, obtain local maximum coenvelope umax；

Step 5: if local maximum is counted out and counted out equal to local minimum in the window w of point centered by umax current data, then using the average in window w as the current point of coenvelope u, and go to step (7)；

Step 6:w window value increases by 2, if w < p, goes to step (b5), otherwise using the average in window w as the current point of coenvelope u；

Step 7: go to umax next one data point, and make w=3, goes to step (5).Travel through all data points, obtain coenvelope u；

Step 8: similar repetition step (b2) arrives (b7), obtains lower envelope l, in the process calculated, the word " maximum " occurred in step is become " minimum ", parameter " umax " becomes " lmin ", and parameter " u " becomes " l ", other parameter constant.

Super empirical mode decomposition achieves the image high frequency multi-scale self-adaptive catabolic process to low frequency.First, accumulateing mode component in the 1st grade of resolution process is the highest frequency component contained by image, and source data deducts accumulates mode component and obtain the 1st grade of residual components in the 1st grade；1st grade of residual components is decomposed again, obtains in the 2nd grade, accumulateing mode component and the 2nd grade of residual components；By that analogy, obtain in n level, accumulateing mode component and n-th grade of residual components.Lena image is done 2 grades of decomposition, obtain in 2 grades, accumulateing mode component and a residual components, as shown in Figure 2, mode component and residual components is accumulate in the 1st, 2 grades that in figure, the first behavior Conventional wisdom Mode Decomposition obtains, accumulate mode component and residual components in the 1st, 2 grades that second behavior window empirical mode decomposition obtains, in the 1st, 2 grades that the third line is that super empirical mode decomposition obtains, accumulate mode component and residual components.Visible multidirectional empirical mode decomposition also can solve the problem that skin dark stain occurs in Conventional wisdom Mode Decomposition well.Multidirectional empirical mode decomposition inherits the advantage of Conventional wisdom Mode Decomposition, and its also base produces according to signal adaptive so that it has good time-frequency locality.

2. by the imf of image to be reinforced_jCarry out enhancement process with r according to non-linear enhancing rule, produce to strengthen the imf of image_j' and r '.

Non-linear enhanced processes comprises the steps:

Step 1: obtain the intensification factor of each coefficient in imf or r according to following formula,

$\left\{\begin{array}{c}y(x,\mathrm{\&sigma;})=1;\left|x\right|<\mathrm{c\&sigma;}\\ y(x,\mathrm{\&sigma;})=\frac{\left|x\right|-\mathrm{c\&sigma;}}{\mathrm{c\&sigma;}}{\left(\frac{z}{\mathrm{c\&sigma;}}\right)}^q+\frac{2\mathrm{c\&sigma;}-\left|x\right|}{\mathrm{c\&sigma;}};\left|x\right|<2\mathrm{c\&sigma;}\\ y(x,\mathrm{\&sigma;})={\left(\frac{z}{\left|x\right|}\right)}^q;2\mathrm{c\&sigma;}<\left|x\right|<z\\ y(x,\mathrm{\&sigma;})=1;\left|x\right|>z\end{array}\right.,$

Y (x, σ) is intensification factor, and x is the coefficient in imf or r, and σ is the noise variance of current imf or r；Q determines the degree of crook of power function curve；Q is between 0 and 1；σ is the noise variance of imf, incorporating parametric c and z, is divided into different grades to carry out appropriate enhancing the coefficient in imf or r, c between 1 and 2, z=gM_imf, M_imfFor the maximum absolute value value in imf or r, g is between 0 and 1.From above formula, if the absolute value of x is less than c σ, it is believed that this coefficient is affected by noise relatively big, it is not strengthened, the impact of reducing noise；If greater than z, it is believed that the detailed information that this coefficient represents is more satisfactory, it is not necessary to strengthen, reduce owing to excessively strengthening the details distortion caused.

Step 2: utilize following formula to obtain imf_jImf enhanced with r_j' and r ',

${\mathrm{imf}}_j{\left(s\right)}^{\&prime;}={\mathrm{imf}}_j\left(s\right)\&times;y({\mathrm{imf}}_j\left(s\right),{\mathrm{\&sigma;}}_{{\mathrm{imf}}_j}),$

r(s)′=r(s)×y(r(s),σ_r),

S is the position parameter of the coefficient in imf and r matrix,With y (r (s), σ_r) for imf_jIntensification factor with r each coefficient calculated.

