CN104463822A - Multi-focus image fusing method and device based on multi-scale overall filtering - Google Patents

Abstract

The invention discloses a multi-focus image fusing method based on multi-scale overall filtering. Multi-scale decomposition is carried out on a plurality of multi-focus images to be fused according to the multi-scale overall filtering, a plurality of multi-scale sub-band images of the obtained multi-focus images are fused and decomposed according to a fusion rule, and inverse multi-scale overall filtering is carried out on the fused multi-scale sub-band images to obtain a final fused image. The invention further discloses a multi-focus image fusing device based on multi-scale overall filtering, the definition, the information amount and the like of the fused image are improved, and the fused image with the higher image quality can be obtained.

Description

Based on multi-focus image fusing method and the device thereof of multiple dimensioned global filtering

Technical field

The invention belongs to image co-registration processing technology field, be specifically related to a kind of multi-focus image fusing method based on multiple dimensioned global filtering and device thereof.

Background technology

Image co-registration has great significance in graphical analysis and computer vision.The image organic assembling obtaining Same Scene from different imaging sensor can be become piece image by image fusion technology, its can effectively complementary different imaging sensor obtain the advantage of image, forming a width can the image of true clear reflection objective scene, to analyze further image, understand and the detection and indentification etc. of target.From the 1980s, Multi-sensor Image Fusion has caused interest and research boom widely.Wherein, multi-focus image fusion is an important research direction, has a wide range of applications in machine learning, remote sensing, computer vision, Medical Image Processing and Military Application.Through the development of nearly 30 years, image fusion technology reached a certain scale, also all developed many emerging systems for different field both at home and abroad, but this does not also mean that image fusion technology is quite ripe.From research conditions current both at home and abroad, image fusion technology all has problem to be solved at theoretical and technical elements.Compared with abroad, domesticly carry out image co-registration research work and start late, although domestic research in recent years in image co-registration achieves larger achievement; But, compare the state being still in and comparatively falling behind abroad.Therefore, we carry out basic theory extensively and profoundly and basic technology research to image co-registration is badly in need of.

Along with the fast development of infotech, in practical application, the demand of people to quantity of information is increasing.Under these conditions, typical image interfusion method, as the image interfusion method based on multiresolution analysis, see document " Image sequence fusion using a shift-invariant wavelet transform ", ImageProcessing, 1997.Proceedings., International Conference on.IEEE, 1997,3:288-291, because wavelet transformation can not catch edge and the texture information of image well, and, the simple absolute coefficient of the method gets large fusion rule, and the fused images effect obtained is undesirable, see document " Feature level fusion ofmultimodal medical images in lifting wavelet transformdomain ", Engineering in Medicine and Biology Society, 2004.IEMBS'04.26thAnnual International Conference ofthe IEEE.IEEE, 2004, 1:1479-1482, the method is by calculating the gradient of wavelet conversion coefficient, and determine fusion coefficients by the size of the difference of wavelet conversion coefficient gradient and the threshold value of setting that compare two width images, although the fused images effect that the method obtains makes moderate progress, but, syncretizing effect is still not ideal enough.In recent years, some scholars also been proposed the method adopting image filtering theory to merge image, see document " The multiscale directional bilateral filterand its application to multisensor image fusion ", Information Fusion, 2012,13 (3): 196-206. the method add directivity effectively to extract image information by multiple dimensioned bilateral filtering, obtain good syncretizing effect, there is higher spatial resolution and contrast; See article " Image fusionbased on pixel significance using cross bilateral filter ", Signal, Image and VideoProcessing, 2013:1-12. the filtering core of the method to bilateral filtering exchanges, employing field window statistical property obtains and merges weight, adopt weighted average method to obtain fused images, its quantity of information, contrast, spatial resolution etc. all increase; But detailed information, sharpness etc. are still difficult to meet request for utilization, the whole structure of fused images or not fully up to expectations.

Summary of the invention

For solving the technical matters of existing existence, the embodiment of the present invention provides a kind of multi-focus image fusing method based on multiple dimensioned global filtering and device thereof.

