CN108921809A - Multispectral and panchromatic image fusion method under integral principle based on spatial frequency - Google Patents
Multispectral and panchromatic image fusion method under integral principle based on spatial frequency Download PDFInfo
- Publication number
- CN108921809A CN108921809A CN201810593011.3A CN201810593011A CN108921809A CN 108921809 A CN108921809 A CN 108921809A CN 201810593011 A CN201810593011 A CN 201810593011A CN 108921809 A CN108921809 A CN 108921809A
- Authority
- CN
- China
- Prior art keywords
- image
- coefficient
- fusion
- nsct
- multispectral
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000007500 overflow downdraw method Methods 0.000 title claims abstract description 19
- 230000009466 transformation Effects 0.000 claims abstract description 41
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 33
- 230000004927 fusion Effects 0.000 claims description 68
- 238000000034 method Methods 0.000 claims description 28
- 238000005070 sampling Methods 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000005303 weighing Methods 0.000 claims description 3
- 238000002156 mixing Methods 0.000 abstract description 23
- 238000012545 processing Methods 0.000 abstract description 9
- 230000000007 visual effect Effects 0.000 abstract description 6
- 238000000605 extraction Methods 0.000 abstract description 4
- 230000003595 spectral effect Effects 0.000 abstract description 4
- 239000000203 mixture Substances 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 7
- 238000001228 spectrum Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10041—Panchromatic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The present invention provides the multispectral and panchromatic image fusion methods under integral principle based on spatial frequency, if A is multispectral image, B is full-colour image；IHS transformation is carried out to A first, obtains brightness IA, coloration HA, saturation degree SA；Then NSCT decomposition is carried out respectively to IA and B；The NSCT coefficient of IA and B is merged again；Finally according to determining low pass subband coefficient and band logical sub-band coefficients, NCST inverse transformation is carried out, reconstruct, which generates, merges new IA component, and Inew and HA, SA are carried out IHS inverse transformation, be transformed into rgb space, obtain fused new images by referred to as Inew.The present invention is multispectral by NSCT based on spatial frequency under integral principle and full-colour image blends, high-quality, the spectral information high resolution of blending image, informative, it is well arranged, there is better visual effect and statistical indicator, be conducive to the subsequent processings such as remote sensing images extraction, classification.
Description
Technical field
The invention belongs to remote sensing science and technology field, be specifically related under integral principle multispectral based on spatial frequency and
Panchromatic image fusion method.
Background technique
Remote sensing is that have different reflection of electromagnetic wave or radiation feature using any object, by obtaining ground target reflection
Or the electromagnetic wave radiated to be to obtain target information, and handles information obtained, realizes the positioning of target, knows
Not, description etc. qualitatively or quantitatively.
Remote sensing science is a Comprehensive Science, it is with Spatial Information Science, electronic information science, optics, computer section
It is the mankind's activities such as national economy and scientific research by the observation to spatial object based on the subjects such as, biology, chemistry,
Provide remote sensing image data abundant.
Contourlet transformation and NSCT transformation (non-downsampling Contourlet conversion) are answered in field of remote sensing image processing
It is very much, including denoising, compression, coding, fusion etc..Because Contourlet and NSCT transformation can effectively capture image
Edge feature, and the most information of image has been concentrated in the marginal portion of image, and is the important feature of image, this is being handled
It is equal to catch " center of gravity " of image when image.
The super more wave band number of high spectrum image forms the data of magnanimity, and processing high spectrum image is just faced with dimension mistake
The difficulties such as how more bring wave bands select, operation efficiency is low.Therefore in the processing of high spectrum image, pressure is always existed
Contracting, waveband selection, fusion research, many research achievements, which can play, reduces EO-1 hyperion dimension, the purpose that improves efficiency.
The conventional method of image co-registration is very more, very mature, but does not represent these methods all very " excellent ".Mesh
Before, the hot spot for merging field is concentrated on based on multi-scale geometric analysis tool method.This is because NSCT has translation invariant
Property, all subgraphs decomposed have pixel space identical with source images, can make full use of the region of source images segmentation
Information guiding fusion process；Compared to small echo, Laplacian-pyramid method etc., the NSCT of non-lower sampling has to image geometry spy
Stronger ability to express is levied, wherein small wave converting method only has limited directionality, and the NSCT of non-lower sampling has random angle
The directionality of degree " can more capture " in image " edge " marginal information.Some scholars do in terms of the image co-registration based on NCST
A large amount of work, mainly there is the fusion for medical image, infrared image and visual image fusion, full-colour image, EO-1 hyperion
Fusion, SAR image and fusion of multispectral image of fusion, polarimetric SAR image between multispectral image etc..It is wherein big
Quantity research, which works, all concentrates on the fusions of Medical image fusion or infrared image and visible images, and other aspects be related to compared with
It is few.
