CN104318526B - MTF based remote sensor on-track automatic on-track parameter optimization method - Google Patents

Abstract

The invention provides an MTF based remote sensor on-track automatic on-track parameter optimization method. The MTF based remote sensor on-track automatic on-track parameter optimization method comprises the following steps of obtaining an original remote sensing image under the condition of current on-track parameter running; conducting frequency domain based denoising and MTF stretching operation on the original remote sensing image based on an MTF so as to improve image quality and judging whether the improved restored image is obviously improved or not; conducting next processing if the restored image is obviously improved; terminating on-track optimization operation if the restored image is not obviously improved, wherein image parameters of the quality-improved restored image extracted by means of an image batch processing method include edge energy, contrast ratio, definition, information entropy, detail energy, variance, image correlation, mean value, power spectrum and signal to noise ratio; inputting the image parameters extracted in the previous step into inversion models corresponding to focal length, phase shift and a forward parameter to obtain new parameters; using the new parameters to replace originally corresponding on-track parameters so as to produce new remote sensing image and performing optimization processing.

Description

The method of the in-orbit parameter of the in-orbit Automatic Optimal of the remote sensor based on MTF

Technical field

The present invention relates to instructing the in-orbit parametric technique of the in-orbit Automatic Optimal of remote sensor, more particularly, to the remote sensing figure based on MTF As quality optimization and the method based on the in-orbit parameter of the anti-satellite that spreads out of remote sensing images parameter, belong to technical field of remote sensing image processing.

Background technology

Due to being affected by factors such as sensor performance decline, atmospheric perturbations in remote sensing images imaging process, cause image matter Amount is degenerated.Image restoration is to suppress image degradation to carry out Image semantic classification using the knowledge about image degradation property Method.The purpose of image restoration is the image of Quality Down (degeneration) to be carried out respective handling, to improve picture quality.Due to The reason image degradation, may have many kinds, the relative motion such as between the non-linear, object of photoelectric sensor and video camera, air Disturbance etc., therefore according to different causes for Degradations, occur in that a variety of image recovery methods, such as liftering (deconvoluting), dimension Nanofiltration ripple, constrained least square filtering etc..The picture quality that how could effectively improve remote sensing images has important science Meaning and application value.

But go the method improving distant picture quality to have a lot of limitation to lead to it simultaneously based on traditional image recovery method Preferable effect can not be reached, its main cause has：

1) traditional image recovery method is the restorative procedure for single image, it not can overcome the disadvantages that sensor performance with The aging of sensor hardware, or the setting of the hardware that leads to of change with hardware performance and in-orbit parameter can not be perfect The harmful effect that brings such as coupling, that is, it can not fundamentally solve the problems, such as to lead to remote sensor shooting image Quality Down；

2) scope of application of single traditional images restored method is limited, and lead to the induction of image quality decrease because Element is diversified, does not therefore ensure that it is applied to all of image degradation situation；

3) most of image recovery method is all given that it is known that corresponding point spread function (point spread at present Function, PSF) on the basis of do, such as deconvolution method, Wiener filtering method etc., therefore this assume it is difficult to ensure that Its correctness and applicability.

It is it will be apparent that and it is not provided that an effective machine based on the limitation of traditional image recovery method System is asked with fundamentally solving the Remote Sensing Image Quality decline that the degeneration of remote sensor hardware and in-orbit parameter mismatch etc. lead to Topic.

Content of the invention

It is contemplated that overcoming the deficiencies in the prior art, provide a kind of in-orbit Automatic Optimal of remote sensor based on MTF in-orbit Parametric technique, using the Remote Sensing Image Quality optimization based on MTF and the mould based on the in-orbit parameter of the anti-satellite that spreads out of remote sensing images parameter Type, thus realize the real-time automatic majorization function of in-orbit parameter.

The above-mentioned purpose of the present invention is realized by the technical characteristic of independent claims, and dependent claims are to select else or to have The mode of profit develops the technical characteristic of independent claims.