For achieving the above object, the technical scheme of the embodiment of the present invention is achieved in that

The embodiment of the present invention provides a kind of multi-focus image fusing method based on multiple dimensioned global filtering, this fusion method is: carry out multi-resolution decomposition to some width multiple focussing images to be fused respectively according to multiple dimensioned global filtering, merge the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule, carry out obtaining final fused images against multiple dimensioned global filtering to the multiple dimensioned sub-band images after merging.

In such scheme, describedly respectively multi-resolution decomposition is carried out to some width multiple focussing images to be fused according to multiple dimensioned global filtering and be: carry out multiple dimensioned global filtering decomposition according to the two width multiple focussing images of multiple dimensioned global filtering to input, obtain the low frequency sub-band image of two width multiple focussing images and high-frequency sub-band images wherein, K=5, K represent Decomposition order, and m represents the m layer of decomposition, and A represents the first width multiple focussing image, and B represents the second width multiple focussing image.

In such scheme, the described multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained that merges according to fusion rule is: to the low frequency sub-band coefficient of the two width multiple focussing images decomposed through multiple dimensioned global filtering merge, and to high-frequency sub-band images $\{H_m^A,m=\mathrm{1,2},...,K\},\{H_m^B,m=\mathrm{1,2},...,K\}$ Merge.

The embodiment of the present invention also provides a kind of multi-focus image fusion device based on multiple dimensioned global filtering, and this device comprises resolving cell, integrated unit, inverse multiple dimensioned global filtering unit,

Described resolving cell, for carrying out multi-resolution decomposition to some width multiple focussing images to be fused respectively according to multiple dimensioned global filtering;

Described integrated unit, for merging the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule;

Described inverse multiple dimensioned global filtering unit, obtains final fused images for carrying out inverse multiple dimensioned global filtering to the multiple dimensioned sub-band images after fusion.

In such scheme, described resolving cell, specifically for carrying out multiple dimensioned global filtering decomposition according to the two width multiple focussing images of multiple dimensioned global filtering to input, obtains the low frequency sub-band image of two width multiple focussing images and high-frequency sub-band images $\{H_m^A,m=\mathrm{1,2},...,K\},\{H_m^B,m=\mathrm{1,2},...,K\},$ Wherein, K=5, K represent Decomposition order, and m represents the m layer of decomposition, and A represents the first width multiple focussing image, and B represents the second width multiple focussing image.

In such scheme, described integrated unit, specifically for the low frequency sub-band coefficient to the two width multiple focussing images decomposed through multiple dimensioned global filtering merge, and to high-frequency sub-band images $\{H_m^A,m=\mathrm{1,2},...,K\},\{H_m^B,m=\mathrm{1,2},...,K\}$ Merge.

Compared with prior art, the present invention carries out multi-resolution decomposition to some width multiple focussing images to be fused respectively according to multiple dimensioned global filtering, merge the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule, carry out obtaining final fused images against multiple dimensioned global filtering to the multiple dimensioned sub-band images after merging; The present invention, by carrying out repeatedly filtering to multiple focussing image, obtains the high-frequency sub-band images on low frequency sub-band image and different scale; It has taken into full account the Space Consistency of image in decomposable process.Adopt top-hat conversion and higher order statistical characteristic to obtain the fusion weight of low frequency sub-band image and high-frequency sub-band images respectively respectively, the method increase the sharpness of fused images, quantity of information etc. to obtain the better fused images of picture quality; The important information that multiscalization can catch image is better carried out to global filtering.

Accompanying drawing explanation

Fig. 1 is overall flow figure of the present invention;

Fig. 2 is the source images of two groups of multiple focussing images that the present invention uses;

Fig. 3 is the result figure that the present invention and existing four kinds of image interfusion methods merge multi-focus bottle image;

Fig. 4 is the result figure that the present invention and existing four kinds of image interfusion methods merge multi-focus pepsi image;

Fig. 5 is that device of the present invention connects block diagram.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

The invention provides a kind of multi-focus image fusing method based on multiple dimensioned global filtering, respectively multi-resolution decomposition is carried out to some width multiple focussing images to be fused according to multiple dimensioned global filtering, merge the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule, carry out obtaining final fused images against multiple dimensioned global filtering to the multiple dimensioned sub-band images after merging.