Image Fusion Rule can be divided into three classes：It is pixel-based fusion, the fusion based on neighborhood window and be based on region
Fusion.Based on above-mentioned three kinds of basic skills also have weighting, it is adaptive weighted, also have and preferentially chosen according to certain standard
's.Fusion based on single pixel is a kind of relatively common method, but this method is equal to every the back for having split whole image
Scape generally requires the detection for doing consistency.And the fusion rule based on neighborhood window is using neighborhood window as investigation object, using working as
The statistical property in mask window calculation neighborhood centered on preceding pixel, such as variance, shade of gray, local gray level energy.?
In one neighborhood window, statistical property is more obvious, and illustrates that image gray levels variation is bigger, details is abundanter.Melting based on region
Normally using the region after image segmentation as research object, a part of the corresponding target in each region or target, part point
Cutting region is the part that information is rich in fusion, it is desirable that the registration accuracy of two width blending images must be very high, is additionally based on region
Fusion increase fusion steps, while also increasing calculation amount.
The fusion of low-frequency information can be and simply use some methods of weighting or adaptive method of weighting.
These schemes all do not fully consider the characteristics of image such as edge, the texture of image, as a result, reducing the contrast of image.
In conclusion current image interfusion method there are many defects, need further in image co-registration quality
It improves.
Summary of the invention
The technical problem to be solved by the present invention is to how improve the quality of image co-registration, make blending image that there is preferably view
Feel effect and statistical indicator.
In order to solve the above-mentioned technical problems, the present invention provides multispectral based on spatial frequency under a kind of integral principle and
Panchromatic image fusion method, which employs the following technical solutions：
Multispectral and panchromatic image fusion method based on spatial frequency under a kind of integral principle, it is characterised in that：Specifically
Steps are as follows：
Step 1：If A represents multispectral image, B represents full-colour image；Step 2：IHS transformation is carried out to multispectral image A,
Obtain brightness IA, coloration HA, saturation degree SAThree components；
Step 3：To IANSCT decomposition is carried out respectively with B；
Step 4：To IAIt is merged with the NSCT coefficient of B；
Step 5：NSCT reconstruct；
According to the low pass subband coefficient and band logical sub-band coefficients determined in step 4, NCST inverse transformation is carried out, reconstruct, which generates, melts
Close new IAComponent, referred to as Inew；
Step 6：IHS inverse transformation；
It will obtain new component InewWith the coloration H in step 2A, saturation degree SAThree components carry out IHS inverse transformation, conversion
To rgb space, fused new images are obtained.
Preferably, in the step 3, the scale number of plies that NSCT is decomposed is 3, and corresponding direction number is [2,4,8]；IAFigure
The NSCT decomposition coefficient of picture is LAWithThe NSCT decomposition coefficient of B image is LBWithWherein, LA、LBRespectively represent IAAnd B
Low pass subband coefficient；Respectively represent IAWith the band logical sub-band coefficients of B；K=1,2,3, it is multi-resolution decomposition layer
Number；Corresponding j=2k=[1,2；1,2,3,4；It 1,2 ..., 8], is that non-lower sampling under Decomposition order j decomposes direction number, this
The j circular in definition at place.
Preferably, in the step 4, the fusion rule of low pass subband coefficient：Low frequency fusion is using the low of multispectral image
Frequency sub-band coefficients LF=LA,LAIt is the low pass subband coefficient of A image.
Preferably, in the step 4, the fusion rule of band logical sub-band coefficients is：Band logical subband will be according under integral principle
Regional space frequency be weighing criteria, merged；Image A and B carries out NSCT decomposition all in accordance with the setting in step 2, obtains
To band logical subbandWithIt is finally to be decomposed by non-lower sampling anisotropic filter, but it is big
Small all directions subband saves the coefficient of direction decomposition respectively as image A, B, also saves the letter in non-supported region
Breath；In addition to calculating k=1,2,3, corresponding j=[1,2；1,2,3,4；1,2 ..., 8] whenDirectional subband coefficient,
Also calculate k=1,2,3, j=[1；1；Directional subband coefficient when 1], at this moment the direction number of every layer of band logical subband is all 1, note
For：Size be also with B with image A as, such structure can be generated multiple, it is therefore an objective to
For savingDirectional subband coefficient.
