CN107689038A - A kind of image interfusion method based on rarefaction representation and circulation guiding filtering - Google Patents
A kind of image interfusion method based on rarefaction representation and circulation guiding filtering Download PDFInfo
- Publication number
- CN107689038A CN107689038A CN201710724551.6A CN201710724551A CN107689038A CN 107689038 A CN107689038 A CN 107689038A CN 201710724551 A CN201710724551 A CN 201710724551A CN 107689038 A CN107689038 A CN 107689038A
- Authority
- CN
- China
- Prior art keywords
- image
- fusion
- smoothed
- images
- width
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000001914 filtration Methods 0.000 title claims abstract description 32
- 230000004927 fusion Effects 0.000 claims abstract description 65
- 230000008569 process Effects 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 7
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 6
- 238000012545 processing Methods 0.000 abstract description 5
- 238000007500 overflow downdraw method Methods 0.000 abstract description 4
- 230000000717 retained effect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000003706 image smoothing Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000004297 night vision Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
- G06T2207/20024—Filtering details
-
- 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 Analysis (AREA)
- Image Processing (AREA)
Abstract
A kind of image interfusion method based on rarefaction representation and circulation guiding filtering, belongs to image processing field.Source images are decomposed into smoothed image and detail pictures by the present invention, and rarefaction representation and circulation guiding filtering fusion smoothed image and detail pictures is then respectively adopted, is finally added the smoothed image after fusion with detail pictures to obtain fused images.Rarefaction representation has preferable syncretizing effect to the smoothed data of low-rank, and the edge and profile of detail data, the valid data of prominent detail pictures can be retained by circulating guiding filtering so that of the invention compared with traditional fusion method, syncretizing effect is obvious, and picture appraisal parameter is higher.
Description
Technical field
The invention belongs to image processing field, relates generally to image fusion technology, specifically discloses one kind and is based on sparse table
Show (Sparse Representation, SR) and circulate the image of guiding filtering (Rolling Guidance Filter, RGF)
Fusion method.
Background technology
Image co-registration is that two width or multiple image that are shot to same target are synthesized to the process of piece image, after processing
It can obtain having that the advantages of source images, information are more accurate, are more suitable for human-eye visual characteristic or are adapted to the new figure of computer disposal
Picture.Image to be fused each possesses different features and prominent information, therefore is melted some images by certain algorithm
Close, the target information of obtained fused images is more accurate comprehensive, the more conducively analysis and research of view data.
With the continuous development of science and technology, the species and application field of imaging sensor constantly expand, different sensors into
As principle difference, the feature of image of acquisition is also just different, and the view data for causing to obtain both has redundancy, exists again mutual
Benefit property.Relatively reliable result can be obtained using redundancy, and improves signal to noise ratio;It can tie fusion using complementary information
Fruit, which includes, more enriches comprehensive detailed information.Therefore, image fusion technology is exactly the same field got to different sensors
Scape or multiple source images of target are merged according to actual demand, are allowed to the processing for being adapted to visually-perceptible and computer.
Image co-registration is divided into 3 processing levels, is Pixel-level fusion, feature-based fusion and decision level fusion respectively.Pixel
The raw information of original image has directly been merged in level fusion, can keep the trickle information content of image well, and fusion purpose is
Multiple input pictures are fused to an image, it is more beneficial for the mankind or machine perception compared to any input picture;
These advantages cause it to have important application value in remote sensing, medical imaging and night vision application.Image interfusion method includes
The conversion of three Main Stages, i.e. image, the fusion of conversion coefficient, inverse transformation.Varied one's tactics, image can be melted based on used
Conjunction method is divided into four classes:(1) method based on multi-resolution decomposition;(2) method based on rarefaction representation;(3) directly to image slices
The method that element or other transform domains are merged, such as bulk composition space or intensity form and aspect saturation color space;(4) combine a variety of
The method of conversion.In existing fusion method, the average calculating fusion amount of linear weighted function is small, time-consuming short, but can not
Retain the information of source images well;Wavelet transform (Discrete Wavelet Transform, DWT) is by data conversion
For frequency domain, it is easy to calculate, but lacking direction property;Contourlet transformation can catch the inherent geometry of image, can be very well
Ground handles 2D signal, but does not have translation invariance, it is impossible to represents complicated space structure.In recent years, edge preserving filter
It is widely used on image procossing, typical boundary filter has two-sided filter, wave filter etc., and it can be exactly
Grain details, medium scale edge and the large-scale space structure of the small yardstick of separate picture, therefore in image smoothing, denoising
Etc. have good effect.