The embodiment of the present invention provides a kind of multi-focus image fusing method based on multiple dimensioned global filtering, and as shown in Figure 1, the method is realized by following steps:

Step 101: respectively multi-resolution decomposition is carried out to some width multiple focussing images to be fused according to multiple dimensioned global filtering.

Concrete, carry out multiple dimensioned global filtering decomposition according to the two width multiple focussing images of multiple dimensioned global filtering to input, obtain the low frequency sub-band image of two width multiple focussing images and high-frequency sub-band images wherein, K=5, K represent Decomposition order, and m represents the m layer of decomposition, and A represents the first width multiple focussing image, and B represents the second width multiple focussing image.

Described according to multiple dimensioned global filtering to input two width multiple focussing images carry out multiple dimensioned global filtering decomposition, obtain the low frequency sub-band image of two width multiple focussing images and high-frequency sub-band images $\{H_m^A,m=\mathrm{1,2},...,K\},\{H_m^B,m=\mathrm{1,2},...,K\},$ Be specially:

For image A and image B, formula (1) and formula (2) is utilized to obtain filtering image sequence:

$A_{m+1}=W_m^A{A}_m---\left(1\right)$

$B_{m+1}=W_m^B{B}_m---\left(2\right)$

Wherein, when m gets 1, i.e. A ₁with B ₁represent original image A and B respectively, the wave filter of global filtering $W_m^A=\left[\begin{array}{c}{w}_{m,1}^T\\ w_{m,2}^T\\ \·\\ \·\\ \·\\ w_{m,n}^T\end{array}\right],W_m^B=\left[\begin{array}{c}{\mathrm{\λ}}_{m,1}^T\\ {\mathrm{\λ}}_{m,2}^T\\ \·\\ \·\\ \·\\ {\mathrm{\λ}}_{m,n}^T\end{array}\right]$

$w_i=\frac{1}{{\mathrm{\Σ}}_{j=1}^n{Y}_{\mathrm{ij}}}{[Y_{i1},Y_{i2},...,Y_{\mathrm{in}}]}^T---\left(3\right)$

${\mathrm{\λ}}_i=\frac{1}{{\mathrm{\Σ}}_{j=1}^n{Z}_{\mathrm{ij}}}{[Z_{i1},Z_{i2},...,Z_{\mathrm{in}}]}^T---\left(4\right)$

Wherein, [Y _i1, Y _i2..., Y _in] and [Z _i1, Z _i2..., Z _in] representing i-th row of symmetrical Gaussian nuclear matrix Y and Z respectively, n represents the n-th pixel of image to be filtered.

Obtain the low frequency sub-band coefficient of image A and B, through type (5) and formula (6) just can obtain:

$L_K^A=A_{K+1}---\left(5\right)$

$L_K^B=B_{K+1}---\left(6\right)$

Obtain the high-frequency sub-band coefficient of image A and B, the difference between adjacent yardstick approximate image illustrates the detail pictures of high-frequency sub-band:

$H_m^A=A_{m+1}-A_m---\left(7\right)$

$H_m^B=B_{m+1}-B_m---\left(8\right)$

Wherein, m=1,2 ..., K.

Step 102: merge the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule.

Concrete, to the low frequency sub-band coefficient of the two width multiple focussing images decomposed through multiple dimensioned global filtering merge, and to high-frequency sub-band images $\{H_m^A,m=\mathrm{1,2},...,K\},\{H_m^B,m=\mathrm{1,2},...,K\}$ Merge.

Due to, low frequency sub-band coefficient is the approximate of image, that reflects the energy distribution of original image.Therefore, the primary information resource of image co-registration, in low frequency sub-band coefficient, directly affects the quality of final fusion results on the quality of the selection of the fusion rule of low frequency sub-band image.The present invention obtains the fusion weight of low frequency sub-band coefficient by carrying out top-hat conversion to former multiple focussing image and then merges low frequency sub-band coefficient, to improve the quality of image co-registration.

The low frequency sub-band coefficient of the described two width multiple focussing images to decomposing through multiple dimensioned global filtering merge, be specially:

Step 201: respectively top-hat conversion is carried out to former multiple focussing image A and B, obtains changing image A ^*with B ^*.If utilize the gray scale of structural element U to image f (f represents A and B) to expand, corrode and be expressed as:

$\left[\mathrm{f\⊕U}\right](p,q)=\max \{f(p-a,q-b)+U(a,b)\}---\left(9\right)$

[fΘU](p,q)＝min{f(p+a,q+b)-U(a,b)} (10)

Wherein, (p, q) is the coordinate of image pixel, and (a, b) is the coordinate of pixel in structural element, representing the dilation and erosion operational symbol of mathematical morphology respectively with Θ, is 3 × 3 sized by the structural element chosen.

After defining dilation and erosion computing, opening operation and closed operation can be carried out by structural element U to image f (f represents A and B), specifically be expressed as follows:

$f\·U=(f\⊕U)\mathrm{\ΘU}---\left(12\right)$

Wherein, o with to represent that the opening operation of mathematical morphology accord with and closed operation accords with respectively.

Step 202: obtain low frequency sub-band coefficient with fusion weight, on the basis of the opening operation and closed operation that define mathematical morphology, utilize structural element U to image A and B carry out top-hat conversion just obtain low frequency sub-band coefficient with fusion weight with

Step 203: low frequency sub-band coefficient with fusion:

$L_K(p,q)=\left\{\begin{array}{c}\frac{L_K^A(p,q)+L_K^B(p,q)}{2}\mathrm{if}{\mathrm{\ω}}_K^A(p,q)={\mathrm{\ω}}_K^B(p,q)=0\\ {\mathrm{\ω}}_K^A(p,q)X_K^A(p,q)+{\mathrm{\ω}}_K^B(p,q)X_K^B(p,q)\mathrm{else}\end{array}\right.---\left(15\right)$

Due to the detailed information of the high-frequency sub-band coefficient representative image of image, therefore, the detailed information of fused images is mainly derived from the high-frequency sub-band coefficient of source images, and the quality choosing high-frequency sub-band coefficient fusion rule directly will affect the degree of enriching of final fused images detailed information and the sharpness of image.The present invention, by the eight rank degree of correlation CF adopting higher order statistical characteristic to obtain the high-frequency sub-band coefficient of source images, determines the fusion rule of final high-frequency sub-band coefficient.

To decomposing the high-frequency sub-band coefficient obtained through multiple dimensioned global filtering with merge, be specially:

Step 301: solve degree of correlation CF:

${\mathrm{CF}}_{A,B}=\frac{1}{P\×Q}\frac{{\mathrm{\Σ}}_p^P{\mathrm{\Σ}}_q^Q{(I_1(p,q)-{\mathrm{\μ}}_{I_1})}^4{(I_2(p,q)-{\mathrm{\μ}}_{I_3})}^4}{\sqrt{\left({\mathrm{\Σ}}_p^P{\mathrm{\Σ}}_q^Q{(I_1(p,q)-{\mathrm{\μ}}_{I_1})}^8\right)\left({\mathrm{\Σ}}_p^P{\mathrm{\Σ}}_q^Q{(I_2(p,q)-{\mathrm{\μ}}_{I_2})}^8\right)}}---\left(16\right)$

Wherein, I ₁with I ₂represent respectively with at the gray-scale value of P × Q local neighborhood, with represent respectively with average, (p, q) represents the location of pixels of high-frequency sub-band coefficient, and P × Q represents the size of neighborhood window, and P × Q gets 3 × 3 in the present invention.

Step 302: high-frequency sub-band coefficient $\{H_m^A,m=\mathrm{1,2},...,K\}$ With $\{H_m^B,m=\mathrm{1,2},...,K\}$ Fusion:

If during degree of correlation CF (p, q) > Th, show that the high-frequency sub-band coefficient of two width images is higher in the correlativity at location of pixels (p, q) place, the high-frequency sub-band coefficient of fused images is now expressed as:

$H_m(p,q)=\left\{\begin{array}{c}{H}_m^A(p,q),\mathrm{if}\left|H_m^A(p,q)\right|>\left|H_m^B(p,q)\right|\\ H_m^B(p,q),\mathrm{else}\end{array}\right.---\left(17\right)$

If degree of correlation CF is (p, q), during < Th, show that the high-frequency sub-band coefficient of two width images is lower in the correlativity at location of pixels (p, q) place, in order to preserve the high-frequency information of two width multiple focussing images, the high-frequency sub-band coefficient of fused images is now expressed as:

$H_m(p,q)=H_m^A(p,q)+H_m^B(p,q)---\left(18\right)$

Wherein, with H _m(p, q) represents the high-frequency sub-band coefficient of two width multiple focussing image A, B and fused images F at location of pixels (p, q) place respectively, and Th is default threshold value, gets 0.9 in the present invention.

