CN1581230A - Remote-senstive image interfusion method based on image local spectrum characteristic - Google Patents
Remote-senstive image interfusion method based on image local spectrum characteristic Download PDFInfo
- Publication number
- CN1581230A CN1581230A CN 200410018480 CN200410018480A CN1581230A CN 1581230 A CN1581230 A CN 1581230A CN 200410018480 CN200410018480 CN 200410018480 CN 200410018480 A CN200410018480 A CN 200410018480A CN 1581230 A CN1581230 A CN 1581230A
- Authority
- CN
- China
- Prior art keywords
- image
- component
- fusion
- wavelet
- local
- 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.)
- Pending
Links
Images
Abstract
The present invention relates to a remote-sensing image fusion method based on image local spectral characteristics. Said method combines local correlation moment and local variance of remote-sensing image to make fusion, on the basis of making multispectral image undergo the process of IHS transformation makes I component obtained after transformation and high-spectral image undergo the process of histogram matching treatment, then makes the matched I component and high-spectral image respectively undergo the process of vavelet transformation. For low-frequency component obtained after wavelet decomposition said method adopts the fusion criterion based on local normalized correlation moment to make fusion, and for high-frequency component it adopts the fusion criterion based on variance to make fusion, then reconstructes all the wavelet components after fusion to obtain new I componnet, finally, uses new I component and original components of H and S to make IHS inverse transformation so as to obtain the fusion result.
Description
Technical field
The present invention relates to a kind of based on image local light spectral property, the method (LCMD method) that merges in conjunction with the local correlation square and the local variance of remote sensing image, be a Multi-Sensory Image Fusion at Pixel Level in the art of image analysis, in fields such as agricultural, military affairs, have a wide range of applications.
Background technology
Development along with remote sensing technology, modern remote sensing system can provide the view data of multiple high spatial resolution, spectral resolution and temporal resolution for the user, and the application of remote sensing technology progressively turns to the analysis-by-synthesis and the application of multiband, multisensor, multi-platform, multidate, many resolving powers data from the analytical applications of single-sensor data.
Obtaining at present of high-resolution satellite image can be carried out with two kinds of different modes: a kind of is the panchromatic mode of high spatial resolution, and another kind is the multispectral mode of high spectral resolution.The feature of panchromatic image is to have very high spatial resolution, and multispectral image provides atural object abundant spectral information.For the enrichment of advantage separately of the basic spectral information of the high spatial resolution of panchromatic image and low resolution multispectral image is got up, this two classes image can be merged, the image that obtains should have higher geological information content, has still kept good spectral information quality simultaneously.
The image and the multispectral image (low resolution) of high spatial resolution are merged, and method representative in this technology has: color space transformation (IHS conversion), PCA (principal component analysis (PCA)) conversion, HPF (high-pass filtering) method, multiresolution wavelet analysis etc.
The multiresolution wavelet analysis is a kind of new time domain/frequency-region signal analysis tool, it can so just can make wavelet coefficient and the corresponding relation of original image content aspect space and frequency field two after people find conversion at an easy rate to the signal indication of signal decomposition on the low resolution level more.Panchromatic image and multispectral image merge, normally to a kind of balance from the spatial information of the spectral information of low spatial-high spectral resolution sensor and low spectrum-high spatial resolution sensor, and, then can control this balance with comparalive ease by the progression of adjusting wavelet decomposition for wavelet transform fusion.In recent years, the multiresolution wavelet analytical approach has been widely used among the fusion of multi-sensor image data, and this method has kept the spectral information of former multispectral image to greatest extent.
The general 2-d wavelet that adopts in remote sensing image is handled.Image merges purpose according to difference after wavelet decomposition, take specific blending algorithm to handle respectively to wavelet decomposition subband and base band, and last inverse transformation promptly gets merges image.