Using the high frequency coefficient fusion rule of maximum absolute value pixel-based, its essence is compare the single picture of high-frequency sub-band
The order of magnitude of element therefrom chooses the biggish fusion results as fusion high-frequency sub-band of absolute value, model formula
(1.1) it indicates：
In formula (1.1), CnewTo merge obtained maximum high frequency coefficient, CAFor the high frequency coefficient of A image, CBFor B image
High frequency coefficient；
It is a kind of extended method pixel-based based on the adaptive weighted of field window statistical property；By currentElement
Centered on, the window that a size is 3 × 3 or 5 × 5 delimited, the statistical property of the window is calculated；Then by statistical property
As index, then or size comparison is carried out, directly accepted or rejected, or calculated weight and be weighted fusion；Its model formula
(1.2) it indicates：
Cnew=qCA+pCB, p+q=1 (1.2)
CnewTo merge obtained maximum high frequency coefficient, CAFor the high frequency coefficient of A image, CBFor the high frequency coefficient of B image, p
For the power of the high frequency coefficient of B image, q is the power of the high frequency coefficient of A image.The spatial frequency of image reflects image on airspace
Overall active degree, spatial frequency is bigger, and image is more clear, conversely, image is fuzzyyer；Size is the space frequency of M × N image
The definition of rate SF is：
Wherein
M, N is the positive integer not less than 2；F (m, n) is the pixel value of image m row n column.
Taking size is 3 × 3 or 5 × 5 window, calculates band logical subband according to formula (1.3), (1.4) and (1.5)Regional space frequency, corresponding result is stored inBased on band logical subband according under integral principle
Regional space frequency calculates band logical sub-band coefficientsIt is as follows：
Method provided by the invention overcomes the deficiencies in the prior art, based on spatial frequency that NSCT is more under integral principle
Spectrum and full-colour image blend, high-quality, the spectral information high resolution of blending image, image clearly, informative, layer
It is secondary clearly demarcated, there is better visual effect and statistical indicator, be conducive to the subsequent processings such as extraction, the classification of remote sensing images；This hair
The numerical procedure calculation amount of bright proposition reduces, and computational efficiency is higher.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
The neighborhood window schematic diagram that Fig. 1 is 3 × 3；
The neighborhood window schematic diagram that Fig. 2 is 5 × 5；
Fig. 3 is the blending image that IHS is converted；
Fig. 4 is the blending image that Wavelet Transform Fusion obtains；
Fig. 5 is the blending image that contourlet transformation merges；
Fig. 6 is the blending image that traditional NSCT fusion method obtains；
Fig. 7 is the blending image that the method for the present invention obtains.
Specific embodiment
Hereinafter reference will be made to the drawings, describes method provided by the invention in detail by way of example.It needs to illustrate herein
, the descriptions of the manners of these embodiments are used to facilitate the understanding of the present invention, but and does not constitute a limitation of the invention.
The terms "and/or", only a kind of incidence relation for describing affiliated partner, indicates that there may be three kinds of passes
System, for example, A and/or B, can indicate：Individualism A, individualism B exist simultaneously tri- kinds of situations of A and B, the terms
"/and " it is to describe another affiliated partner relationship, indicate may exist two kinds of relationships, for example, A/ and B, can indicate：Individually deposit
In A, two kinds of situations of individualism A and B, in addition, character "/" herein, typicallying represent forward-backward correlation object is a kind of "or" pass
System.
Embodiment one
The present embodiment mainly introduces the base of the multispectral and panchromatic image fusion method under integral principle based on spatial frequency
Present principles and step.
One, basic definition
High frequency coefficient representative is image " obvious " feature, the details of image has been reacted, such as edge, Texture eigenvalue；
NSCT transformation is good at capturing image " edge " again, and therefore, the fusion of high frequency coefficient is very crucial.The present invention, which uses, is based on pixel
Maximum absolute value high frequency coefficient fusion rule, its essence is the orders of magnitude for comparing high-frequency sub-band single pixel, therefrom
The biggish fusion results as fusion high-frequency sub-band of absolute value are chosen, model can be indicated with formula (1.1).
In formula (1.1), CnewTo merge obtained maximum high frequency coefficient, CAFor the high frequency coefficient of A image, CBFor B image
High frequency coefficient.
Based on the adaptive weighted of field window statistical property, it may be said that be a kind of extended method pixel-based.It is
Centered on currentElement, the window that size is 3 × 3 or 5 × 5 usually delimited, (attached drawing 1 is as shown in attached drawing 1-2
3 × 3 neighborhood window schematic diagram, the neighborhood window schematic diagram that attached drawing 2 is 5 × 5), calculate the statistical property of the window；Then will
Statistical property then or carries out size comparison, directly accepts or rejects as index, or calculates weight and be weighted fusion.With public affairs
The model of formula (1.2) description Weighted Fusion.The present embodiment selection is based on 3 × 3 field window, 3 × 3 windows under calculating integrally
Spatial frequency compare size as index, take absolute value big spatial frequency, the high frequency coefficient as fusion.Formula
(1.2) as follows：
Cnew=qCA+pCB, p+q=1 (1.2)
In formula (1.2), CnewTo merge obtained maximum high frequency coefficient, CAFor the high frequency coefficient of A image, CBFor B image
High frequency coefficient, p are the power of the high frequency coefficient of B image, and q is the power of the high frequency coefficient of A image；
The spatial frequency of image reflects overall active degree of the image on airspace, and spatial frequency is bigger, and image is more clear
It is clear, conversely, image is fuzzyyer.
Size is that the definition of the spatial frequency SF of M × N image is：
Wherein
M, N is the positive integer not less than 2, and f (m, n) is the pixel value of image m row n column.
The biggish coefficient of absolute value corresponds to some mutation in the band logical subband of NSCT transformation, edge, texture such as image
Equal important features information.Select to carry out the fusion based on region by module of spatial frequency, can reflect out image compared with
For true high-frequency information.
Two, the multispectral and panchromatic image fusion method under integral principle based on spatial frequency
The specific step of multispectral and panchromatic image fusion method under a kind of integral principle of the invention based on spatial frequency
Suddenly include：
Step 1：If A represents multispectral image, B represents full-colour image；
Step 2：IHS transformation is carried out to multispectral image A, obtains brightness IA, coloration HA, saturation degree SAThree components, then
IAThe NSCT that component participates in below step is decomposed, fusion and NSCT are changed against transformation；
Step 3：To IANSCT decomposition is carried out respectively with B；
Specifically, the scale number of plies that NSCT is decomposed is 3, and corresponding direction number is [2,4,8] in this fusion method；IA
The NSCT decomposition coefficient of image is LAWithThe NSCT decomposition coefficient of B image is LBWithWherein, LA、LBRespectively represent IA
With the low pass subband coefficient of B；Respectively represent IAWith the band logical sub-band coefficients of B；K=1,2,3, it is multi-resolution decomposition
The number of plies；Corresponding j=2k=[1,2；1,2,3,4；It 1,2 ..., 8], is the non-lower sampling decomposition of corresponding decomposition under Decomposition order
Direction number, j circular in definition herein.
Step 4：To IAIt is merged with the NSCT coefficient of B；
Specifically, the fusion rule of low pass subband coefficient：When multi-resolution decomposition, (high-resolution is complete for discovery PAN image
Chromatic graph picture) spatial information is weaker, and low frequency energy is very low, so low frequency fusion uses the low frequency sub-band L of multispectral imageF=LA,
LAIt is the low pass subband coefficient of A image.
Specifically, the fusion rule of band logical sub-band coefficients：Band logical subband will be according to the regional space frequency under integral principle
For weighing criteria, merged.Image A and B carries out NSCT decomposition all in accordance with the setting in step 2, obtains band logical subband WithIt is finally to be decomposed by non-lower sampling anisotropic filter, but its size is as image A, B,
All directions subband saves the coefficient of direction decomposition respectively, also saves the information in non-supported region, and these are non-supported
If the calculating that area information participates in spatial frequency will weaken direction coefficient.It is therefore proposed that calculating area under integral principle
Domain space frequency, the measurement standard as fusion.In addition to calculating k=1,2,3, corresponding j=[1,2；1,2,3,4；1,2,…,
When 8]Directional subband coefficient, also calculating k=1,2,3, j=[1；1；Directional subband coefficient when 1], at this moment often
The direction number of layer band logical subband is all 1, is denoted as：Size be also with image A, B it is the same, save
The direction coefficient of kth layer.Taking size is 3 × 3 or 5 × 5 window Q (taking into account calculation amount, the present embodiment takes 3 × 3), according to public affairs
Formula (1.3), (1.4) and (1.5) calculates band logical subbandRegional space frequency, be stored in accordingly
So far, band logical subband direction fusion coefficients can be calculated based on band logical subband by according to the regional space frequency under integral principleIt is as follows：
Step 5：NSCT reconstruct：According to the low pass subband L determined in above-mentioned stepsFWith band logical subbandIt is inverse to carry out NCST
Transformation, reconstruct, which generates, merges new IAComponent, referred to as Inew。
Step 6：IHS inverse transformation：It will obtain new measurement component InewWith the coloration H in step 2A, saturation degree SAThree points
Amount carries out IHS inverse transformation, is transformed into rgb space, obtains fused new images.
In the present embodiment, by the luminance component I of multispectral image AATwo kinds of NSCT transformation are all carried out respectively with full-colour image B
(one is hierarchy number k=3, l=[1,2,3], direction number is [21,22,23]；Another is Decomposition order k=3, and l=
[0,0,0], direction number are [20,20,20]) (l is the index in the decomposition direction of corresponding decomposition layer, for example corresponds to Decomposition order
Direction number is 20, i.e. decomposition direction number is 1), wherein subsequent this decomposition method is for calculating window under integral principle
Spatial frequency, and the foundation as high frequency fusion；The fusion of low frequency sub-band has directlyed adopt the low frequency sub-band of full-colour image, high frequency
The fusion of band logical subband is the maximum absolute value principle according to spatial frequency；By fused each subband according to NSCT inverse transformation
Reconstruct obtains new luminance component Inew；By Inew、HA、SAThree components carry out IHS inverse transformation and obtain fusion new images, such as attached drawing 7
It is shown.
Blending image low frequency coefficient of the present invention has directlyed adopt the low frequency coefficient of full-colour image, and reason is multispectral image
The more spectral informations for including；And full-colour image resolution ratio with higher, it includes contain much information, edge, profile and
The characteristics of image such as texture become apparent from deeply, and compared to multispectral image, its spatial information is more acurrate, the higher full-colour picture of resolution ratio
As more can approximatively reflect image.
The present invention by NSCT and PCA (Principal Component Analysis) transformation combines, merge high-spectrum remote sensing, by pair
Than analysis, the remote sensing images of better quality are obtained, the subsequent processings such as extraction, the classification of remote sensing images will be also more advantageous to.
Embodiment two
For the technical effect of further instruction technical solution of the present invention, the present embodiment is by the place of technical solution of the present invention
It manages result and is compared analysis with the processing result of conventional method.The blending image that IHS is converted is as shown in Fig. 3.
Wavelet Transform Fusion principle is the IHS transformation of multispectral image A first, then to its luminance component IAWith it is panchromatic distant
Feel image B and carry out three layers of small echo decomposition, wavelet function uses " db45 ", and low-frequency data is carried out average fusion, high frequency
It according to the principle according to maximum absolute value, is merged, then wavelet inverse transformation obtains new luminance component Inew, by Inew、HA、SA
IHS inverse transformation is carried out, obtains fusion new images, as shown in Fig. 4.
Contourlet transformation fusion method is classical traditional fusion rule, and it is inverse that multispectral image A is equally carried out IHS
Transformation, the luminance component I that transformation is obtainedcContourlet transformation is carried out with panchromatic remote sensing images, three layers of k=3 decompose,
The direction number of band logical subband is l=[1,2,3]；The two is merged according to low frequency sub-band average value, its absolute value of high-frequency sub-band
Maximum principle；Then I is reconstructed according to Contourlet inverse transformationcComponent, the I that will be obtainedc, that tri- components of H, S carry out IHS is inverse
Transformation obtains the new images of contourlet transformation fusion, as shown in Fig. 5.
Traditional fusion rule is combined to do test for fusion using NSCT, it is more convincing by comparison.Traditional NSCT
Its process of fusion method is similar with the NSCT integration program proposed by the present invention based on the spatial frequency under integral principle, decomposition layer
Several and band logical subband direction number be all it is identical, only traditional NSCT fusion uses more common fusion rule, i.e., low
Frequency subband removes average value, and high-frequency sub-band is chosen according to the principle of maximum absolute value, and processing result is as shown in Fig. 6.
Using the multispectral and panchromatic image fusion method based on spatial frequency under a kind of integral principle of the invention, place
It is as shown in Fig. 7 to manage result.
Abundant and picture quality the improvement of blending image information content is the basic goal of image co-registration always, and measures
The basic norm of various fusion method effects.Ideal fusion process should also have both to the intake of useful information to original useful
The reservation of information, syncretizing effect evaluation should include innovative and inheritance.
Clearance observation analysis Fig. 3-7 carry out subjective visual evaluation.Fig. 3 is that the blending algorithm IHS based on color space becomes
Change, fusion mass be it is worst in several method, image is relatively fuzzy, and spatial resolution is for other blending images
Lower, reason is that luminance component is changed.Whole visual observation：Method blending image (attached drawing proposed by the present invention
7) processing result is best, and on spectral information, resolution ratio all preferably, image clearly is well arranged, is better than other algorithms
Fusion results.Generally speaking, the blending image of the method for the present invention, tradition NSCT blending image, contourlet transformation fusion
Image is better than Wavelet Transform Fusion image, IHS transformation blending image, and quality is successively gradually poor.
For the fusion mass of further more several fusion methods, the present embodiment calculates the entropy, average of blending image
Gradient, standard deviation, concrete outcome are shown in Table 1.
By analytical table 1, it can be found that objectively evaluating in index, the fusion results of algorithm proposed by the present invention are preferable, this
It is consistent with above-mentioned subjective visual evaluation.Method IHS transformation, wavelet transformation WT, contourlet transformation, traditional NSCT,
The quality of the blending image of the method for the present invention is successively got higher, and the index of tri- wave bands of RGB of blending image successively improves, and is shown
Information content successively becomes richer.It is compared with WT with HIS transformation, based on the fusion that Contourlet and NSCT is converted, in spectrum
It is all improved on information and spatial information, effect is more excellent；Based on the blending image that Contourlet and NSCT is converted, side
Edge, detailed information are also more obvious, and which reflects advantage of the multi-scale geometric analysis tool in image procossing；Due to NSCT's
Translation invariance, result is also due to the fusion results based on contourlet transformation.The blending image solution that NSCT is converted
It translates interpretation effect more preferably, classifies for follow-up work, extraction process is also advantageously.