Step 103: carry out obtaining final fused images against multiple dimensioned global filtering to the multiple dimensioned sub-band images after merging.

Concrete, in order to obtain final fused images F, need to the fusion low frequency sub-band coefficient L obtained in step 102 _k(p, q) and high-frequency sub-band coefficient H _m(p, q) be reconstructed, reconstructing method of the present invention carries out inverse multiple dimensioned global filtering to the multiple dimensioned sub-band coefficients merged, namely to the low frequency sub-band coefficient merged and high-frequency sub-band coefficient be added just fused images F finally, specifically such as formula shown in (19):

F(p,q)＝L _K(p,q)+H _m(p,q) (19)

Effect of the present invention can be illustrated by emulation experiment:

1. experiment condition

The view data adopted in experiment is the multiple focussing image of two groups of registrations, size is 128 × 128, and first group is bottle image, as Fig. 2 (a) and Fig. 2 (b), second group is pepsi image, as Fig. 2 (c) and Fig. 2 (d).

2. experiment content

Experiment 1, by method of the present invention and existing four kinds of fusion methods, fusion experiment is carried out to bottle image, fusion results is as Fig. 3, wherein Fig. 3 (a) is document " Image sequence fusion using ashift-invariantwavelet transform ", Image Processing, 1997.Proceedings., International Conference on.IEEE, 1997, 3:288-291. Fig. 3 (b) is document " Feature levelfusion ofmultimodal medical images in liftingwavelettransform domain ", Engineering inMedicine andBiology Society, 2004.IEMBS'04.26thAnnualInternational Conference ofthe IEEE.IEEE, 2004, 1:1479-1482. Fig. 3 (c) is document " Image fusion based on pixel significance using cross bilateral filter ", Signal, Image and Video Processing, 2013:1-12. Fig. 3 (d) is " The multiscale directionalbilateral filter and its application to multisensor image fusion.Information Fusion ", Information Fusion, 2012, 13 (3): 196-206. Fig. 3 (e) are image co-registration result figure of the present invention.

As seen from Figure 3, fusion method of the present invention is compared with existing four kinds of fusion methods, visual effect is better, letter on two bottles and gear all more clear, document " Image sequence fusion using ashift-invariant wavelet transform ", Image Processing, 1997.Proceedings., International Conference on.IEEE, 1997, 3:288-291, document " Feature level fusion ofmultimodal medical images in lifting wavelet transform domain ", Engineering inMedicine and Biology Society, 2004.IEMBS'04.26thAnnual InternationalConference ofthe IEEE.IEEE, 2004, 1:1479-1482 and document " Image fusion based onpixel significance using cross bilateral filter ", Signal, Image and Video Processing, the fusion method of 2013:1-12. and document " The multiscale directional bilateral filter and itsapplication to multisensor image fusion.Information Fusion ", Information Fusion, 2012, the fusion results of the fusion method of 13 (3): 196-206. is compared with the inventive method, sharpness is lower, detailed information amount is less.

Experiment 2, by method of the present invention and existing four kinds of fusion methods, fusion experiment is carried out to pepsi image, fusion results is as Fig. 4, wherein Fig. 4 (a) is document " Image sequence fusion using a shift-invariantwavelet transform ", Image Processing, 1997.Proceedings.International Conferenceon.IEEE, 1997, 3:288-291. Fig. 4 (b) is document " Feature level fusion ofmultimodalmedical images in liftingwavelet transform domain ", Engineering in Medicine andBiology Society, 2004.IEMBS'04.26thAnnual International Conference oftheIEEE.IEEE, 2004, 1:1479-1482. Fig. 4 (c) is document " Image fusion based on pixelsignificance using cross bilateral filter ", Signal, Image and Video Processing, 2013:1-12. Fig. 4 (d) is document " The multiscale directional bilateral filter and its application tomultisensor image fusion.Information Fusion ", Information Fusion, 2012, 13 (3): 196-206. Fig. 4 (e) are image co-registration result figure of the present invention.