In fusion process, the fusion criterion of being taked will determine the resultant quality that merges image, general criterion have based on pixel with based on the zone.Criterion based on pixel is the maximal value of getting corresponding point gray scale absolute value; Fusion criterion based on local variance is: as in the figure A image wavelet coefficient image certain a bit around variance in the window more than or equal to the variance in the window around the corresponding point in the B image wavelet coefficient image, then the wavelet coefficient image of this grade fusion is the wavelet coefficient of image A at this point value, otherwise is the wavelet coefficient of image B.More than 2 kinds of criterions all can be described as maximal criterion.In fusion process, also have to adopt based on partial gradient with based on the fusion criterion of local energy, and the fusion criterion of taking different weights to merge in different base band.
Maximal value criterion and partial gradient criterion because of being subjected to the influence of isolated noise point easily, thereby may cause selecting wrong coefficient of dissociation when choosing wavelet coefficient.For local variance criterion and energy criteria, all be again that the coefficient of dissociation in the local window is done whole consideration, and consider not enough the independent information that each pixel comprised in the window.
Summary of the invention
The objective of the invention is to deficiency, a kind of remote sensing image fusing method based on image local light spectral property is provided, when improving the multispectral image spatial resolution, keep the spectral information of multispectral image better at the prior art existence.
For realizing such purpose, the present invention is carrying out multispectral image on the basis of IHS conversion, and I component and the Hyperspectral imaging that obtains after the conversion carried out the histogram coupling, then I component and Hyperspectral imaging after the coupling is carried out wavelet transformation respectively.In order to protect the spectral information of multispectral image; adopt the relevant square of local variance to merge respectively to high and low frequency with local normalization; each Wavelet Component after the fusion is rebuild and is obtained new I component; carry out the IHS inverse transformation with new I component and original H, S component again, fusion results to the end.
Remote sensing image fusion based on image local light spectral property of the present invention comprises following concrete steps:
1. multispectral image is carried out the IHS color space transformation, obtain strength component I, chromatic component H, the saturation degree component S of multispectral image, keep H, the S component is constant.
2. the I component that obtains after the IHS conversion and Hyperspectral imaging are carried out the histogram coupling: calculating earlier the histogram of multispectral image I component and Hyperspectral imaging respectively, is reference with the histogram of multispectral image I component, carries out the histogram coupling.
3. I component and Hyperspectral imaging after the coupling are carried out wavelet transformation respectively, obtain the low frequency component and the high fdrequency component of multispectral image I component and Hyperspectral imaging.
4. the low frequency component after the wavelet decomposition is taked to merge based on the relevant square fusion criterion of local normalization: adopt the 3*3 window, calculate the relevant square of local normalization of multispectral image I component and Hyperspectral imaging low frequency component respectively.Fusion rule is: as in the piece image wavelet coefficient image certain a bit around the relevant square of local normalization in the window more than or equal to the relevant square of the local normalization in the window around the corresponding point in another width of cloth image wavelet coefficient image, then the wavelet coefficient image of this grade fusion is this width of cloth wavelet subband coefficients of images just at this point value, otherwise is another width of cloth wavelet subband coefficients of images.
5. the high fdrequency component after the wavelet decomposition is taked to merge based on the variance fusion criterion: adopt the 3*3 window, calculate the local variance of multispectral image I component and Hyperspectral imaging high fdrequency component respectively.Take fusion criterion in the fusion process based on variance: as in the piece image wavelet coefficient image certain a bit around variance in the window more than or equal to the variance in the window around the corresponding point in another width of cloth image wavelet coefficient image, then the wavelet coefficient image of this grade fusion is this width of cloth wavelet subband coefficients of images at this point value, otherwise is another width of cloth wavelet subband coefficients of images.
6. low frequency component and high fdrequency component after merging are rebuild, obtain new I component.
7. the H that obtains with new I component and step 1, S component carry out the IHS inverse transformation, obtain fusion results.
Image interfusion method of the present invention has following beneficial effect:
To the panchromatic image of high spatial resolution and multispectral image (low resolution) when merging, to the fusion criterion of the employing of the low frequency component behind the wavelet transformation based on the relevant square of local normalization, to the fusion criterion of high fdrequency component employing, merge based on local variance.The image that merges when well keeping the spectral information of former multispectral image, has improved the resolution of former multispectral image.