The statistical appraisal of 1 distinct methods of table
Note：What TM was represented is multispectral image, contains tri- wave bands of RGB.
The above, only presently preferred embodiments of the present invention, not to the present invention in any form with substantial limitation,
It should be pointed out that under the premise of not departing from the method for the present invention, can also be made for those skilled in the art
Several improvement and supplement, these are improved and supplement also should be regarded as protection scope of the present invention.All those skilled in the art,
Without departing from the spirit and scope of the present invention, when made using disclosed above technology contents it is a little more
Dynamic, modification and the equivalent variations developed, are equivalent embodiment of the invention；Meanwhile all substantial technologicals pair according to the present invention
The variation, modification and evolution of any equivalent variations made by above-described embodiment, still fall within the range of technical solution of the present invention
It is interior.
Claims (4)
1. the multispectral and panchromatic image fusion method under a kind of integral principle based on spatial frequency, it is characterised in that：This method
Specific steps include：
Step 1：If A represents multispectral image, B represents full-colour image；
Step 2：IHS transformation is carried out to multispectral image A, obtains brightness IA, coloration HA, saturation degree SAThree components；
Step 3：To IANSCT decomposition is carried out respectively with B；
Step 4：To IAIt is merged with the NSCT coefficient of B；
Step 5：NSCT reconstruct；According to the low pass subband coefficient and band logical sub-band coefficients determined in step 4, NCST inversion is carried out
It changes, reconstruct, which generates, merges new IAComponent, referred to as Inew；
Step 6：IHS inverse transformation；It will obtain new component InewWith the coloration H in step 2A, saturation degree SAThree components carry out IHS
Inverse transformation is transformed into rgb space, obtains fused new images.
2. the multispectral and panchromatic image fusion method based on spatial frequency under a kind of integral principle as described in claim 1,
It is characterized in that：In the step 3, the multi-resolution decomposition number of plies that NSCT is decomposed is 3, and corresponding direction number is [2,4,8]；IA
The NSCT decomposition coefficient of image is LAWithThe NSCT decomposition coefficient of B image is LBWithWherein, LA、LBRespectively represent IA
With the low pass subband coefficient of B；Respectively represent IAWith the band logical sub-band coefficients of B；K=1,2,3, it is multi-resolution decomposition
The number of plies；Corresponding j=2k=[1,2；1,2,3,4；It 1,2 ..., 8], is that non-lower sampling under Decomposition order decomposes direction number.
3. the multispectral and panchromatic image fusion method based on spatial frequency under a kind of integral principle as described in claim 1,
It is characterized in that：It include the fusion of low pass subband coefficient in the step 4, rule is：Low frequency fusion uses multispectral image
Low frequency sub-band coefficient LF=LA, LAIt is the low pass subband coefficient of A image.
4. the multispectral and full-colour image fusion side under a kind of integral principle as claimed in claim 1 or 3 based on spatial frequency
Method, it is characterised in that：It include the fusion of band logical sub-band coefficients in the step 4, rule is：Band logical subband will be according to whole original
Regional space frequency under then is weighing criteria, is merged；Image A and B carries out NSCT points all in accordance with the setting in step 2
Solution, obtains band logical sub-band division coefficient WithIt is finally to decompose to obtain by non-lower sampling anisotropic filter
, but its size, as image A, B, all directions subband saves the coefficient of direction decomposition respectively, also saves non-
The information of supporting zone, and if the calculating of these non-supported area informations participation spatial frequencys will weaken direction coefficient；
In addition to calculate=1,2,3, it is corresponding=[1,2；1,2,3,4；1,2 ..., 8] whenDirectional subband coefficient, is also counted
Calculation=1,2,3 ,=[1；1；Directional subband coefficient when 1], at this moment the direction number of every layer of band logical subband is all 1, is denoted as：Size be also with image A, B it is the same, such structure can be generated multiple, it is therefore an objective to for protecting
It depositsDirectional subband coefficient；
Using the high frequency coefficient fusion rule of maximum absolute value pixel-based, model is indicated with formula (1.1)：
In formula (1.1), CnewTo merge obtained maximum high frequency coefficient, CAFor the high frequency coefficient of A image, CBFor the high frequency of B image