For reaching above-mentioned purpose, the technical solution adopted in the present invention is as follows：

A kind of method of the in-orbit parameter of the in-orbit Automatic Optimal of the remote sensor based on MTF, its realization comprises the following steps：

Step 1, the original remote sensing images obtaining under currently in-orbit parameter is run；

Step 2, be based on modulation transfer function (modulation transfer function, MTF), to original remote sensing figure As carrying out the operation of the denoising based on frequency domain and MTF stretching, to improve picture quality, and whether judge the restored image after improvement Have and be obviously improved：If having clear improvement, entering next step and processing；If being not improved, terminating in-orbit optimization and grasping Make；

Step 3, by image batch processing method extract through quality improve after restored image image parameter, including：Side Edge energy, contrast, definition, comentropy, details energy, variance, image correlation, average, power spectrum, signal to noise ratio；

Anti- corresponding to step 4, the image parameter input focal length by extracting abovementioned steps, phase shift, front tropism parameter Spread out in model, obtain new parameter；

Step 5, by new parameter is replaced originally corresponding in-orbit parameter, produce new remote sensing images, and return step Rapid 2.

In further embodiment, in abovementioned steps 2, using blade method solve former be remote sensing images MTF curve, then Carry out MTF stretching, that is,：

The mtf value of the remote sensing images that abovementioned steps 1 are got is rewritten into following form：

MTF_new=(MTF_old)^t

In formula, t is regulation parameter, 0 ＜ t ＜ 2, MTF_oldRepresent the MTF curve of original remote sensing images, MTF_newRepresent through ginseng MTF curve after number t regulation.

In further embodiment, in front step 2, the denoising based on frequency domain is carried out to original remote sensing images, it realizes bag Include：

1) as Fourier transformation, spectral image is obtained to original remote sensing images；

2) remaining each pixel (x, y) in addition to isolated bright spot in spectral image is handled as follows：

T (i, j)=R (x, y)-R (x-2+i, y-2+j)

I, j=1,2,3

In formula, R (x, y) represents the gray value of pixel (x, y), and T represents current pixel point and centered on this pixel It is the matrix of differences of nine pixel gray values in the range of the nine grids of point；S is the Description Matrix of T, to the matrix T of 3*3 each The gray scale difference value of point, if it is more than 0.3 times of the corresponding central point gray value of original remote sensing images, the corresponding points of matrix S are designated as 1, otherwise it is designated as 0；If the point being recorded as 1 in S is more than 6 then it represents that the gray value significantly greater than point about of current point, it is judged as Isolated bright spot, updates the half that its gray value is former ash angle value, otherwise, gray value is constant；

3) Fourier inversion is done to the spectral image after processing, gray average is adjusted to and original remote sensing images simultaneously Phase

4) MTF recovery is carried out to the spectral image removing after isolated bright spot, just have the following mesh recovering original remote sensing images Mark frequency domain figure is as the method for F：

F=(G-N)/(MTF_new·k)

G is the degraded image observed, N is noise image, makes the spectrogram that R=G-N is after removing isolated bright spot

Picture, with season k=1, then simplified formula be：

F=R/MTF_new

5) Fourier inversion is carried out to the spectral image F after aforementioned MTF stretching, carry out image restoration, and will restore Afterwards the gray average of image be adjusted to identical with original remote sensing images.

In further embodiment, preceding method further includes following steps：Obtained through quality by image batch processing method After the image parameter of the restored image after improvement, also include the normalized to image parameter：

Processed using following method for normalizing：

Wherein, x_iFor the corresponding image parameter of n-th image sample.

In further embodiment, preceding method is further comprising the steps of：

The anti-model that spreads out corresponding to aforementioned focal length, phase shift, front tropism parameter is set：

1) the anti-model that spreads out of the parameter of focal length：

Focal length=9.003 × Image³+71.480×Image²+268.990×Image+735.662

Wherein：Image is defined as follows：

Image=0.43667230 × edge energy+0.45446772 × contrast+0.45446772 × definition+

0.43439084 × comentropy+0.45583660 × details energy；

2) the anti-model that spreads out of the parameter of phase shift

Phase shift=0.28+0.89 × 0.253^x

Wherein, x representative image overall target, it is defined as follows：

Image synthesis index=0.123* contrast+0.134* variance+0.132* definition+0.133* comentropy+ 0.126* details energy -0.128* image correlation+0.128* average+0.132* power spectrum；

3) the anti-model that spreads out of the parameter of tropism before

Y=-0.053 × a³+0.150×a²-0.193×a+0.101

Wherein, y is anti-calculated new front tropism parameter of spreading out, and a is defined as follows：

A=

0.3548 × contrast+0.3607 × comentropy+0.3613 × variance+0.3624* definition -0.3576* noise Than

+ 0.3597 × details energy -0.3344 × image correlation+0.3363 × edge energy.