The content of the invention
The present invention proposes a kind of based on rarefaction representation (Sparse Representation, SR) and circulation guiding filtering
The image interfusion method of (Rolling Guidance Filter, RGF), this method maintain image while smooth background
Edge and profile, remain the detailed information of image well.
Technical scheme is as follows:
A kind of image interfusion method based on rarefaction representation and circulation guiding filtering, comprises the following steps:
Step 1, source images are decomposed into by smoothed image and detail pictures using mean filter;
Step 2, the smoothed image for obtaining step 1 are using the method fusion based on rarefaction representation:
The smoothed image that step 1 is obtained first resolves into a series of images block, and each image block is converted into column vector
And subtracting its column vector average makes average value be classified as 0, pass through OMP (Orthogonal Matching Pursuit, orthogonal matching
Tracking) the obtained vectorial sparse coefficient of Algorithm for Solving, then by " max-L1 " rule fusion sparse coefficient, after fusion
The image block merged after sparse coefficient inverse transformation, according to the smooth figure after being merged after the position grouping of different images block
Picture;
Step 3, the detail pictures for obtaining step 1 use the method fusion based on circulation guiding filtering:
Pending image is built using DoG operators to source images, and circulation guiding filtering is carried out to pending image;Then
Weight coefficient is obtained through normalization, the details after obtained weight coefficient is merged with corresponding detail pictures weighting summation
Image;
Step 4, step 2 is merged after smoothed image merged with step 3 after detail pictures be added after, obtain fusion figure
Picture.
A kind of image interfusion method based on rarefaction representation and circulation guiding filtering, specifically includes following steps:
The decomposition of step 1, source images:
Source images are decomposed into by smoothed image and detail pictures using mean filter, are specially:
Pn=In*Lave
Dn=In-Pn
Wherein, InFor the n-th width source images, LaveFor mean filter, PnFor the smooth figure corresponding with the n-th width source images
Picture, DnFor the detail pictures corresponding with the n-th width source images, n=1,2 ..., N;
The fusion of step 2, smoothed image
2.1 use sliding window technique, according to the order from the upper left corner to the lower right corner by PnIt is decomposed intoSize
Image block, set PnResolve into T image blockWherein, l is study dictionary D width;
2.2 for each image blockIt is converted into column vectorThen column vector is subtractedThe average of middle element
Its average value is classified as 0, obtain
Wherein, 1 be the 1 × l for being all 1 column vector,It is correspondingAll elements average;
2.3 use what OMP algorithms calculation procedure 2.2 obtainedSparse coefficientObject function is as follows:
Wherein, D is study dictionary, and ε is serious forgiveness;
2.4 repeat step 2.1-2.3 method, obtain correspondence image block in N width smoothed imagesSparse coefficientThen using the sparse coefficient of " max-L1 " rule fusion N width smoothed images, obtain corresponding i-th in N width smoothed images
The fusion sparse coefficient of individual image block:
Wherein,The sparse coefficient being calculated for step 2.3,For obtained fusion sparse coefficient;
2.5 according to the fusion sparse coefficient that i-th of image block is corresponded in obtained N width smoothed imagesInverse transformation obtains it
Merge column vector
Wherein, 1 be the 1 × l for being all 1 column vector,For corresponding column vectorAverage;
2.6 are directed to T image blockIt can obtain TBy eachRemold as image blockThen according to not
Position with image block is by TReconfigure, the smoothed image P after being mergedF;
The fusion of step 3, detail pictures
3.1 are directed to each width source images, and pending image S is built using DoG operatorsn:
Sn=abs (In*Gσ1-In*Gσ2)
Wherein, G is Gaussian filter, and σ 1, σ 2 are the Gaussian kernel of various criterion difference, SnFor n-th pending image, n=
1,2,…,N;
3.2 are directed to n-th pending image SnIt is compared to each other to obtain characteristic image Pn, it is specially:
Wherein,The value of pixel k in n-th pending image is represented,Represent pixel k in the n-th width characteristic image
Value, k=1,2 ..., K, K is pending image SnWith characteristic image PnIn pixel total number, PnIt is pending for n-th
Characteristic image corresponding to image, n=1,2 ..., N, N represent the quantity of source images;
The characteristic image P that 3.3 pairs of steps 3.2 obtainnCirculation guiding filtering is carried out, and is normalized:
Rn=RGD (Pn)
Wherein, RenFor the weight map picture finally obtained, RGF represents circulation guiding filtering operation;
3.4 are directed to N width source images, repeat 3.1-3.3 process, obtain N number of weight map pictureThen will obtain
Weight map merge as corresponding detail pictures, the detail pictures after being merged:
Smoothed image P after step 4, the fusion for obtaining step 2FDetail pictures D after the fusion obtained with step 3FPhase
Fused images are obtained after adding:
F=PF+DF
Beneficial effects of the present invention are:
Source images are decomposed into smoothed image and detail pictures by the present invention, and rarefaction representation is then respectively adopted and circulation guides
Smoothed image after fusion, is finally added to obtain fused images by filtering fusion smoothed image and detail pictures with detail pictures.