3. reversely reconstruct the enhancing image I ' obtained:

In order to verify effectiveness of the invention, remote sensing images are used to carry out enhancement process.In Fig. 3, the upper left corner is remote sensing images to be reinforced, and the upper right corner is curvelet reinforced effects figure, and the lower left corner is window empirical mode decomposition reinforced effects figure, and the lower right corner is this reinforced effects figure based on the Enhancement Methods about Satellite Images of super empirical mode decomposition.Contrasting visible, there is local distortion in the image of super Wavelet Fusion, it is impossible to optimum expression image enhaucament detail textures information, and window empirical mode decomposition contrast in strengthening image is not strong.This is clear, undistorted based on the enhancing image detail of the Enhancement Methods about Satellite Images of super empirical mode decomposition, and after enhancing, information preserves complete, optimally enhances image detail texture information.For objective evaluation syncretizing effect, following evaluation index is selected to carry out strengthening the objective evaluation of result here.

Use details variance (DV), background variance (BV) and DV/BV value to analyze and strengthen image, " DV " is the local variance average of image all detail areas pixel, " BV " is the local variance average of all background area pixels, and they can strengthen ability and noise sensitivity as evaluating the details strengthening algorithm.DV is more big, and the visible details of image is more abundant；BV is more little, and noise is more little, and image background is more uniform.

The computational methods of DV, BV are as follows:

(1) sliding window utilizing 5 × 5 calculates artwork f (x, y) the local variance lv (x of each pixel, y), then utilize maximum variance between clusters to determine details/background area, pixel is belonged to detail areas (being labeled as 1), pixel belongs to background area (being labeled as 0), obtain background/detail pictures n (x, y)；

(2) by n, (x, (x, y) is accumulated in together, and divided by detail areas pixel count, just obtains strengthening the DV value of image to be labeled as the local variance lv ' of the pixel of 1 in y).By n, (x, (x y) is accumulated in together and divided by background area pixel count, just obtains strengthening the BV value of image to be labeled as the local variance lv ' of the pixel of 0 in y).

Table 1 strengthens image evaluation parameter

From the data in table 1, it can be seen that the DV based on the Enhancement Methods about Satellite Images of super empirical mode decomposition is maximum, and BV second is little, it is seen that DV has more detailed information, and the noise introduced is less.The DV that curvelet strengthens algorithm is second largest, and BV is second largest, causes local detail comparatively fuzzy, and window empirical mode decomposition strengthens algorithm in remote sensing images strengthen, and overall effect might as well original image.

In sum, the image enhaucament resultant effect based on the Enhancement Methods about Satellite Images of super empirical mode decomposition is best.

It should be noted last that, above example is only in order to illustrate technical scheme and unrestricted.Although the present invention has been described in detail by embodiment, it will be understood by those within the art that, technical scheme being modified or equivalent replacement, without departure from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of scope of the presently claimed invention.

Claims (6)

1. based on the Enhancement Methods about Satellite Images of super empirical mode decomposition, it is characterised in that the step of the method is as follows:

(1) remote sensing images to be reinforced are carried out the multiple dimensioned multi-direction decomposition to low frequency of the image high frequency, it is thus achieved that in the high fdrequency component of yardsticks at different levels, accumulate mode component and low frequency residual components；

(2) adopt and accumulate mode component in the non-linear enhancing function high frequency to yardsticks at different levels and low frequency residual components carries out enhancement process；

(3) mode component will be accumulate in the high frequency of enhanced yardsticks at different levels and low frequency residual components reversely reconstructs and obtains strengthening image；

Described step (1) is to adopt super empirical mode decomposition that remote sensing images are decomposed, and concrete catabolic process is as follows:

(11) remote sensing images to be reinforced are carried out Radon and converts Raf (b, θ), b, θ respectively position and rotation angle parameter；

(12) use one-dimensional empirical mode decomposition that the radial section that each θ angle is corresponding is decomposed, obtain accumulateing mode component Rimf under Radon territory_jWith residual components Rr；

(13) to Rimf_jWith Rr carry out Radon inverse transformation obtain final in accumulate mode component imf_jWith residual components r.

2. the Enhancement Methods about Satellite Images based on super empirical mode decomposition according to claim 1, it is characterised in that in described step (12), the detailed process of one-dimensional empirical mode decomposition is as follows:

(121) definition initializing variable r_j-1, and make r_j-1=I, I are the radial section data that in step (12), each θ angle is corresponding, j=1, and j represents that decomposition accumulates mode component quantity in obtaining；

(122) mode component quantity is accumulate according in decomposing,

(a) definition intermediate variable h_i-1, make h_i-1=r_j-1, i=1, i represents screening number of times；

B () uses partial estimation to calculate h respectively_i-1Coenvelope line u_i-1With lower envelope line l_i-1；

C () calculates average envelope line m_i-1, m_i-1=(u_i-1+l_i-1)/2；

D () updates h_i=h_i-1-m_i-1, i=i+1；

E () repeats (b) to (d) until i < AI, AI is the screening number of times specified, then, imf_j=h_i, j=j+1；

(123) r is updated_j=r_j-1–imf_j；

(124) step (122), (123) are repeated, until r_jMaximum and minimum number and less than 3.