From the above technical solution of the present invention shows that, the in-orbit Automatic Optimal of the remote sensor based on MTF proposed by the invention The method of in-orbit parameter, by state and the property of remote sensor on-orbit performance parameter and image quality evaluation parameter prediction remote sensor Can, it is that the adjustment of remote sensor parameter (including focal length, phase shift, front tropism) provides reference with optimizing.

Compared with prior art, the beneficial effects of the present invention is：

Based on real-time remote sensing images MTF indicatrix image restoration algorithm, known or accurately calculated MTF and decline In the case of curve, can restored image effectively, to improve picture quality.In order to suppress the impact of noise, employ novelty Frequency domain technique, that is, by from cluster MTF decline curve select a suitable curve stretch spectral image, so The scope of application can be expanded and improve recovery effect；Simultaneously using the anti-model that spreads out of the parameter set up based on specific remote sensing images, that is, When the image parameter information according to the high-quality remote sensing images restoring accurately calculate corresponding in-orbit parameter value, this allows for The automatic real-time optimization of rail parameter is possibly realized.

The additional aspect of the present invention and its advantage will be illustrated in following specific embodiments and describe, or It is verified in practice according to the following embodiment of the present invention.

Brief description

Fig. 1 is the reality of the method for the in-orbit parameter of the in-orbit Automatic Optimal of the remote sensor based on MTF of an embodiment of the present invention Existing schematic flow sheet.

Specific embodiment

In order to know more about the technology contents of the present invention, especially exemplified by specific embodiment and coordinate institute's accompanying drawings to be described as follows.

As shown in figure 1, according to the preferred embodiment of the present invention, a kind of in-orbit Automatic Optimal of the remote sensor based on MTF is in-orbit The method of parameter, using the Remote Sensing Image Quality optimization based on MTF and based on the in-orbit parameter of the anti-satellite that spreads out of remote sensing images parameter Model, thus realizing the real-time automatic majorization function of in-orbit parameter, its realization comprises the following steps：

Step 1, the remote sensing images obtaining under currently in-orbit parameter is run；

Step 2, be based on modulation transfer function (modulation transfer function, MTF), remote sensing images are entered The denoising based on frequency domain for the row and the operation of MTF stretching, to improve picture quality, and judge whether the restored image after improvement has It is obviously improved：If having clear improvement, entering next step and processing；If being not improved, terminating in-orbit optimization and operating；

Step 3, by image batch processing method extract through quality improve after restored image image parameter, including：Side Edge energy, contrast, definition, comentropy, details energy, variance, image correlation, average, power spectrum, signal to noise ratio；

Anti- corresponding to step 4, the image parameter input focal length by extracting abovementioned steps, phase shift, front tropism parameter Spread out in model, obtain new parameter；

Step 5, by new parameter is replaced originally corresponding in-orbit parameter, produce new remote sensing images, and return step Rapid 2.

Using the above-mentioned implementation process of the present invention, real-time remote sensing images MTF indicatrix image restoration algorithm can be based on, Known or in the case of accurately calculating MTF decline curve, restored image effectively, to improve picture quality.Meanwhile, it is The impact of suppression noise, using novel frequency domain technique in previous embodiment, that is, passes through from the cluster MTF decline curve A suitable curve is selected to stretch spectral image, to expand the scope of application and to improve recovery effect；And using based on spy Determine the anti-model that spreads out of parameter of remote sensing images foundation, the instant image parameter information according to the high-quality remote sensing images restoring is accurate Calculate corresponding in-orbit parameter value, the automatic real-time optimization that this allows for in-orbit parameter is possibly realized, be that remote sensor parameter (includes Focal length, phase shift, front tropism) adjustment with optimize provide reference.

Due to being affected by factors such as sensor performance decline, atmospheric perturbations in remote sensing images imaging process, cause image matter Amount is degenerated.In order to improve Remote Sensing Image Quality, from deteriroation of image quality principle analysis, in the step 2 of the present embodiment, by calculating The MTF curve of image sampling system counter to push away the upper of real image frequency spectrum medium-high frequency part in the decline degree of different spaces frequency range The degree of liter.Due to image gather digitized during loss mainly high-frequency information, its show as on image details, Edge etc. fuzzy, the high-frequency information of therefore lifting image can improve image resolution ratio, also just improves picture quality.