Description of drawings
Fig. 1 uses respectively based on pixel for the set of diagrams picture and chooses method, chooses the contrast figure of method and LCMD method fusion results of the present invention based on the zone.
Among the figure, (a) and (b) be respectively original color image and original full-colour image; (c), (d), (e) be respectively based on pixel and choose method, choose method and LCMD method fusion results of the present invention based on the zone.
Embodiment
In order to understand technical scheme of the present invention better, below embodiments of the present invention are further described.The concrete implementation detail of each several part is as follows:
1. multispectral image is carried out the IHS conversion:
With following formula multispectral image is handled, multispectral image is transformed to the IHS space, obtain three components of I, H, S of multispectral image;
H=tan wherein
-1(V
2/ V
1),
I is a strength component, and H is a chromatic component, and S is the saturation degree component; V
1, V
2It is intermediate variable.
2. I component and the Hyperspectral imaging that obtains after the conversion carried out the histogram coupling:
Specific practice is: calculating earlier the histogram of multispectral image I component and Hyperspectral imaging respectively, is reference with the histogram of multispectral image I component, carries out the histogram coupling;
3. I component and Hyperspectral imaging after the coupling are carried out wavelet transformation respectively, obtain the low frequency component and the high fdrequency component of multispectral image I component and Hyperspectral imaging;
4. the low frequency component after the wavelet decomposition is taked to merge based on the relevant square fusion criterion of local normalization:
Specific practice is: adopt the 3*3 window, calculate the relevant square of multispectral image I component and the relevant square of Hyperspectral imaging low frequency component respectively according to following formula:
Wherein,
The wavelet coefficient values of j interior i the pixel correspondence of window in the little wave image that expression is compared, μ
jThe average of representing wavelet coefficient in j the window, σ
jThe standard deviation of representing the corresponding wavelet coefficient of all pixels in j the window, k=1,2,3 is respectively horizontal high fdrequency component coefficient, vertical high frequency component coefficient and diagonal angle high fdrequency component coefficient symbols.
When low frequency component is merged, take fusion criterion based on the relevant square of local normalization: as in the figure A image wavelet coefficient image certain a bit around the relevant square of normalization in the window more than or equal to the relevant square of the normalization in the window around the corresponding point in the B image wavelet coefficient image, then the wavelet coefficient image of this grade fusion is the wavelet coefficient of image A at this point value, otherwise is the wavelet coefficient of image B.
5. the high fdrequency component on all directions after the wavelet decomposition is taked fusion criterion based on local variance, merges:
Specific practice is: adopt the 3*3 window, calculate the local variance of multispectral image I component and Hyperspectral imaging high fdrequency component respectively.Take fusion criterion in the fusion process based on variance: as in the figure A image wavelet coefficient image certain a bit around variance in the window more than or equal to the variance in the window around the corresponding point in the B image wavelet coefficient image, then the wavelet coefficient image of this grade fusion is the wavelet coefficient of image A at this point value, otherwise is the wavelet coefficient of image B.
6. low frequency component and high fdrequency component after merging are rebuild, obtain new I component.
7. the H that obtains with new I component and step 1, S component carry out the IHS inverse transformation, obtain fusion results.
According to experimental result following conclusion is arranged: aspect the protection local light spectrum information, syncretizing effect is relatively good based on the fusion criterion of relevant square; And aspect the guard space object information, syncretizing effect is not so good as based on the local variance criterion.This mainly be because: high-resolution remote sensing image information on grain details is abundanter, and the level of detail of image mainly is contained in its radio-frequency component.And the spectral information of multispectral image is abundanter, mainly concentrates in the low-frequency component.According to the image spectrum analysis, the image that areal is dissimilar, the low frequency part difference is little, and radio-frequency component differs greatly.Because two width of cloth image low frequency part to be merged are very approaching; low frequency component is adopted the local variance fusion criterion; whole window is merged object as one consider that the difference of two width of cloth image low frequency components can not obtain careful embodiment, spectral information can not well be protected.And the HFS difference of image to be merged is very big; adopt relevant square fusion criterion; given prominence to independently information of each pixel, but the aspects such as protection of larger-size texture information are just not as the local variance criterion in the continuity of considering texture information, to image.