Rarefaction representation has preferable syncretizing effect to the smoothed data of low-rank, and the side of detail data can be retained by circulating guiding filtering
Edge and profile, the valid data of prominent detail pictures so that it is of the invention that syncretizing effect is obvious compared with traditional fusion method,
Picture appraisal parameter is higher.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of image interfusion method of the present invention based on rarefaction representation and circulation guiding filtering;
Fig. 2 is source images decomposable process schematic diagram of the present invention;
Fig. 3 is the fusion process schematic diagram of smoothed image in the present invention;
Fig. 4 is the fusion process schematic diagram of detail pictures in the present invention;
Fig. 5 is multi-focus image fusion schematic diagram;Wherein, (a) is rear focused view, and (b) is front focused view, and (c) is
Image after fusion;
Fig. 6 is Medical image fusion schematic diagram;Wherein, (a) is CT images, and (b) is MR images, and (c) is the figure after fusion
Picture;
Fig. 7 is that visible ray-infrared image merges schematic diagram;Wherein, (a) is visible images, and (b) is infrared image, (c)
For the image after fusion;
Embodiment
With reference to the accompanying drawings and examples, technical scheme is described in detail.
A kind of image interfusion method based on rarefaction representation and circulation guiding filtering, specifically includes following steps:
The decomposition of step 1, source images:
N source images to be fused, it is to be clapped under synchronization by the sensor of different imaging characteristicses for same target
Take the photograph what is obtained.Filtered for each width source images in N width source images using mean filter, the result for filtering to obtain is smooth
What image, source images and smoothed image subtracted each other to obtain is detail pictures.Specially:
Pn=In*Lave
Dn=In-Pn
Wherein, InFor the n-th width source images, LaveFor mean filter, PnFor the smooth figure corresponding with the n-th width source images
Picture, DnFor the detail pictures corresponding with the n-th width source images, n=1,2 ..., N.
The fusion of step 2, smoothed image
2.1 use sliding window technique, according to the order from the upper left corner to the lower right corner by PnIt is decomposed intoSize
Image block, set PnResolve into T image blockWherein, l is study dictionary D width, sets the step-length of sliding window
For s pixel, s is less than
2.2 for each image blockIt is converted into column vectorThen column vector is subtractedThe average of middle element makes its average value be classified as 0, obtains
Wherein, 1 be the 1 × l for being all 1 column vector,It is correspondingAll elements average;
2.3 are obtained using OMP (Orthogonal Matching Pursuit, orthogonal matching pursuit) algorithm calculation procedure 2.2
ArriveSparse coefficientObject function is as follows:
Wherein, D is study dictionary, and ε is serious forgiveness;
2.4 repeat step 2.1-2.3 method, obtain corresponding image block in N width smoothed imagesSparse coefficientThen using the sparse coefficient of " max-L1 " rule fusion N width smoothed images, obtain corresponding i-th in N width smoothed images
The fusion sparse coefficient of individual image block:
Wherein,The sparse coefficient being calculated for step 2.3,For obtained fusion sparse coefficient;
2.5 according to the fusion sparse coefficient that i-th of image block is corresponded in obtained N width smoothed imagesInverse transformation obtains it
Merge column vector
Wherein, 1 be the 1 × l for being all 1 column vector,For corresponding column vectorAverage;
2.6 are directed to all image blocksIt can obtain TBy eachRemold as image blockThen root
According to the position of different images block by TReconfigure, the smoothed image P after being mergedF;Wherein, when different images block has
When overlapping, each pixel divided by overlapping number are averaged in overlapping region.