In abovementioned steps 2, solve the MTF curve of original remote sensing images initially with blade method, draw followed by MTF Stretch process, for example with Chen et al. propose based on modulation transfer function (modulation transfer function, MTF) theoretical image recovery method.

In order to simulate cluster MTF curve, mtf value is rewritten as following form：

MTF_new=(MTF_old)^t, 0 ＜ t ＜ 2

Here MTF_oldRepresent the MTF curve of original image, MTF_newRepresent the MTF curve after adjusting through parameter t, adjust Parameter t adjusts the decline degree of MTF curve, and t is bigger, and MTF curve declines faster；Conversely, MTF curve declines slower.? During with MTF curve restored image, can pass through to adjust t, that is, by selecting suitable MTF decline curve come restored image, so that Obtain preferable recovery effect.

In conjunction with shown in Fig. 1, during using MTF theory restored image, mainly it is an up the HFS of image, but due to noise It is also HFS, if so not suppressing to noise, likely resulting in the image after recovery and containing a lot of noises, with As for covering effective information.Therefore in the present embodiment, also carry out following frequency domain denoising.

As an example specifically, the denoising based on frequency domain is carried out to remote sensing images, it realizes including：

1) as Fourier transformation, spectral image is obtained to original remote sensing images；

2) remaining each pixel (x, y) in addition to isolated bright spot in spectral image is handled as follows：

T (i, j)=R (x, y)-R (x-2+i, y-2+j)

I, j=1,2,3

In formula, R (x, y) represents the gray value of pixel (x, y), and T represents current pixel point and centered on this pixel It is the matrix of differences of nine pixel gray values in the range of the nine grids of point；S is the Description Matrix of T, to the matrix T of 3*3 each The gray scale difference value of point, if it is more than 0.3 times of the corresponding central point gray value of original remote sensing images, the corresponding points of matrix S are designated as 1, otherwise it is designated as 0；If the point being recorded as 1 in S is more than 6 then it represents that the gray value significantly greater than point about of current point, it is judged as Isolated bright spot, updates the half that its gray value is former ash angle value, otherwise, gray value is constant；

3) Fourier inversion is done to the spectral image after processing, gray average is adjusted to and original remote sensing images simultaneously Phase

4) MTF recovery is carried out to the spectral image removing after isolated bright spot, just have the following mesh recovering original remote sensing images Mark frequency domain figure is as the method for F：

F=(G-N)/(MTF_new·k)

G is the degraded image observed, N is noise image, makes the spectral image that R=G-N is after removing isolated bright spot, with Seasonal k=1, then simplified formula be：

F=R/MTF_new

5) Fourier inversion is carried out to the spectral image F after aforementioned MTF stretching, carry out image restoration, in order that multiple Image F after former is consistent substantially with the energy of the former Degenerate Graphs observed, after restoring in the lump, the gray average of image is adjusted Whole to identical with original remote sensing images.So, the image after being restored.

Due to being affected by image texture itself, some image parameter (edge energy, contrast, definition, information Entropy, details energy, variance, image correlation, average, power spectrum, signal to noise ratio) numerical value reaches the difference of tens units, in order to disappear Except this impact, make optimum results more accurate, therefore also include to image parameter result of calculation in the method for the present embodiment Normalized.

As preferred scheme, preceding method further includes following steps：Obtain changing through quality by image batch processing method After the image parameter of the restored image dealt with problems arising from an accident, also include the normalized to image parameter：

Processed using following method for normalizing：

Wherein, x_iFor the corresponding image parameter of n-th image sample.

In conjunction with shown in Fig. 1, then spread out in model by counter for the image parameter input after normalized, you can obtain new remote sensing The new in-orbit parameter of remote sensor is now replaced originally corresponding in-orbit parameter by device in-orbit parameter again, produces new remote sensing images, And return to step 2.

In the present embodiment, determined whether based on the BRISQUE non-reference picture quality method of NSS characteristic using existing Need persistently to improve picture quality or terminate operation, evaluate score value more low image quality higher.Based on NSS characteristic BRISQUE non-reference picture quality method, consists of shape and the variance parameter of model the MSCN parameter calculating each pixel Finally feature learning is carried out using SVR approximating method, finally by the characteristic model learning, image is scored.