Therefore when merging,,, merge the fusion criterion of high fdrequency component employing based on local variance to the fusion criterion of low frequency component employing based on the relevant square of local normalization.The image that merges when well keeping the spectral information of former multispectral image, has improved the resolution of former multispectral image.
The 3 groups of fused images that provided table 1 adopt respectively based on pixel and choose method, choose the comparison of method and LCMD method fusion results of the present invention based on the zone.
Table 1 is chosen method, is chosen the comparison of method and LCMD method fusion results based on the zone based on pixel
Average gradient | Mutual information | ||
First group of image | Original multispectral image (I component) | ????10.991 | |
Choose method (fused image I component) based on pixel | ????26.858 | ??1.6882 | |
Choose method (fused image I component) based on the zone | ????32.297 | ??2.7135 | |
LCMD method (fused image I component) | ????39.833 | ??3.1611 | |
Second group of image | Original multispectral image (I component) | ????11.751 | |
Choose method (fused image I component) based on pixel | ????28.349 | ??2.0382 | |
Choose method (fused image I component) based on the zone | ????33.921 | ??3.1137 | |
LCMD method (fused image I component) | ????39.787 | ??3.5151 | |
The 3rd group of image | Original multispectral image (I component) | ????7.6948 | |
Choose method (fused image I component) based on pixel | ????14.429 | ??2.9998 | |
Choose method (fused image I component) based on the zone | ????16.102 | ??3.6132 | |
LCMD method (fused image I component) | ????19.823 | ??3.9414 |
As can be seen from Table 1 for the average gradient index, the fusion results of three groups of images shows, the fusion results average gradient maximum of LCMD method of the present invention, illustrate the LCMD method improve aspect the image definition effect than choose method based on pixel, to choose method based on the zone better.And for the mutual information index, the fusion results mutual information maximum of LCMD method, illustrate the LCMD method keep original image message context effect than choose method based on pixel, to choose method based on the zone good.
Claims (1)
1, a kind of remote sensing image fusing method based on image local light spectral property is characterized in that merging in conjunction with the local correlation square and the local variance of remote sensing image, specifically comprises the steps:
1) multispectral image is carried out the IHS color space transformation, obtain strength component I, chromatic component H, the saturation degree component S of multispectral image, keep H, the S component is constant;
2) I component that obtains after the IHS conversion and Hyperspectral imaging are carried out the histogram coupling: calculating earlier the histogram of multispectral image I component and Hyperspectral imaging respectively, is reference with the histogram of multispectral image I component, carries out the histogram coupling;
3) I component and Hyperspectral imaging after the coupling are carried out wavelet transformation respectively, obtain the low frequency component and the high fdrequency component of multispectral image I component and Hyperspectral imaging;
4) low frequency component after the wavelet decomposition is taked to merge based on the relevant square fusion criterion of local normalization: adopt the 3*3 window, calculate the relevant square of local normalization of multispectral image I component and Hyperspectral imaging low frequency component respectively, fusion rule is: as in the piece image wavelet coefficient image certain a bit around the relevant square of local normalization in the window more than or equal to the relevant square of the local normalization in the window around the corresponding point in another width of cloth image wavelet coefficient image, then the wavelet coefficient image of this grade fusion is this width of cloth wavelet subband coefficients of images just at this point value, otherwise is another width of cloth wavelet subband coefficients of images;