The fusion of step 3, detail pictures
3.1 are directed to each width source images, and pending image S is built using DoG (difference of Gaussian) operatorn:
Sn=abs (In*Gσ1-In*Gσ2)
Wherein, G is Gaussian filter, and σ 1, σ 2 are the Gaussian kernel of various criterion difference, reference value σ 1 used herein, σ 2
Respectively 1 and 0.3, SnFor n-th pending image, n=1,2 ..., N;
3.2 are directed to n-th pending image SnIt is compared to each other to obtain characteristic image Pn, it is specially:
Wherein,The value of pixel k in n-th pending image is represented,Represent pixel in the 1st pending image
Point k value,The value of pixel k in the n-th width characteristic image, k=1,2 ..., K are represented, K is pending image SnAnd characteristic pattern
As PnIn pixel total number, PnFor characteristic image corresponding to n-th pending image, n=1,2 ..., N, N represents source figure
The quantity of picture;
It is compared for the size of the value of the pixel of same position in N pending image, is obtained according to above-mentioned formula
To characteristic image P corresponding to n-th pending imagen;
3.3 circulation guiding filterings have the characteristics of eliminating fine structure and keeping edge simultaneously, by being changed to object use
The two-sided filter in generation completes filtering.The realization of present invention circulation guiding filtering is realized by the iteration of wave filter
, the process for circulating guiding filtering is expressed from the next:
Wherein,
PnCharacteristic image corresponding to the n-th pending image obtained for step 3.2, Pn(m) n-th pending figure is represented
Pixel m value, H in the characteristic image as corresponding totFor the image of the t times iteration, Ht(k) in the image for representing the t times iteration
Pixel k value, Ht+1(k) value of the pixel k in the image of the t+1 times iteration, H are representedt(m) figure of the t times iteration is represented
The value of pixel m as in;M (k) represents pixel k neighborhood, and m represents the point in pixel k fields, σs、σrTo need to set
Parameter, representation space weight and distance weighting, are respectively 3 and 0.05 herein with reference to value respectively;Wherein, initial iterative image
H0It is arranged to and source images size identical full 0 matrix;
The value (k=1,2 ..., K) of all pixels point in the image of the t+1 times iteration is calculated according to above formula, and then is obtained
Image after the t+1 times iteration, the weight map as obtained after circulation guiding filtering operation is as Rn。
To characteristic image PnCirculation guiding filtering is carried out, and normalized process is represented by:
Rn=RGF (Pn)
Wherein, RenFor the weight map picture finally obtained, RGF represents circulation guiding filtering operation;
3.4 are directed to N width source images, repeat 3.1-3.3 process, obtain N number of weight map pictureThen will obtain
Weight map as corresponding detail pictures multiplication Weighted Fusion, the detail pictures after being merged:
Smoothed image P after step 4, the fusion for obtaining step 2FDetail pictures D after the fusion obtained with step 3FIt is inverse
Conversion, fused images are obtained after addition:
F=PF+DF
Embodiment
Merged according to the method described above for two width source images, the image after being merged.Multi-focus figure as shown in Figure 5
As fusion, Medical image fusion shown in Fig. 6, visible ray shown in Fig. 7-infrared image fusion.
The present invention has carried out evaluation index emulation, and and discrete wavelet transformer to the fusion results of accompanying drawing 5, accompanying drawing 6 and accompanying drawing 7
Change (discrete wavelet transform, DWT), warp wavelet (curvelet transform, CVT) is contrasted,
Obtain such as the result of following table one, table two and table three.Wherein, the evaluation index of selection is:
(1)Qmi:A kind of evaluating based on information theory, the mutual information of image can be calculated;
(2)Qy:Retain the integrated degree of original image structural information;
(3)EN:The number of the average information of video source;
(4)SSIM:Structural similarity, it measures image similarity in terms of brightness, contrast, structure three respectively.