Therefore, in abovementioned steps 2, by judging the objective evaluation score value of the restored image after optimizing and original remote sensing images Determine whether picture quality has clear improvement, that is,：

IfThen it is judged to that picture quality is not improved, terminate operation, otherwise, proceed excellent Change operation, wherein：

P_newAnd P_oldRepresent the objective evaluation score value of the restored image after optimizing and original remote sensing images respectively, ε is to set Threshold value.

Rule of thumb set, threshold value ε is usually arranged as 0.1.Certainly, in other examples, according to image and The actual conditions processing can also be set as other threshold values, but its value is in [0,1].

As optional embodiment, preceding method is further comprising the steps of：

The anti-model that spreads out corresponding to aforementioned focal length, phase shift, front tropism parameter is set：

1) the anti-model that spreads out of the parameter of focal length：

Focal length=9.003 × Image³+71.480×Image²+268.990×Image+735.662

Wherein：Image is defined as follows：

Image=0.43667230 × edge energy+0.45446772 × contrast+0.45446772 × definition+

0.43439084 × comentropy+0.45583660 × details energy；

2) the anti-model that spreads out of the parameter of phase shift

Phase shift=0.28+0.89 × 0.253^x

Wherein, x representative image overall target, it is defined as follows：

Image synthesis index=0.123* contrast+0.134* variance+0.132* definition+0.133* comentropy+ 0.126* details energy -0.128* image correlation+0.128* average+0.132* power spectrum；

3) the anti-model that spreads out of the parameter of tropism before

Y=-0.053 × a³+0.150×a²-0.193×a+0.101

Wherein, y is anti-calculated new front tropism parameter of spreading out, and a is defined as follows：

A=

0.3548 × contrast+0.3607 × comentropy+0.3613 × variance+0.3624* definition -0.3576* noise Than

+ 0.3597 × details energy -0.3344 × image correlation+0.3363 × edge energy.

Shown by above one or more embodiments of the invention, proposed by the invention is existed based on the remote sensor of MTF The method of the in-orbit parameter of rail Automatic Optimal, based on real-time remote sensing images MTF indicatrix image restoration algorithm, known or In the case of accurately calculating MTF decline curve, can restored image effectively, to improve picture quality.In order to suppress noise Impact, employ the frequency domain technique of novelty, that is, by select from cluster MTF decline curve suitable curve Lai Stretching spectral image, so can expand the scope of application and improve recovery effect；Set up using based on specific remote sensing images simultaneously The anti-model that spreads out of parameter, the instant image parameter information according to the high-quality remote sensing images restoring accurately calculates corresponding in-orbit Parameter value, the automatic real-time optimization that this allows for in-orbit parameter is possibly realized, by remote sensor on-orbit performance parameter and image matter Amount evaluating estimates state and the performance of remote sensor, be remote sensor parameter (including focal length, phase shift, front tropism) adjustment with excellent Change and reference is provided.

Although the present invention is disclosed above with preferred embodiment, so it is not limited to the present invention.The affiliated skill of the present invention Has usually intellectual, without departing from the spirit and scope of the present invention, when can be used for a variety of modifications and variations in art field.Cause This, protection scope of the present invention ought be defined depending on those as defined in claim.

Claims (4)

1. a kind of method of the in-orbit parameter of the in-orbit Automatic Optimal of the remote sensor based on MTF is it is characterised in that comprise the following steps：

Step 1, the original remote sensing images obtaining under currently in-orbit parameter is run；

Step 2, be based on modulation transfer function modulation transfer function, MTF, original remote sensing images are carried out The operation that denoising based on frequency domain is stretched with MTF, to improve picture quality, and it is bright to judge whether the restored image after improvement has Aobvious improvement：If having clear improvement, entering next step and processing；If being not improved, terminating in-orbit optimization and operating；

Step 3, by image batch processing method extract through quality improve after restored image image parameter, including：Edge energy Amount, contrast, definition, comentropy, details energy, variance, image correlation, average, power spectrum, signal to noise ratio；

The anti-mould that spreads out corresponding to step 4, the image parameter input focal length by extracting abovementioned steps, phase shift, front tropism parameter In type, obtain new parameter；

Step 5, by new parameter is replaced originally corresponding in-orbit parameter, produce new remote sensing images, and return to step 2；

Wherein, preceding method is further comprising the steps of：

The anti-model that spreads out corresponding to aforementioned focal length, phase shift, front tropism parameter is set：