Coefficient；
It is a kind of extended method pixel-based based on the adaptive weighted of field window statistical property；Using currentElement as
Center delimit the window that a size is 3 × 3 or 5 × 5, calculates the statistical property of the window；Then using statistical property as
Index then or carries out size comparison, directly accepts or rejects, or calculates weight and be weighted fusion；Its model is with formula (1.2)
It indicates：
Cnew=qCA+pCB, p+q=1 (1.2)
In formula (1.2), CnewTo merge obtained maximum high frequency coefficient, CAFor the high frequency coefficient of A image, CBFor the high frequency of B image
Coefficient, p are the power of the high frequency coefficient of B image, and q is the power of the high frequency coefficient of A image；
The spatial frequency of image reflects overall active degree of the image on airspace, and spatial frequency is bigger, and image is more clear, instead
It, image is fuzzyyer；Size is that the definition of the spatial frequency SF of M × N image is：
Wherein
In formula, M, N are the positive integer not less than 2, and f (m, n) is the pixel value of image m row n column；Taking size is 3 × 3 or 5 × 5
Window calculates band logical subband according to formula (1.3), (1.4) and (1.5)Regional space frequency, corresponding result protects
In the presence ofBased on band logical subband by according to the regional space frequency under integral principle, band logical sub-band coefficients are calculatedIt is as follows：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810593011.3A CN108921809B (en) | 2018-06-11 | 2018-06-11 | Multispectral and panchromatic image fusion method based on spatial frequency under integral principle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810593011.3A CN108921809B (en) | 2018-06-11 | 2018-06-11 | Multispectral and panchromatic image fusion method based on spatial frequency under integral principle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108921809A true CN108921809A (en) | 2018-11-30 |
| CN108921809B CN108921809B (en) | 2022-02-18 |
Family
ID=64418746
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810593011.3A Expired - Fee Related CN108921809B (en) | 2018-06-11 | 2018-06-11 | Multispectral and panchromatic image fusion method based on spatial frequency under integral principle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108921809B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111259A (en) * | 2019-05-15 | 2019-08-09 | 电子科技大学 | A kind of multisource image anastomosing method based on regional guidance |
| CN111563866A (en) * | 2020-05-07 | 2020-08-21 | 重庆三峡学院 | Multi-source remote sensing image fusion method |
| CN112699832A (en) * | 2021-01-12 | 2021-04-23 | 腾讯科技（深圳）有限公司 | Target detection method, device, equipment and storage medium |
| CN113554112A (en) * | 2021-07-30 | 2021-10-26 | 西安交通大学 | Remote sensing image fusion method, system, equipment and medium |
| CN114897882A (en) * | 2022-06-10 | 2022-08-12 | 大连民族大学 | Remote sensing image fusion method based on weighted average curvature filter decomposition |
| CN115564692A (en) * | 2022-09-07 | 2023-01-03 | 宁波大学 | Panchromatic-multispectral-hyperspectral integrated fusion method considering width difference |
| CN115861141A (en) * | 2022-12-02 | 2023-03-28 | 北京领云时代科技有限公司 | Unmanned aerial vehicle image acquisition processing system and method based on PCNN neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521818A (en) * | 2011-12-05 | 2012-06-27 | 西北工业大学 | Fusion method of SAR (Synthetic Aperture Radar) images and visible light images on the basis of NSCT (Non Subsampled Contourlet Transform) |
| US20140119668A1 (en) * | 2012-10-31 | 2014-05-01 | Samsung Electronics Co., Ltd. | Medical image processing apparatus and method for synthesizing and displaying medical image generated using multi-energy x-ray |
| CN103985105A (en) * | 2014-02-20 | 2014-08-13 | 江南大学 | Contourlet domain multi-modal medical image fusion method based on statistical modeling |
| CN107194904A (en) * | 2017-05-09 | 2017-09-22 | 西北工业大学 | NSCT area image fusion methods based on supplement mechanism and PCNN |
-
2018
- 2018-06-11 CN CN201810593011.3A patent/CN108921809B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521818A (en) * | 2011-12-05 | 2012-06-27 | 西北工业大学 | Fusion method of SAR (Synthetic Aperture Radar) images and visible light images on the basis of NSCT (Non Subsampled Contourlet Transform) |
| US20140119668A1 (en) * | 2012-10-31 | 2014-05-01 | Samsung Electronics Co., Ltd. | Medical image processing apparatus and method for synthesizing and displaying medical image generated using multi-energy x-ray |
| CN103985105A (en) * | 2014-02-20 | 2014-08-13 | 江南大学 | Contourlet domain multi-modal medical image fusion method based on statistical modeling |
| CN107194904A (en) * | 2017-05-09 | 2017-09-22 | 西北工业大学 | NSCT area image fusion methods based on supplement mechanism and PCNN |