Commenting using the inventive method, wavelet transform and warp wavelet when table one is 5 multi-focus image fusion of accompanying drawing
Valency parameter comparison
Using the evaluation of the inventive method, wavelet transform and warp wavelet when table two is 6 Medical image fusion of accompanying drawing
Parameter comparison
Table three uses the inventive method, wavelet transform and warp wavelet when being 7 visible rays of accompanying drawing-infrared image fusion
Evaluating contrast
Be directed to different types of source images it can be seen from table one, two, three, it is proposed by the present invention based on rarefaction representation and
The fused images that the image interfusion method of circulation guiding filtering obtains are respectively provided with excellent performance.Wherein, evaluation index Qy, SSIM
The fusion degree at main measurement picture structure edge, in the fusion results of three kinds of different types of source images, the present invention is most
It is excellent;Evaluation index Qmi, EN mainly measure image carry information content number, the fusion results of three kinds of different types of source images
In, the present invention is mostly in optimal effectiveness, and minority also achieves preferable effect.
Claims (2)
1. a kind of image interfusion method based on rarefaction representation and circulation guiding filtering, comprises the following steps:
Step 1, source images are decomposed into by smoothed image and detail pictures using mean filter;
Step 2, the smoothed image for obtaining step 1 are using the method fusion based on rarefaction representation:
The smoothed image that step 1 is obtained first resolves into a series of images block, and each image block is converted into column vector and subtracted
Go its column vector average average value is classified as 0, the vectorial sparse coefficient obtained by OMP Algorithm for Solving, then pass through " max-
L1 " rule fusion sparse coefficients, the image block that will be merged after the sparse coefficient inverse transformation after fusion, according to different images block
Position grouping after merged after smoothed image;
Step 3, the detail pictures for obtaining step 1 use the method fusion based on circulation guiding filtering:
Pending image is built using DoG operators to source images, and circulation guiding filtering is carried out to pending image;Then through returning
One change obtains weight coefficient, the detail view after obtained weight coefficient is merged with corresponding detail pictures weighting summation
Picture;
Step 4, step 2 is merged after smoothed image merged with step 3 after detail pictures be added after, obtain fused images.
2. a kind of image interfusion method based on rarefaction representation and circulation guiding filtering, specifically includes following steps:
The decomposition of step 1, source images:
Source images are decomposed into by smoothed image and detail pictures using mean filter, are specially:
Pn=In*Lave
Dn=In-Pn
Wherein, InFor the n-th width source images, LaveFor mean filter, PnFor the smoothed image corresponding with the n-th width source images, Dn
For the detail pictures corresponding with the n-th width source images, n=1,2 ..., N;
The fusion of step 2, smoothed image
2.1 use sliding window technique, according to the order from the upper left corner to the lower right corner by PnIt is decomposed intoThe image of size
Block, set PnResolve into T image blockWherein, l is study dictionary D width;
2.2 for each image blockIt is converted into column vectorThen column vector is subtractedThe average of middle element makes it
Average value is classified as 0, obtains
Wherein, 1 be the 1 × l for being all 1 column vector,It is correspondingAll elements average;
2.3 use what OMP algorithms calculation procedure 2.2 obtainedSparse coefficientObject function is as follows:
Wherein, D is study dictionary, and ε is serious forgiveness;
2.4 repeat step 2.1-2.3 method, obtain correspondence image block in N width smoothed imagesSparse coefficientSo
Afterwards using the sparse coefficient of " max-L1 " rule fusion N width smoothed images, corresponding i-th of image block in N width smoothed images is obtained
Fusion sparse coefficient:
Wherein,The sparse coefficient being calculated for step 2.3,For obtained fusion sparse coefficient;
2.5 according to the fusion sparse coefficient that i-th of image block is corresponded in obtained N width smoothed imagesInverse transformation obtains its fusion