As seen from Figure 4, fusion method of the present invention is compared with existing four kinds of fusion methods, visual effect is better, word in image is more clear, document " Image sequence fusion using a shift-invariantwavelet transform ", Image Processing, 1997.Proceedings., International Conferenceon.IEEE, 1997, 3:288-291, document " Feature level fusion ofmultimodal medicalimages in lifting wavelet transform domain ", Engineering in Medicine and BiologySociety, 2004.IEMBS'04.26thAnnual International Conference ofthe IEEE.IEEE, 2004, 1:1479-1482 and document " Image fusion based on pixel significance using crossbilateral filter ", Signal, Image andVideo Processing, the fusion method of 2013:1-12. and document " The multiscale directional bilateral filter and its application to multisensor imagefusion.Information Fusion ", Information Fusion, 2012, it is lower that fusion results and the inventive method of the fusion method of 13 (3): 196-206. compare sharpness, quantity of information is less.

By fusion method of the present invention and document " Image sequence fusion using a shift-invariantwavelet transform ", Image Processing, 1997.Proceedings., International Conferenceon.IEEE, 1997, 3:288-291. fusion method, document " Feature level fusion ofmultimodalmedical images in liftingwavelet transform domain ", Engineering in Medicine andBiology Society, 2004.IEMBS'04.26thAnnual International Conference oftheIEEE.IEEE, 2004, 1:1479-1482. fusion method, document " Image fusion based on pixelsignificance using cross bilateral filter ", Signal, Image and Video Processing, the fusion method of 2013:1-12. and document " The multiscale directional bilateral filter and itsapplication to multisensor image fusion.Information Fusion ", Information Fusion, 2012, the fusion method of 13 (3): 196-206. compares in four kinds of image quality evaluation indexs, carry out objective evaluation effect of the present invention.Five kinds of fusion methods fusion objective evaluation index on first group of bottle, second group of pepsi multiple focussing image is as table 1 and table 2:

Table 1. first group of multiple focussing image bottle merges objective evaluation index

Method	SD	AG	Entropy	SF
					SWT	65.2281	23.0340	7.6856	34.5349
LWT	66.2794	27.7749	7.6824	41.9111

CBF	66.6634	26.9130	7.6794	41.2268
					MDBF	68.8676	29.4745	7.6255	45.3892
The present invention	71.3053	37.3358	7.5510	56.2755

Table 2. second group of multiple focussing image pepsi merges objective evaluation index

Method	SD	AG	Entropy	SF
					SWT	43.2108	8.4721	6.9742	13.2138
LWT	43.2108	9.4786	6.9834	15.2527
					CBF	43.6270	9.4341	6.9924	15.5338
MDBF	44.6036	10.1409	7.0392	16.7478
					The present invention	45.6118	13.6810	7.2870	22.3411

Table 1 is with table 2:

SWT represents document Rockinger O, " Image sequence fusion using a shift-invariantwavelet transform; " Image Processing, 1997.Proceedings., International Conferenceon.IEEE, the fusion method of 1997,3:288-291..

LWT represents document Kor S, Tiwary U, " Feature level fusion ofmultimodal medicalimages in lifting wavelet transform domain; " Engineering in Medicine and BiologySociety, 2004.IEMBS'04.26thAnnual International Conference ofthe IEEE.IEEE, the fusion method of 2004,1:1479-1482..

CBF represents document Kumar B K S.Image fusion based on pixel significance usingcross bilateral filter.Signal, Image and Video Processing, the image interfusion method of 2013:1-12..

MDBF represents document Hu J, Li S.The multiscale directional bilateral filter and itsapplication to multisensor image fusion.Information Fusion, 2012, the fusion method of 13 (3): 196-206..

SD represents standard deviation, and AG represents average gradient, and Entropy represents information entropy, SF representation space frequency.

From table 1, method of the present invention is obviously better than the method for four sections of above-mentioned documents in three indexs, an other index also relatively other method; And the global index of the inventive method is higher.

From table 2, method of the present invention is obviously better than the method for above-mentioned four sections of documents on four indices.