1) the anti-model that spreads out of the parameter of focal length：

Focal length=9.003 × Image³+71.480×Image²+268.990×Image+735.662

Wherein：Image is defined as follows：

Image=0.43667230 × edge energy+0.45446772 × contrast+0.45446772 × definition+ 0.43439084 × comentropy+0.45583660 × details energy；

2) the anti-model that spreads out of the parameter of phase shift

Phase shift=0.28+0.89 × 0.253^x

Wherein, x representative image overall target, it is defined as follows：

Image synthesis index=0.123* contrast+0.134* variance+0.132* definition+0.133* comentropy+0.126* is thin Amount of energy saving -0.128* image correlation+0.128* average+0.132* power spectrum；

3) the anti-model that spreads out of the parameter of tropism before

Y=-0.053 × a³+0.150×a²-0.193×a+0.101

Wherein, y is anti-calculated new front tropism parameter of spreading out, and a is defined as follows：A=0.3548 × contrast+0.3607 × comentropy+0.3613 × variance+0.3624* definition -0.3576* signal to noise ratio+0.3597 × details energy -0.3344 × figure As related+0.3363 × edge energy.

2. the method for the in-orbit parameter of the in-orbit Automatic Optimal of the remote sensor based on MTF according to claim 1, its feature exists In, in abovementioned steps 2, using blade method solve original remote sensing images MTF curve, then carry out MTF stretching, that is,：

The mtf value of the remote sensing images that abovementioned steps 1 are got is rewritten into following form：

MTF_new=(MTF_old)^t

In formula, t is regulation parameter, 0 ＜ t ＜ 2, MTF_oldRepresent the MTF curve of original remote sensing images, MTF_newRepresent through parameter t MTF curve after regulation.

3. the method for the in-orbit parameter of the in-orbit Automatic Optimal of the remote sensor based on MTF according to claim 2, its feature exists In, in front step 2, the denoising based on frequency domain is carried out to original remote sensing images, its realize include：

1) as Fourier transformation, spectral image is obtained to original remote sensing images；

2) remaining each pixel (x, y) in addition to isolated bright spot in spectral image is handled as follows：

T (i, j)=R (x, y)-R (x-2+i, y-2+j)

I, j=1,2,3

$S(i,j)=\left\{\begin{array}{cc}1& ifT(i,j)>0.3\·R(x,y)\\ 0& otherwise\end{array}\right.$

$R(x,y)=\left\{\begin{array}{cc}0.5\·R(x,y)& ifsum\left(S\right)>6\\ R(x,y)& else\end{array}\right.$

In formula, R (x, y) represents the gray value of pixel (x, y), and T represents current pixel point and point centered on this pixel It is the matrix of differences of nine pixel gray values in the range of nine grids；S is the Description Matrix of T, to each point of the matrix T of 3*3 Gray scale difference value, if it is more than 0.3 times of the corresponding central point gray value of original remote sensing images, the corresponding points of matrix S are designated as 1, no Then it is designated as 0；If the point being recorded as 1 in S is more than 6 then it represents that the gray value significantly greater than point about of current point, it is judged as isolating Bright spot, updates the half that its gray value is former ash angle value, otherwise, gray value is constant；

3) Fourier inversion is done to the spectral image after processing, gray average is adjusted to and original remote sensing images phase simultaneously With；

4) MTF recovery is carried out to the spectral image removing after isolated bright spot, just have the following target frequency recovering original remote sensing images The method of area image F：

F=(G-N)/(MTF_new·k)

G is the degraded image observed, N is noise image, makes the spectral image that R=G-N is after removing isolated bright spot, with season K=1, then simplified formula be：

F=R/MTF_new

5) Fourier inversion is carried out to the spectral image F after aforementioned MTF stretching, carry out image restoration, and will scheme after restoring The gray average of picture is adjusted to identical with original remote sensing images.

4. the method for the in-orbit parameter of the in-orbit Automatic Optimal of the remote sensor based on MTF according to claim 3, its feature exists In preceding method further includes following steps：The image of the restored image after image batch processing method obtains improving through quality After parameter, also include the normalized to image parameter：

Processed using following method for normalizing：

$x_i=\frac{x_i-min(x_1,x_2,...,x_n)}{max(x_1,x_2,...,x_n)-\min (x_1,x_2,...,x_n)}$

Wherein, x_iFor the corresponding image parameter of i-th image pattern.