Column vector
Wherein, 1 be the 1 × l for being all 1 column vector,For corresponding column vectorAverage;
2.6 are directed to T image blockIt can obtain TBy eachRemold as image blockThen according to different figures
As the position of block is by TReconfigure, the smoothed image P after being mergedF;
The fusion of step 3, detail pictures
3.1 are directed to each width source images, and pending image S is built using DoG operatorsn:
Sn=abs (In*Gσ1-In*Gσ2)
Wherein, G is Gaussian filter, and σ 1, σ 2 are the Gaussian kernel of various criterion difference, SnFor n-th pending image, n=1,
2,…,N;
3.2 are directed to n-th pending image SnIt is compared to each other to obtain characteristic image Pn, it is specially:
Wherein,The value of pixel k in n-th pending image is represented,The value of pixel k in the n-th width characteristic image is represented,
K=1,2 ..., K, K are pending image SnWith characteristic image PnIn pixel total number, PnFor n-th pending image
Corresponding characteristic image, n=1,2 ..., N, N represent the quantity of source images;
The characteristic image P that 3.3 pairs of steps 3.2 obtainnCirculation guiding filtering is carried out, and is normalized:
Rn=RGF (Pn)
Wherein, RenFor the weight map picture finally obtained, RGF represents circulation guiding filtering operation;
3.4 are directed to N width source images, repeat 3.1-3.3 process, obtain N number of weight map pictureThen the weight that will be obtained
The corresponding detail pictures fusion of image, the detail pictures after being merged:
Smoothed image P after step 4, the fusion for obtaining step 2FDetail pictures D after the fusion obtained with step 3FAfter addition
Obtain fused images:
F=PF+DF。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710724551.6A CN107689038A (en) | 2017-08-22 | 2017-08-22 | A kind of image interfusion method based on rarefaction representation and circulation guiding filtering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710724551.6A CN107689038A (en) | 2017-08-22 | 2017-08-22 | A kind of image interfusion method based on rarefaction representation and circulation guiding filtering |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107689038A true CN107689038A (en) | 2018-02-13 |
Family
ID=61153608
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710724551.6A Pending CN107689038A (en) | 2017-08-22 | 2017-08-22 | A kind of image interfusion method based on rarefaction representation and circulation guiding filtering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107689038A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109754384A (en) * | 2018-12-18 | 2019-05-14 | 电子科技大学 | A kind of uncooled ir divides the infrared polarization image interfusion method of focal plane arrays (FPA) |
CN110211080A (en) * | 2019-05-24 | 2019-09-06 | 南昌航空大学 | It is a kind of to dissect and functional medicine image interfusion method |
CN110738677A (en) * | 2019-09-20 | 2020-01-31 | 清华大学 | Full-definition imaging method and device for camera and electronic equipment |
CN110956592A (en) * | 2019-11-14 | 2020-04-03 | 北京达佳互联信息技术有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
CN111833284A (en) * | 2020-07-16 | 2020-10-27 | 昆明理工大学 | Multi-source image fusion method based on low-rank decomposition and convolution sparse coding |
CN115358963A (en) * | 2022-10-19 | 2022-11-18 | 季华实验室 | Image fusion method based on extended Gaussian difference and guided filtering |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104077761A (en) * | 2014-06-26 | 2014-10-01 | 桂林电子科技大学 | Multi-focus image fusion method based on self-adaption sparse representation |
US9152881B2 (en) * | 2012-09-13 | 2015-10-06 | Los Alamos National Security, Llc | Image fusion using sparse overcomplete feature dictionaries |
CN106447640A (en) * | 2016-08-26 | 2017-02-22 | 西安电子科技大学 | Multi-focus image fusion method based on dictionary learning and rotating guided filtering and multi-focus image fusion device thereof |
CN106886986A (en) * | 2016-08-31 | 2017-06-23 | 电子科技大学 | Image interfusion method based on the study of self adaptation group structure sparse dictionary |
-
2017
- 2017-08-22 CN CN201710724551.6A patent/CN107689038A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9152881B2 (en) * | 2012-09-13 | 2015-10-06 | Los Alamos National Security, Llc | Image fusion using sparse overcomplete feature dictionaries |
CN104077761A (en) * | 2014-06-26 | 2014-10-01 | 桂林电子科技大学 | Multi-focus image fusion method based on self-adaption sparse representation |
