CN105225213B - A kind of Color Image Fusion method based on S PCNN and laplacian pyramid - Google Patents
A kind of Color Image Fusion method based on S PCNN and laplacian pyramid Download PDFInfo
- Publication number
- CN105225213B CN105225213B CN201510630936.7A CN201510630936A CN105225213B CN 105225213 B CN105225213 B CN 105225213B CN 201510630936 A CN201510630936 A CN 201510630936A CN 105225213 B CN105225213 B CN 105225213B
- Authority
- CN
- China
- Prior art keywords
- fusion
- pcnn
- component
- image
- color image
- 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.)
- Expired - Fee Related
Links
Abstract
The invention provides a kind of Color Image Fusion method based on S PCNN and laplacian pyramid, and a kind of effective Color Image Fusion is proposed with Lapland Lars pyramid algorith based on pulse-couple network (S PCNN) is simplified.In HSV color spaces, after adding local entropy to carry out characteristic area cluster to H components using S PCNN, the H components fusion of each source images is realized based on pulse oscillating frequency figure；S, V component are carried out to resolution decomposition using laplacian pyramid, then S, V component merged using different fusion rules.Color space inverse transformation finally is carried out to new H, S, V component, realizes the fusion of RGB color image.The present invention's test result indicates that, inventive algorithm better than other conventional Image Fusions either in subjective vision effect, or objective evaluation standard.
Description
Technical field
The invention belongs to technical field of image processing, and in particular to a kind of coloured silk based on S-PCNN and laplacian pyramid
Color image fusion method.
Background technology
Image co-registration is the consistent branch of information fusion, is one of the study hotspot of current information fusion.Coloured image melts
It is more accurate, comprehensive, reliable to close description of the image after making fusion to Same Scene.The research of Color Image Fusion is relative at present
It is less, and human vision to color information can identification be far above gray level image.With the improvement and offer of sensor technology,
Color Image Fusion can more and more be paid attention to.Different brightness and different color are combined and represent image by coloured image
Information.So the fusion based on color space is usually that each component is merged respectively.Algorithms most in use such as IHS, weighting and
PCA converter techniques scheduling algorithm is easily realized still ineffective.And the blending algorithm based on multiresolution analysis, typically first treat and melt
The multi-source image of conjunction carries out image conversion, and then the coefficient after conversion is reconfigured.According to picture breakdown mode not
It is same to be broadly divided into the blending algorithm based on pyramid transform, the blending algorithm based on wavelet transformation and become based on multiple dimensioned geometry
The blending algorithm changed, this kind of algorithm are used for the image co-registration of Pixel-level.Pixel-level fusion is classified most in image co-registration
One layer of bottom, each of which pixel are all that other source images respective pixels are determined.Pulse Coupled Neural Network (PCNN) according to
Its image procossing, pattern-recognition, route solve etc. field premium properties, be described as the 3rd ground artificial neural network.PCNN
A kind of new neural network different from traditional artificial neural network, it be interconnected together by several neurons it is anti-
Feedback type network, form PCNN neurons be a comprehensive dynamic non linear system, than it have traditional artificial neural network without
The advantage of method analogy.Laplacian pyramid algorithm is a kind of multiple dimensioned, multiresolution, the image processing method of multilayer decomposition
Method, it can by the key character (such as edge, texture) of image according on different Scale Decomposition to different decomposition layers, with
Simple image blending algorithm is compared, and it can obtain more preferable syncretizing effect, be widely used among image co-registration.
The content of the invention
In order to solve above mentioned problem existing for prior art, the invention provides one kind based on S-PCNN and Laplce's gold
The Color Image Fusion method of word tower.
The technical solution adopted in the present invention is：
The present invention's is studied based on the Color Image Fusion of S-PCNN and laplacian pyramid, is comprised the steps of：
A. first, registered good RGB color image space is transformed into HSV color spaces；
B. after carrying out characteristic area cluster to H components using S-PCNN, each source images are realized with based on pulse oscillating frequency figure
H components fusion；H component processing procedures：First H components feeding S-PCNN models are iterated, obtain igniting frequency figure OFG, then
Local entropy calculating processing is carried out to S-PCNN igniting frequency figure OFG corresponding to H components, obtains OFG local entropies matrix as H components
Eigenmatrix, the H component respective pixel local entropies LE of different source images size is finally contrasted, is taken with larger local entropy
Pixel is as fusion pixel；
C. image Laplacian pyramid, wherein, S, V component, S, V component are carried out using laplacian pyramid
To resolution decomposition, then S, V component are merged using different fusion rules；
D. new H, S, the V component finally obtained to step A, B, C carries out HSV color space inverse transformations, finally gives and melts
RGB color image after conjunction.