5) high fdrequency component after the wavelet decomposition is taked to merge based on the variance fusion criterion: adopt the 3*3 window, calculate the local variance of multispectral image I component and Hyperspectral imaging high fdrequency component respectively, take fusion criterion in the fusion process based on variance: as in the piece image wavelet coefficient image certain a bit around variance in the window more than or equal to the variance in the window around the corresponding point in another width of cloth image wavelet coefficient image, then the wavelet coefficient image of this grade fusion is this width of cloth wavelet subband coefficients of images at this point value, otherwise is another width of cloth wavelet subband coefficients of images;
6) low frequency component and high fdrequency component after merging are rebuild, obtain new I component;
7) H that obtains with new I component and step 1, S component carry out the IHS inverse transformation, obtain fusion results.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 200410018480 CN1581230A (en) | 2004-05-20 | 2004-05-20 | Remote-senstive image interfusion method based on image local spectrum characteristic |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 200410018480 CN1581230A (en) | 2004-05-20 | 2004-05-20 | Remote-senstive image interfusion method based on image local spectrum characteristic |
Publications (1)
Publication Number | Publication Date |
---|---|
CN1581230A true CN1581230A (en) | 2005-02-16 |
Family
ID=34581801
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 200410018480 Pending CN1581230A (en) | 2004-05-20 | 2004-05-20 | Remote-senstive image interfusion method based on image local spectrum characteristic |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN1581230A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN100365435C (en) * | 2005-12-05 | 2008-01-30 | 牛铮 | Analogue technology for imaging spectrograph remote-sensing image in satellite |
CN100383805C (en) * | 2005-11-03 | 2008-04-23 | 复旦大学 | Method for sorting characters of ground object through interfusion of satellite carried microwave and infrared remote sensing |
CN100410684C (en) * | 2006-02-23 | 2008-08-13 | 复旦大学 | Remote sensing image fusion method based on Bayes linear estimation |
CN101799915A (en) * | 2010-02-26 | 2010-08-11 | 中北大学 | Bicolor medium wave infrared image fusion method |
CN101930604A (en) * | 2010-09-08 | 2010-12-29 | 中国科学院自动化研究所 | Infusion method of full-color image and multi-spectral image based on low-frequency correlation analysis |
CN101540045B (en) * | 2009-03-25 | 2011-07-27 | 湖南大学 | Multi-source image fusion method based on synchronous orthogonal matching pursuit |
CN102176066A (en) * | 2011-01-24 | 2011-09-07 | 南京理工大学 | Target optimal detection spectral coverage imaging detecting method based on narrow band scanning |
CN102254311A (en) * | 2011-06-10 | 2011-11-23 | 中国科学院深圳先进技术研究院 | Method and system for fusing remote sensing images |
CN102063710B (en) * | 2009-11-13 | 2013-01-16 | 中国科学院烟台海岸带研究所 | Method for realizing fusion and enhancement of remote sensing image |
CN103065282A (en) * | 2012-12-27 | 2013-04-24 | 浙江大学 | Image fusion method based on sparse linear system |
CN105469364A (en) * | 2015-10-26 | 2016-04-06 | 厦门理工学院 | Medical image fusion method combined with wavelet transformation domain and spatial domain |
CN109584210A (en) * | 2018-10-30 | 2019-04-05 | 南京理工大学 | Multispectral three-dimensional vein imaging system |
CN114331936A (en) * | 2021-12-24 | 2022-04-12 | 郑州信大先进技术研究院 | Remote sensing image fusion method based on wavelet decomposition and improved IHS algorithm |
-
2004
- 2004-05-20 CN CN 200410018480 patent/CN1581230A/en active Pending
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN100383805C (en) * | 2005-11-03 | 2008-04-23 | 复旦大学 | Method for sorting characters of ground object through interfusion of satellite carried microwave and infrared remote sensing |
CN100365435C (en) * | 2005-12-05 | 2008-01-30 | 牛铮 | Analogue technology for imaging spectrograph remote-sensing image in satellite |
CN100410684C (en) * | 2006-02-23 | 2008-08-13 | 复旦大学 | Remote sensing image fusion method based on Bayes linear estimation |
CN101540045B (en) * | 2009-03-25 | 2011-07-27 | 湖南大学 | Multi-source image fusion method based on synchronous orthogonal matching pursuit |
CN102063710B (en) * | 2009-11-13 | 2013-01-16 | 中国科学院烟台海岸带研究所 | Method for realizing fusion and enhancement of remote sensing image |