Above-mentioned experiment proves, the present invention can obtain good visual effect to multi-focus image fusion problem.

The embodiment of the present invention also provides a kind of multi-focus image fusion device based on multiple dimensioned global filtering, and as shown in Figure 5, this device comprises resolving cell 1, integrated unit 2, inverse multiple dimensioned global filtering unit 3,

Described resolving cell 1, for carrying out multi-resolution decomposition to some width multiple focussing images to be fused respectively according to multiple dimensioned global filtering;

Described integrated unit 2, for merging the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule;

Described inverse multiple dimensioned global filtering unit 3, obtains final fused images for carrying out inverse multiple dimensioned global filtering to the multiple dimensioned sub-band images after fusion.

Described resolving cell 1, specifically for carrying out multiple dimensioned global filtering decomposition according to the two width multiple focussing images of multiple dimensioned global filtering to input, obtains the low frequency sub-band image of two width multiple focussing images and high-frequency sub-band images wherein, K=4, K represent Decomposition order, and m represents the m layer of decomposition.

Described integrated unit 2, specifically for the low frequency sub-band coefficient to the two width multiple focussing images decomposed through multiple dimensioned global filtering merge, and to high-frequency sub-band images $\{H_m^B,m=\mathrm{1,2},...,K\}$ Merge.

Content of the present invention is not limited to cited by embodiment, and the conversion of those of ordinary skill in the art by reading instructions of the present invention to any equivalence that technical solution of the present invention is taked, is claim of the present invention and contains.

Claims (6)

1. the multi-focus image fusing method based on multiple dimensioned global filtering, it is characterized in that, this fusion method is: carry out multi-resolution decomposition to some width multiple focussing images to be fused respectively according to multiple dimensioned global filtering, merge the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule, carry out obtaining final fused images against multiple dimensioned global filtering to the multiple dimensioned sub-band images after merging.

2. the multi-focus image fusing method based on multiple dimensioned global filtering according to claim 1, it is characterized in that, describedly respectively multi-resolution decomposition is carried out to some width multiple focussing images to be fused according to multiple dimensioned global filtering and be: carry out multiple dimensioned global filtering decomposition according to the two width multiple focussing images of multiple dimensioned global filtering to input, obtain the low frequency sub-band image of two width multiple focussing images and high-frequency sub-band images wherein, K=5, K represent Decomposition order, and m represents the m layer of decomposition, and A represents the first width multiple focussing image, and B represents the second width multiple focussing image.

3. the multi-focus image fusing method based on multiple dimensioned global filtering according to claim 1, it is characterized in that, the described multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained that merges according to fusion rule is: to the low frequency sub-band coefficient of the two width multiple focussing images decomposed through multiple dimensioned global filtering merge, and to high-frequency sub-band images merge.

4. based on a multi-focus image fusion device for multiple dimensioned global filtering, it is characterized in that, this device comprises resolving cell, integrated unit, inverse multiple dimensioned global filtering unit,

Described resolving cell, for carrying out multi-resolution decomposition to some width multiple focussing images to be fused respectively according to multiple dimensioned global filtering;

Described integrated unit, for merging the multiple dimensioned sub-band images of decomposing the some width multiple focussing images obtained according to fusion rule;

Described inverse multiple dimensioned global filtering unit, obtains final fused images for carrying out inverse multiple dimensioned global filtering to the multiple dimensioned sub-band images after fusion.

5. the multi-focus image fusion device based on multiple dimensioned global filtering according to claim 4, it is characterized in that: described resolving cell, specifically for carrying out multiple dimensioned global filtering decomposition according to the two width multiple focussing images of multiple dimensioned global filtering to input, obtain the low frequency sub-band image of two width multiple focussing images and high-frequency sub-band images wherein, K=5, K represent Decomposition order, and m represents the m layer of decomposition, and A represents the first width multiple focussing image, and B represents the second width multiple focussing image.

6. the multi-focus image fusion device based on multiple dimensioned global filtering according to claim 4, is characterized in that: described integrated unit, specifically for the low frequency sub-band coefficient to the two width multiple focussing images decomposed through multiple dimensioned global filtering merge, and to high-frequency sub-band images merge.