3. the Enhancement Methods about Satellite Images based on super empirical mode decomposition according to claim 2, it is characterised in that the concrete calculating process of (b) in described step (122) is as follows:

(b1) h is extracted_i-1All Local Extremum, calculate the distance of all adjacent Local Extremum, obtain the p that is sized to of maximum local window, and set initial window as w=3；

(b2) if local maximum is counted out and counted out equal to local minimum in the window w of point centered by current data, then the current point using the local maximum in window w as local maximum coenvelope line umax, and step (b4) is gone to；Otherwise go to step (b3)；

(b3) w window value increases by 2, if w < p, goes to step (b2)；If w > p, with the current point as local maximum coenvelope line umax of the local maximum in current window w；And go to step (b4)；

(b4) go to next data point, and make w=3, go to step (b2)；Travel through all data points, obtain local maximum coenvelope line umax；

(b5) if in the window w of point, local maximum is counted out and counted out equal to local minimum centered by umax current data, then using the average in window w as the current point of coenvelope u, and go to step (b7)；

(b6) w window value increases by 2, if w < p, goes to step (b5), otherwise using the average in window w as the current point of coenvelope u；

(b7) go to umax next one data point, and make w=3, go to step (b5)；Travel through all data points, obtain coenvelope u；

(b8) with reference to step (b2) to (b7), lower envelope l is obtained, in calculating process, the word " maximum " occurred in step is become " minimum ", parameter " umax " becomes " lmin ", and parameter " u " becomes " l ", other parameter constant.

4. the Enhancement Methods about Satellite Images based on super empirical mode decomposition according to claim 3, it is characterized in that: in described step (b1), the Local modulus maxima of image is the point that gray value is all higher than 2, surrounding 3 region neighbor pixel gray value, and the minimum point of image is the point that gray value is all lower than 2 neighbor pixel gray values of surrounding.

5. the Enhancement Methods about Satellite Images based on super empirical mode decomposition according to claim 3, it is characterized in that: in described step (b1), maximum local window p is by all local maximums and minimum point Maximal plus algebra s, traversal extreme value point set s finds out the extreme point that next-door neighbour's extreme point is closest, and calculates the distance P of point-to-point transmission_i(i=1,2,3 ...), with P_iMaximum one rounds as maximum local window p intermediate value；When p is even number, perform p=p+1 operation.

6. the Enhancement Methods about Satellite Images based on super empirical mode decomposition according to any one of claim 1-5, it is characterised in that described step (2) step is:

(21) intensification factor of each coefficient in imf or r is obtained according to following formula,

$\left\{\begin{array}{cc}y(x,\mathrm{\&sigma;})=1;& \left|x\right|<c\mathrm{\&sigma;}\\ y(x,\mathrm{\&sigma;})=\frac{\left|x\right|-c\mathrm{\&sigma;}}{c\mathrm{\&sigma;}}{\left(\frac{z}{c\mathrm{\&sigma;}}\right)}^q+\frac{2c\mathrm{\&sigma;}-\left|x\right|}{c\mathrm{\&sigma;}};& \left|x\right|<2c\mathrm{\&sigma;}\\ y(x,\mathrm{\&sigma;})={\left(\frac{z}{\left|x\right|}\right)}^q;& 2c\mathrm{\&sigma;}<\left|x\right|<z\\ y(x,\mathrm{\&sigma;})=1;& \left|x\right|>z\end{array}\right.$

Y (x, σ) is intensification factor, and x is the coefficient in imf or r, and σ is the noise variance of current imf or r；Q determines the degree of crook of power function curve；Q is between 0 and 1；σ is the noise variance of imf, incorporating parametric c and z, is divided into different grades to strengthen the coefficient in imf or r, c between 1 and 2, z=gM_imf, M_imfFor the maximum absolute value value in imf or r, g is between 0 and 1；

(22) following formula is utilized to obtain imf_jImf enhanced with r_j' and r',

${\mathrm{imf}}_j{\left(s\right)}^{\&prime;}={\mathrm{imf}}_j\left(s\right)\&times;y\left({\mathrm{imf}}_j\right(s),{\mathrm{\&sigma;}}_{{\mathrm{imf}}_j}),$

R (s) '=r (s) × y (r (s), σ_r),

S is the position parameter of the coefficient in imf and r matrix,With y (r (s), σ_r) for imf_jIntensification factor with r each coefficient calculated.