Non-Patent Citations (2)
| Title |
|---|
| 杨金库等: "多光谱图像与全色图像融合", 《西安工业大学学报》 * |
| 王峰等: "基于多特征的遥感图像融合算法", 《西北工业大学学报》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111259B (en) * | 2019-05-15 | 2021-03-02 | 电子科技大学 | Multi-source image fusion method based on region guidance |
| CN110111259A (en) * | 2019-05-15 | 2019-08-09 | 电子科技大学 | A kind of multisource image anastomosing method based on regional guidance |
| CN111563866B (en) * | 2020-05-07 | 2023-05-12 | 重庆三峡学院 | Multisource remote sensing image fusion method |
| CN111563866A (en) * | 2020-05-07 | 2020-08-21 | 重庆三峡学院 | Multi-source remote sensing image fusion method |
| CN112699832B (en) * | 2021-01-12 | 2023-07-04 | 腾讯科技（深圳）有限公司 | Target detection method, device, equipment and storage medium |
| CN112699832A (en) * | 2021-01-12 | 2021-04-23 | 腾讯科技（深圳）有限公司 | Target detection method, device, equipment and storage medium |
| CN113554112A (en) * | 2021-07-30 | 2021-10-26 | 西安交通大学 | Remote sensing image fusion method, system, equipment and medium |
| CN113554112B (en) * | 2021-07-30 | 2023-09-19 | 西安交通大学 | Remote sensing image fusion method, system, equipment and medium |
| CN114897882A (en) * | 2022-06-10 | 2022-08-12 | 大连民族大学 | Remote sensing image fusion method based on weighted average curvature filter decomposition |
| CN114897882B (en) * | 2022-06-10 | 2024-04-19 | 大连民族大学 | Remote sensing image fusion method based on weighted average curvature filter decomposition |
| CN115564692A (en) * | 2022-09-07 | 2023-01-03 | 宁波大学 | Panchromatic-multispectral-hyperspectral integrated fusion method considering width difference |
| CN115564692B (en) * | 2022-09-07 | 2023-12-05 | 宁波大学 | Full color-multispectral-hyperspectral integrated fusion method considering breadth difference |
| CN115861141A (en) * | 2022-12-02 | 2023-03-28 | 北京领云时代科技有限公司 | Unmanned aerial vehicle image acquisition processing system and method based on PCNN neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108921809B (en) | 2022-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108921809A (en) | Multispectral and panchromatic image fusion method under integral principle based on spatial frequency | |
| Berman et al. | ICE: A statistical approach to identifying endmembers in hyperspectral images | |
| CN105719263B (en) | Visible ray and infrared image fusion method based on NSCT domains bottom visual signature | |
| CN106846289B (en) | A kind of infrared light intensity and polarization image fusion method | |
| CN110084108A (en) | Pedestrian re-identification system and method based on GAN neural network | |
| CN106503739A (en) | The target in hyperspectral remotely sensed image svm classifier method and system of combined spectral and textural characteristics | |
| CN109191416A (en) | Image interfusion method based on sparse dictionary study and shearing wave | |
| CN105139367A (en) | Visible-light polarization image fusion method based on non-subsampled shearlets | |
| CN108198147A (en) | A kind of method based on the multi-source image fusion denoising for differentiating dictionary learning | |
| CN108765359A (en) | Fusion method of hyperspectral remote sensing image and full-color image based on JSK model and NSCT technology | |
| CN108898065A (en) | Candidate regions quickly screen and the depth network Ship Target Detection method of dimension self-adaption | |
| CN106251368A (en) | SAR image based on BEMD and the fusion method of multispectral image | |
| CN106897986A (en) | A kind of visible images based on multiscale analysis and far infrared image interfusion method | |
| CN108427964B (en) | Method and system for fusing remote sensing image and geochemistry | |
| CN106960182A (en) | A kind of pedestrian integrated based on multiple features recognition methods again | |
| CN110378295A (en) | A kind of coal rock detection method of multispectral image information and the fusion of spectral information isomery | |
| CN111680579B (en) | Remote sensing image classification method for self-adaptive weight multi-view measurement learning | |
| CN110111276A (en) | Based on sky-spectrum information deep exploitation target in hyperspectral remotely sensed image super-resolution method | |
| CN111563866B (en) | Multisource remote sensing image fusion method | |
| Wang et al. | A hyperspectral image NSST-HMF model and its application in HS-pansharpening | |
| CN108846329A (en) | A kind of EO-1 hyperion face identification method based on waveband selection and Fusion Features | |
| Ye et al. | An unsupervised SAR and optical image fusion network based on structure-texture decomposition | |
| CN109146768A (en) | image conversion method, system and application | |
| CN112734683A (en) | Multi-scale SAR and infrared image fusion method based on target enhancement | |
| CN115272873A (en) | Hyperspectral image nonlinear feature preprocessing system and method based on gradient domain |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220218 |