CN106447640A (en) * | 2016-08-26 | 2017-02-22 | 西安电子科技大学 | Multi-focus image fusion method based on dictionary learning and rotating guided filtering and multi-focus image fusion device thereof |
CN106886986A (en) * | 2016-08-31 | 2017-06-23 | 电子科技大学 | Image interfusion method based on the study of self adaptation group structure sparse dictionary |
Non-Patent Citations (2)
Title |
---|
QI ZHANG ET AL: ""Rolling Guidance Filter"", 《EUROPEAN CONFERENCE ON COMPUTER VISION》 * |
SHUTAO LI ET AL: ""Image Fusion with Guided Filtering"", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109754384A (en) * | 2018-12-18 | 2019-05-14 | 电子科技大学 | A kind of uncooled ir divides the infrared polarization image interfusion method of focal plane arrays (FPA) |
CN109754384B (en) * | 2018-12-18 | 2022-11-22 | 电子科技大学 | Infrared polarization image fusion method of uncooled infrared focal plane array |
CN110211080A (en) * | 2019-05-24 | 2019-09-06 | 南昌航空大学 | It is a kind of to dissect and functional medicine image interfusion method |
CN110738677A (en) * | 2019-09-20 | 2020-01-31 | 清华大学 | Full-definition imaging method and device for camera and electronic equipment |
CN110956592A (en) * | 2019-11-14 | 2020-04-03 | 北京达佳互联信息技术有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
CN111833284A (en) * | 2020-07-16 | 2020-10-27 | 昆明理工大学 | Multi-source image fusion method based on low-rank decomposition and convolution sparse coding |
CN111833284B (en) * | 2020-07-16 | 2022-10-14 | 昆明理工大学 | Multi-source image fusion method based on low-rank decomposition and convolution sparse coding |
CN115358963A (en) * | 2022-10-19 | 2022-11-18 | 季华实验室 | Image fusion method based on extended Gaussian difference and guided filtering |
CN115358963B (en) * | 2022-10-19 | 2022-12-27 | 季华实验室 | Image fusion method based on extended Gaussian difference and guided filtering |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107689038A (en) | A kind of image interfusion method based on rarefaction representation and circulation guiding filtering | |
CN104809734B (en) | A method of the infrared image based on guiding filtering and visual image fusion | |
CN106204447A (en) | The super resolution ratio reconstruction method with convolutional neural networks is divided based on total variance | |
CN107451984A (en) | A kind of infrared and visual image fusion algorithm based on mixing multiscale analysis | |
CN103942769B (en) | A kind of satellite remote-sensing image fusion method | |
CN106709948A (en) | Quick binocular stereo matching method based on superpixel segmentation | |
CN103854267A (en) | Image fusion and super-resolution achievement method based on variation and fractional order differential | |
CN104008533B (en) | Multisensor Image Fusion Scheme based on block adaptive signature tracking | |
CN110097617B (en) | Image fusion method based on convolutional neural network and significance weight | |
Zhang et al. | A multi-modal image fusion framework based on guided filter and sparse representation | |
CN107977951A (en) | The multispectral and hyperspectral image fusion method decomposed based on Coupling Tensor | |
CN107274360B (en) | A kind of high spectrum image denoising method based on Fisher dictionary learning, low-rank representation | |
CN102968781A (en) | Image fusion method based on NSCT (Non Subsampled Contourlet Transform) and sparse representation | |
CN102855616A (en) | Image fusion method based on multi-scale dictionary learning | |
CN112669249A (en) | Infrared and visible light image fusion method combining improved NSCT (non-subsampled Contourlet transform) transformation and deep learning | |
Bhataria et al. | A review of image fusion techniques | |
CN101052993A (en) | Multi-scale filter synthesis for medical image registration | |
CN104217430B (en) | Image significance detection method based on L1 regularization | |
CN105809650A (en) | Bidirectional iteration optimization based image integrating method | |
Zhang et al. | Infrared small target detection based on gradient correlation filtering and contrast measurement | |
CN105354798B (en) | SAR image denoising method based on geometry priori and dispersion similarity measure | |
CN110060220A (en) | Based on the image de-noising method and system for improving BM3D algorithm | |
CN106971382B (en) | A kind of SAR image speckle suppression method | |
CN105303538A (en) | Gauss noise variance estimation method based on NSCT and PCA | |
CN105528772B (en) | A kind of image interfusion method based on directiveness filtering |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180213 |