Beneficial effects of the present invention are：The present invention is based on simplified pulse-couple network (S-PCNN) and laplacian pyramid
Algorithm proposes a kind of effective Color Image Fusion；RGB color image space is transformed into HSV color spaces first；
In HSV color spaces, after carrying out characteristic area cluster to H components using S-PCNN+ local entropies, realized based on pulse oscillating frequency figure
The H components fusion of each source images；S, V component are carried out using laplacian pyramid, to resolution decomposition, then to utilize difference
Fusion rule merges to S, V component.HSV color space inverse transformations finally are carried out to new H, S, V component, after obtaining fusion
RGB color image.The present invention is contrasted with other several algorithms most in use, no matter regarded from subjectivity by some objective indicators
Feel that better than other conventional Image Fusions, experimental result all show that inventive algorithm can in effect or objective evaluation standard
To preserve the details of coloured image and color information well, the coloured image after fusion is relatively sharp, reliable.
It is provided by the invention to be based on S-PCNN and laplacian pyramid algorithm, with reference to the hsv color suitable for human eye vision
A kind of model, it is proposed that effective Color Image Fusion.The algorithm, which is first converted to RGB color image, meets human eye vision
The HSV color spaces of characteristic；H component characterizations are clustered using S-PCNN and local entropy, then to it to merging, because
It knows the characteristic that the PCNN models of mechanism cluster to image with provincial characteristics with mammal visual cortex visual sense；To S, V points
Amount carries out Laplacian pyramid, is merged by different strategies, because it is to carry out multiple dimensioned point to characteristics of image
The effective tool of analysis, its Pyramid transform data embody the band logical turriform filtering of image, and its size decomposes chi in each layer
It is consistent under degree；Inverse transformation, the RGB color image merged finally are carried out to the HSV chrominance components that fusion obtains.Cause
This, the advantages of present invention can utilize Laplacian pyramid and S-PCNN, is effectively merged to RGB image.It is real
Verify that bright color fusion algorithms proposed by the present invention can preferably merge not confocal coloured image, and source can be effectively maintained
Details, texture and the principal character information of image.Illustrate that inventive algorithm no matter in visual effect or objectively, is above better than
Other algorithms.
Brief description of the drawings
Fig. 1 are Color Image Fusion flow chart of the present invention；
Fig. 2 is source of embodiment of the present invention Fig. 1；
Fig. 3 is source of embodiment of the present invention Fig. 2；
Fig. 4 schemes for present invention fusion.
Embodiment
As shown in figure 1, the invention provides a kind of Color Image Fusion side based on S-PCNN and laplacian pyramid
Method.The present invention is further detailed explanation with reference to the accompanying drawings and examples.
Embodiment：A kind of Color Image Fusion method based on S-PCNN and laplacian pyramid, specifically comprising following
Step：
A, the Color Image Fusion of S-PCNN and laplacian pyramid；First, RGB color image space is converted to HSV
Image, obtains tri- components of H, S, V, and tri- components of H, S, V are respectively：Hue, saturation, intensity.H is measured with angle, value
Scope is 0 °~360 °；S represents color purity level, and V represents image light levels, is the measurement to gray scale, their scope
All it is 0~l.The transformation relation formula that RGB color image space is transformed into HSV color spaces is as follows：
V=max (R, G, B) (3)
R in above formula is the red component in common rgb format image, and G is green component, B blue components, above-mentioned formula
Statement is how RGB to be transformed into HSV color spaces, representation of the image in HSV domains has been calculated, RGB and HSV are only
It is the different representations of same sub-picture.Because it is more more effective than R, G, B component to handle H, S, V component, use herein
Be to H, S, V component processing, so, the image of rgb format is converted into the images of HSV forms, then to tri- components of HSV
Handled.
B, in HSV color spaces, after carrying out characteristic area cluster to H components using S-PCNN, with based on pulse oscillating
Frequency figure tests the H components fusion of each source images.H component processing procedures：First H components feeding S-PCNN models are iterated, obtained
To OFG (igniting frequency is schemed), then local entropy calculating processing is carried out to S-PCNN igniting frequency figures (OFG) corresponding to H components, obtain OFG
Eigenmatrix of the local entropy matrix as H components.Finally contrast the big of the H component respective pixel local entropies (LE) of different source images
It is small, the pixel with larger local entropy is taken as fusion pixel.Wherein S-PCNN models are as follows；
Fij(n)=Sij (4)
Lij(n)=VLΣklWkjYijkl(n-1) (5)
Uij(n)=Fij(n)[1+βLij(n)] (6)
For neuron Nij, the F passages of formula (7) description input and the L * channel of formula (8) description constitutes its acceptance region,
The F passages of wherein neuron receive external drive input Sij, i.e. the pixel value of image, and L * channel receives neighborhood neuron Nkl's
Pulse excitation inputs Yijkl, WkjLink and weigh for neighborhood, and VLFor passage amplitude.Then exported in modulation domain, the F passages of neuron
The internal state value U, β for foring neuron by the nonlinear modulation that is multiplied with L * channel output are that L * channel exports in modulation domain
Link strength.Finally, the θ when internal state value U is more than the threshold value of neuronijWhen, neuron sends pulse, i.e. Yij=1.
In iterative process, threshold θijIt is nonlinear exponential damping change, damped expoential αθ, but after pulse is sent, θij
An amplitude coefficient has also been superimposed while exponential damping is carried out。
C, image Laplacian pyramid, the conversion of the laplacian pyramid of image is by gaussian pyramid
On the basis of obtain.Therefore, Laplacian pyramid is divided into two steps：Image is first subjected to gaussian pyramid decomposition, then
Obtain laplacian pyramid.Gauss operator and original image (being represented with G0) are subjected to convolution (i.e. Gassian low-pass filter), then
Interlacing is carried out every the down-sampling of row, obtains approximation of the image on low resolution, resolution ratio is half (the Gauss gold word of artwork
The first layer of tower).LPF and down-sampling are entered to the image after sampling again, obtain next layer of gaussian pyramid.So repeatedly
Aforesaid operations are carried out, if obtaining dried layer, to form gaussian pyramid.This tomographic image by gaussian pyramid obtained above is big
It is small for the 1/4 of previous tomographic image, then enter row interpolation expansion to gaussian pyramid using interpolation method, make l tomographic images Gl, expansion
Size afterwards and l-1 tomographic images Gl-1Size is identical, and its calculation is as follows：
Wherein,
Order
Wherein, w (m, n) is low pass window function in above formula, and size is 5 × 5, LPlFor laplacian pyramid l layers,
LPNFor laplacian pyramid n-th layer.
D, original image is rebuild by laplacian pyramid；Each tomographic image of laplacian pyramid is amplified to through progressively interpolation
Big with as next tomographic image, original image can be rebuild by being then added again；When the number of plies of laplacian pyramid is when 0 arrives N, its
Energy size can reflect the quality of image, all that coefficient is chosen by the way of region energy.
S-PCNN is to information sensings such as the details of image, edges, and the high level of laplacian pyramid contains the thin of image
Section and texture variations information, so using S-PCNN models of the present invention in high layer coefficients to choose to retain the details of image and line
Manage information.S, V component are subjected to Laplacian pyramid respectively, and synthesized each tomographic image information newly by above-mentioned strategy
S, V component.The new HSV coloured images that will be obtained, are converted to rgb space, the coloured image after being merged.From table 1
As can be seen that numeric ratio other algorithms of this paper algorithms in these evaluation objective indicators are more effective, wherein, SF expression spaces
Frequency, STD, standard deviation is represented, EN represents entropy, and AV represents average gradient, and M represents average, and these values are bigger expression figures
As quality is better.It can be seen that the present invention is the most clear in Fig. 2, and color and source images are the most close, it was demonstrated that this paper algorithms
To the validity and feasibility of Color Image Fusion.
Fusion mass evaluation of estimate of the table 1. based on fused images obtained by different fusion methods.
| SF | STD | EN | AV | M | |
| Weighting | 12.57 | 65.01 | 7.45 | 4.20 | 116.97 |
| PCA | 12.76 | 65.17 | 7.44 | 4.23 | 166.67 |
| WT | 13.19 | 65.39 | 7.49 | 4.31 | 163.33 |
| PCNN | 17.62 | 66.27 | 7.45 | 5.27 | 168.50 |
| Herein | 20.40 | 67.08 | 7.44 | 5.85 | 166.96 |
The present invention is not limited to above-mentioned preferred forms, and anyone can show that other are various under the enlightenment of the present invention
The product of form, however, make any change in its shape or structure, it is every that there is skill identical or similar to the present application
Art scheme, is within the scope of the present invention.
Claims (1)
- A kind of 1. Color Image Fusion method based on S-PCNN and laplacian pyramid, it is characterised in that：Specifically include down Row step：A. first, registered good RGB color image space is transformed into HSV color spaces；B. after carrying out characteristic area cluster to H components using S-PCNN, divided with the H that each source images are realized based on pulse oscillating frequency figure Amount fusion；H component processing procedures：First H components feeding S-PCNN models are iterated, obtain igniting frequency figure OFG, then to H points S-PCNN igniting frequency figure OFG carries out local entropy calculating processing corresponding to amount, obtains feature of the OFG local entropies matrix as H components Matrix, the H component respective pixel local entropies LE of different source images size is finally contrasted, takes the pixel with larger local entropy As fusion pixel；C. image Laplacian pyramid, wherein, S, V component, S, V component are carried out using laplacian pyramid to dividing Resolution is decomposed, and then S, V component are merged using different fusion rules；D. new H, S, the V component finally obtained to step A, B, C carries out HSV color space inverse transformations, after finally giving fusion RGB color image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510630936.7A CN105225213B (en) | 2015-09-29 | 2015-09-29 | A kind of Color Image Fusion method based on S PCNN and laplacian pyramid |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510630936.7A CN105225213B (en) | 2015-09-29 | 2015-09-29 | A kind of Color Image Fusion method based on S PCNN and laplacian pyramid |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105225213A CN105225213A (en) | 2016-01-06 |
| CN105225213B true CN105225213B (en) | 2018-03-09 |
Family
ID=54994164
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510630936.7A Expired - Fee Related CN105225213B (en) | 2015-09-29 | 2015-09-29 | A kind of Color Image Fusion method based on S PCNN and laplacian pyramid |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105225213B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341502B (en) * | 2017-05-31 | 2020-09-15 | 三峡大学 | Image fusion method and device based on PCNN and linear superposition technology |
| CN107292845B (en) * | 2017-06-26 | 2020-04-17 | 安健科技(重庆)有限公司 | Standard deviation pyramid-based dynamic image noise reduction method and device |
| CN108830237B (en) * | 2018-06-21 | 2020-08-11 | 北京师范大学 | Facial expression recognition method |
| CN110163830B (en) * | 2019-04-19 | 2021-02-19 | 三峡大学 | Image fusion method based on Riesz-Lap transformation and PCNN |
| CN110600109B (en) * | 2019-09-10 | 2023-12-26 | 安徽中科微因健康科技有限公司 | Diagnosis and monitoring comprehensive medical system with color image fusion and fusion method thereof |
| CN116152343B (en) * | 2023-04-18 | 2023-07-04 | 中铁九局集团第一建设有限公司 | Precise measurement method of measurement robot based on vision |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101980290A (en) * | 2010-10-29 | 2011-02-23 | 西安电子科技大学 | Method for fusing multi-focus images in anti-noise environment |
| CN102722877A (en) * | 2012-06-07 | 2012-10-10 | 内蒙古科技大学 | Multi-focus image fusing method based on dual-channel PCNN (Pulse Coupled Neural Network) |
| CN103617604A (en) * | 2013-08-28 | 2014-03-05 | 内蒙古科技大学 | Image fusion method based on characteristic extraction of two dimension empirical mode decomposition method |
| US8913848B2 (en) * | 2012-02-29 | 2014-12-16 | Ningbo University | Microscopic image fusion method based on region growing |
| CN104282007A (en) * | 2014-10-22 | 2015-01-14 | 长春理工大学 | Contourlet transformation-adaptive medical image fusion method based on non-sampling |
| CN104616274A (en) * | 2015-02-09 | 2015-05-13 | 内蒙古科技大学 | Algorithm for fusing multi-focusing image based on salient region extraction |
-
2015
- 2015-09-29 CN CN201510630936.7A patent/CN105225213B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101980290A (en) * | 2010-10-29 | 2011-02-23 | 西安电子科技大学 | Method for fusing multi-focus images in anti-noise environment |
| US8913848B2 (en) * | 2012-02-29 | 2014-12-16 | Ningbo University | Microscopic image fusion method based on region growing |
| CN102722877A (en) * | 2012-06-07 | 2012-10-10 | 内蒙古科技大学 | Multi-focus image fusing method based on dual-channel PCNN (Pulse Coupled Neural Network) |
| CN103617604A (en) * | 2013-08-28 | 2014-03-05 | 内蒙古科技大学 | Image fusion method based on characteristic extraction of two dimension empirical mode decomposition method |
| CN104282007A (en) * | 2014-10-22 | 2015-01-14 | 长春理工大学 | Contourlet transformation-adaptive medical image fusion method based on non-sampling |
| CN104616274A (en) * | 2015-02-09 | 2015-05-13 | 内蒙古科技大学 | Algorithm for fusing multi-focusing image based on salient region extraction |
Non-Patent Citations (3)
| Title |
|---|
| 基于DT-CWT的图像融合算法研究;黄品霞;《中国优秀硕士学位论文全文数据库 信息科技辑》;20150915;第37页第5-21行、图4.4 * |
| 基于改进LP变换及自适应PCNN的多聚焦图像融合方法;严春满 等;《控制与决策》;20120531;第4节 * |
| 脉冲耦合神经网络在图像处理中的应用研究;李建锋;《中国博士学位论文全文数据库 信息科技辑》;20141215;正文第82-83页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105225213A (en) | 2016-01-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105225213B (en) | A kind of Color Image Fusion method based on S PCNN and laplacian pyramid | |
| He et al. | Multimodal medical image fusion based on IHS and PCA | |
| CN105719263B (en) | Visible ray and infrared image fusion method based on NSCT domains bottom visual signature | |
| CN109035166A (en) | Electrical equipment infrared image enhancing method based on non-lower sampling shearing wave conversion | |
| CN106897972A (en) | A kind of self-adapting histogram underwater picture Enhancement Method of white balance and dark primary | |
| CN104240194B (en) | A kind of enhancement algorithm for low-illumination image based on parabolic function | |
| CN105139371B (en) | A kind of multi-focus image fusing method based on PCNN and LP transformation | |
| CN103942769B (en) | A kind of satellite remote-sensing image fusion method | |
| CN103345732B (en) | Based on PCNN algorithm for image enhancement and the device of contourlet transformation | |
| CN103034986A (en) | Night vision image enhancement method based on exposure fusion | |
| CN107977926A (en) | A kind of different machine brain phantom information fusion methods of PET/MRI for improving neutral net | |
| DE10239801A1 (en) | Process for extracting texture features from a multi-channel image | |
| CN106651817A (en) | Non-sampling contourlet-based image enhancement method | |
| CN107103596A (en) | A kind of color night vision image interfusion method based on yuv space | |
| CN102903081A (en) | Low-light image enhancement method based on red green blue (RGB) color model | |
| CN111080568A (en) | Tetrolet transform-based near-infrared and color visible light image fusion algorithm | |
| CN106485674A (en) | A kind of low light image Enhancement Method based on integration technology | |
| CN106971380A (en) | A kind of contrast enhancing and application of the visual saliency optimization method in golf course figure | |
| CN106056565B (en) | A kind of MRI and PET image fusion method decomposed based on Multiscale Morphological bilateral filtering and contrast is compressed | |
| Mingrui et al. | Multi-focus image fusion algorithm using LP transformation and PCNN | |
| Bao et al. | A novel remote sensing image fusion approach research based on HSV space and bi-orthogonal wavelet packet transform | |
| CN109447934A (en) | Based on two step sparse codings and wavelet transformation PAN and multi-spectral image interfusion method | |
| CN109544470A (en) | A kind of convolutional neural networks single image to the fog method of boundary constraint | |
| CN104732507A (en) | Fusion method for two images of different illuminations based on texture information reconstruction | |
| CN112734679A (en) | Fusion defogging method for medical operation video images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180309 Termination date: 20180929 |