CN101799915A (en) * | 2010-02-26 | 2010-08-11 | 中北大学 | Bicolor medium wave infrared image fusion method |
CN101930604B (en) * | 2010-09-08 | 2012-03-28 | 中国科学院自动化研究所 | Infusion method of full-color image and multi-spectral image based on low-frequency correlation analysis |
CN101930604A (en) * | 2010-09-08 | 2010-12-29 | 中国科学院自动化研究所 | Infusion method of full-color image and multi-spectral image based on low-frequency correlation analysis |
CN102176066A (en) * | 2011-01-24 | 2011-09-07 | 南京理工大学 | Target optimal detection spectral coverage imaging detecting method based on narrow band scanning |
CN102176066B (en) * | 2011-01-24 | 2014-09-17 | 南京理工大学 | Target optimal detection spectral coverage imaging detecting method based on narrow band scanning |
CN102254311A (en) * | 2011-06-10 | 2011-11-23 | 中国科学院深圳先进技术研究院 | Method and system for fusing remote sensing images |
CN103065282A (en) * | 2012-12-27 | 2013-04-24 | 浙江大学 | Image fusion method based on sparse linear system |
CN105469364A (en) * | 2015-10-26 | 2016-04-06 | 厦门理工学院 | Medical image fusion method combined with wavelet transformation domain and spatial domain |
CN109584210A (en) * | 2018-10-30 | 2019-04-05 | 南京理工大学 | Multispectral three-dimensional vein imaging system |
CN109584210B (en) * | 2018-10-30 | 2022-08-16 | 南京理工大学 | Multispectral three-dimensional venous imaging system |
CN114331936A (en) * | 2021-12-24 | 2022-04-12 | 郑州信大先进技术研究院 | Remote sensing image fusion method based on wavelet decomposition and improved IHS algorithm |
CN114331936B (en) * | 2021-12-24 | 2024-04-16 | 郑州信大先进技术研究院 | Remote sensing image fusion method based on wavelet decomposition and IHS algorithm improvement |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN1581230A (en) | Remote-senstive image interfusion method based on image local spectrum characteristic | |
US7020335B1 (en) | Methods and apparatus for object recognition and compression | |
CN112991288B (en) | Hyperspectral remote sensing image fusion method based on abundance image sharpening reconstruction | |
CN1244885C (en) | Remote-sensing image mixing method based on local statistical property and colour space transformation | |
CN109685730B (en) | Wavelet denoising method based on self-adaptive non-local mean value | |
Zhou et al. | Method of improved fuzzy contrast combined adaptive threshold in NSCT for medical image enhancement | |
CN109636766A (en) | Polarization differential and intensity image Multiscale Fusion method based on marginal information enhancing | |
CN101835045B (en) | Hi-fidelity remote sensing image compression and resolution ratio enhancement joint treatment method | |
CN1484039A (en) | Image merging method based on inseparable wavelet frame | |
CN1303432C (en) | Remote sensing image picture element and characteristic combination optimizing mixing method | |
Thangaswamy et al. | Developing an efficient technique for satellite image denoising and resolution enhancement for improving classification accuracy | |
George et al. | A survey on full reference image quality assessment algorithms | |
Wang et al. | A fusion algorithm of remote sensing image based on discrete wavelet packet | |
CN1303571C (en) | Image optimum fusing method based on fuzzy integral | |
CN107622479B (en) | Contour wave sub-band self-adaptive detail injection method for multi-spectral-band image panchromatic sharpening | |
Huang et al. | Synthetic aperture radar image compression using tree-structured edge-directed orthogonal wavelet packet transform | |
Yan et al. | Remore Sensing Image Quality Assessment based on the Ratio of Spatial Feature Weighted Mutual Information. | |
CN112734683A (en) | Multi-scale SAR and infrared image fusion method based on target enhancement | |
CN111199530A (en) | Fusion method of SAR image and visible light image | |
CN112508829B (en) | Pan-sharpening method based on shear wave transformation | |
Chai et al. | Self-adaptive image fusion based on multi-resolution decomposition using wavelet packet analysis | |
Xiao et al. | Remote sensing image fusion based on Gaussian mixture model and multiresolution analysis | |
CN113538306B (en) | SAR image and low-resolution optical image multi-image fusion method | |
Ghassemian | Multisensor image fusion by inverse subband coding | |
Choi et al. | Performance analysis of image fusion methods in transform domain |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |