CN106228528A - A kind of multi-focus image fusing method based on decision diagram Yu rarefaction representation - Google Patents
A kind of multi-focus image fusing method based on decision diagram Yu rarefaction representation Download PDFInfo
- Publication number
- CN106228528A CN106228528A CN201610605703.6A CN201610605703A CN106228528A CN 106228528 A CN106228528 A CN 106228528A CN 201610605703 A CN201610605703 A CN 201610605703A CN 106228528 A CN106228528 A CN 106228528A
- Authority
- CN
- China
- Prior art keywords
- image
- fusion
- decision
- focus
- region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000010586 diagram Methods 0.000 title claims abstract description 39
- 230000004927 fusion Effects 0.000 claims abstract description 73
- 230000008569 process Effects 0.000 claims abstract description 22
- 238000005516 engineering process Methods 0.000 claims description 16
- 238000000354 decomposition reaction Methods 0.000 claims description 13
- 238000009499 grossing Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000000052 comparative effect Effects 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 230000014509 gene expression Effects 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 9
- 230000000007 visual effect Effects 0.000 abstract description 5
- 238000013461 design Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 8
- 238000003384 imaging method Methods 0.000 description 6
- 230000009466 transformation Effects 0.000 description 6
- 239000013598 vector Substances 0.000 description 5
- 238000013459 approach Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000006835 compression Effects 0.000 description 3
- 238000007906 compression Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 230000001976 improved effect Effects 0.000 description 3
- 238000002156 mixing Methods 0.000 description 3
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
- 238000013316 zoning Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/77—Retouching; Inpainting; Scratch removal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- 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
- G06T2207/20048—Transform domain processing
- G06T2207/20064—Wavelet transform [DWT]
-
- 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
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/513—Sparse representations
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of multi-focus image fusing method based on decision diagram Yu rarefaction representation, this method characteristic based on human visual system proposes the multi-focus image fusion framework different from conventional Multi-focus image fusion, and the transitional region for multiple focussing image is analyzed and studies, to avoid its impact on fusion results, improve the quality of fusion image.Realize process: on the basis of the low scalogram picture of multiple focussing image is carried out definition analysis, produce decision diagram, and obtain fusion results according to this decision diagram；There is deviation in the definition judgement in view of transitional region, the decision diagram generated can be caused to there is error, it is necessary to determine transitional region and use Multi-focus image fusion based on rarefaction representation to process, it is thus achieved that the fusion results of transitional region；Finally, fusion results based on decision diagram and transitional region fusion results are carried out mean operation, it is thus achieved that final fusion image.
Description
Technical field
The invention belongs to technical field of image processing, relate to compression of images, image sparse represents, image space frequency ratio
Relatively, Mathematical Morphology technology and image fusion technology, can be used for the blur-free imaging of machine vision, target recognition, digital camera
In field.
Background technology
In shooting process, after the focal length of imaging system sets, due to the restriction of the camera lens depth of field, only camera lens conjugate planes
The most a range of object imaging is clearly, and the object imaging in this scope is not then fuzzy.Actual
Time in utilization to certain scene imaging, owing to subject is different with the distance of camera lens, imaging is frequently not all clear
Clear.For obtaining the full picture rich in detail of scene, each object in scene is focused on by a feasible way exactly respectively, obtains it
Picture rich in detail, and its region extracted merge.Thus, extract the clear of image the most exactly
The most how region, obtain and the decision diagram that clear area and fuzzy region accurately divide become the pass weighing fused image quality
Key.Generally, the transitional region between clear area and fuzzy region is difficult to clearly be divided, this work extracted for clear area
Bring difficulty, cause obtained decision diagram to there is error, reduce the quality of fusion image.Have in view of human visual system
There is following characteristic: when quickly distinguishing clear area with fuzzy region, be more likely to directly find clear area and fuzzy region
Border and be not primarily upon the detailed information of image.If human eye vision feature quick obtaining decision diagram can be utilized, and root
The transitional region between clear area and fuzzy region is determined, further to this region according to the boundary line contained by this decision diagram
Carrying out processing to improve the accuracy rate that clear area is extracted, the raising making fusion mass is possibly realized by this.Given this plant it is assumed that
And with the image-forming principle of multiple focussing image and the feature of its low scalogram picture as foundation, it is proposed that based on decision diagram and sparse table
The multi-focus image fusing method shown.The method by compression of images, the rarefaction representation of image, image space frequency ratio relatively, mathematics
Morphology technology and image fusion technology are constituted, and main thought can be divided into five steps to be described；(1) in view of poly
The low scalogram picture of burnt image has main reflection picture structure and does not comprise the feature of too much image detail information, uses many chis
Multiple focussing image is decomposed by degree wavelet transformation, it is thus achieved that size is the low scalogram picture that former multiple focussing image is the most half as large；(2)
Analyze low scalogram and as the definition of each point and each point is divided into focusing, out of focus, uncertain three classes, to produce initial decision map,
And initial decision map is carried out up-sampling operation so that it is size is in the same size with the multiple focussing image of input；(3) space is used
Each point in the uncertain region of initial decision map is further divided into focusing on and out of focus two class by frequency approach, it is thus achieved that final decision
Figure, determines and defines focusing territory, the boundary line in out of focus territory；(4) in view of the transitional region impact on decision-making plot quality, according to
The multiple focussing image of input is merged by whole decision diagram, and detects transitional region according to the boundary line obtained, and utilizes
This region is merged by Multi-focus image fusion based on rarefaction representation；(5) both fusion results are combined, it is thus achieved that
Whole fusion image.Compared to traditional multi-focus image fusion algorithm, there is following two feature in this method: one, traditional poly
Burnt Image Fusion is directly focused feature extraction to multiple focussing image, and we's rule is according to image low frequency sub-band image
The characteristic being had, uses wavelet transformation to decomposite the low scalogram picture of multiple focussing image, and then utilized complete rarefaction representation
The definition i.e. focus characteristics of this low scalogram picture is analyzed by model.So, not only reduce the computation complexity of method, and
And both ensured that the focus characteristics of former multiple focussing image was constant, reduce again the difficulty that multiple focussing image definition is analyzed.Two, tradition
Multi-focus image fusion have ignored the transitional region impact on fusion mass of multiple focussing image, and the method is to poly
The transitional region of burnt image is analyzed and utilizes multi-focus image fusion based on rarefaction representation to merge it, reduces
This region adverse effect, to improve the effect of fusion image.Therefore, utilize the method, fusion image can be improved
Effect, and, the method, compared with multiple focussing image algorithm based on rarefaction representation, has higher ageing.
Wavelet transformation is that function, operator or data are divided into various frequency content, then carries out it multiple dimensioned thin
Change.As a kind of mathematical tool, wavelet transformation has played effect in terms of image co-registration, picture breakdown for having various direction
The subband signal of feature, go forward side by side line frequency domain analysis and time-domain analysis.Generally, wavelet algorithm is used to carry out at image co-registration
During reason, image is considered as 2D signal, and it is carried out multi-level decomposition.Image can generate after wavelet decomposition one low
Frequency component and three high fdrequency components, they are the expressions to the different content in original image in order to simpler obtain
Take the feature in picture.The low frequency sub-band image of wavelet decomposition remains in that general picture and the spatial character of original image, simply size
Diminish.Therefore, the characteristic that the present invention is had according to the low frequency sub-band image of wavelet decomposition, use wavelet transformation to multi-focus
Source images carries out level of decomposition and by acquisition low frequency sub-band as low scalogram picture, for using complete rarefaction representation side further
Method carries out rarefaction representation to it and lays the foundation.
In recent years, cross complete rarefaction representation and effectively represent that as one model is widely used in image denoising, image recovers
Etc. in the task such as image processing field and image recognition.The complete rarefaction representation of mistake of signal is i.e. to seek from complete dictionary excessively
The linear combination of minimum atom represents signal.Generally the data message of natural image also has redundancy, therefore may be used
To carry out rarefaction representation in redundant dictionary.Cross in complete rarefaction representation problem, the complete dictionary of mistake chosen whether appropriate relation
Openness to representing that signal is expressed.For the image of different structure Yu feature, the dictionary that should take various forms or base
Function.The construction method crossing complete dictionary can be divided into two classes: (1) builds dictionary according to mathematical model；(2) according to sample
Design dictionary.Equations of The Second Kind dictionary construction method can learn to be suitable for the dictionary of a certain class signal characteristic by training sample,
Openness to guarantee that signal represents.Can the common method compatible with Its Sparse Decomposition algorithm now be that ELAD proposed in 2006
K-SVD learning algorithm.The method is alternately performed the renewal process of the signal rarefaction representation at current dictionary and atom, reaches
The purpose of study dictionary.In addition to dictionary construction step, designing Its Sparse Decomposition algorithm fast and effectively is to be related to signal equally
Indicate whether the committed step of optimum.Existing Its Sparse Decomposition algorithm is summarized as three classes: greedy tracing algorithm (OMP), based on model
The number algorithm of canonical, iterative shrinkage algorithm.Wherein, greedy algorithm is with Greedy idea as core, in each step iterative process,
Select the atom mated most with residual error to participate in sparse bayesian learning from dictionary.Such algorithm mainly includes matching pursuit algorithm
(and the orthogonal matching pursuit algorithm improved on its basis).The present invention uses based on crossing complete dictionary K-SVD with OMP sparse
The sparse representation model of encryption algorithm, carries out rarefaction representation, and then analyzes each pixel in low scalogram picture low scalogram picture
Definition obtain definition score value figure, the generation for decision diagram lays the foundation.
Spatial frequency reflects the overall active degree of piece image, and the biggest image of spatial frequency is the most active, the most clear.Empty
Between frequency calculated by line frequency and row frequency and get, i.e. the change frequency of horizontal and vertical directions.This concept extensive
Use, for the research such as processing of visual characteristic, figural perception and the transmission of visual system signal, information provide one new
Approach.Therefore, the present invention utilizes spatial frequency that uncertain region is made further definition analysis, the biggest explanation of spatial frequency
The definition of this point is the highest.
Mathematical morphology (Mathematical Morphology) is built upon a subject on the basis of set theory, non-
Often it is suitable for geometrical morphological analysis and the description of signal.Its basic thought is to utilize structural element to carry out signal " detection ", retains
Principal shape, deletes irrelevant shape (such as noise, burr).Form and difference operation, i.e. expanding with corrosion is mathematical morphology
Basis.What first mathematical morphology processed is bianry image, referred to as binary mathematical morphology (Binary Morphology).Two
Value mathematical morphology is a kind of processing procedure for set.It is to move a structural elements in the picture that morphological images processes
Element, then carries out the set operations such as intersecting and merging by structural element and following bianry image.First corrode the process expanded afterwards to be referred to as
Opening operation.It has elimination small objects, in very thin place separating objects and the effect on smooth larger object border.The present invention is just
It it is the small surfaces territory using opening operation to remove in bianry image.
Along with sparse representation theory is widely used in image co-registration process, blending algorithm based on rarefaction representation
Research is also paid close attention to by Chinese scholars.Sparse representation theory is applied in image co-registration by Yang B. etc., and the method is improved
The effect merged, but training dictionary process ratio is relatively time-consuming.2013, Chen L. et al. and Yin H. et al. was by rarefaction representation
Combine for image co-registration with other blending algorithm, improve fused image quality, but the orthogonal coupling in rarefaction representation is calculated
Method amount of calculation is very big, and the image co-registration of two width is the longest.Liu Y. et al. is by by multi-scale geometric analysis method and sparse table
Representation model combines, it is proposed that a kind of general image co-registration framework, overcome simple based on multi-scale geometric analysis method or
Inherent defect existing for the fusion method of person's rarefaction representation.They also sparse representation model is applied simultaneously to image co-registration with
Image denoising.Relative to traditional multi-scale geometric analysis method, the obtained fusion of Image Fusion based on rarefaction representation
The effect of image is preferably and the strong robustness of algorithm.Therefore, the present invention uses multi-focus image fusion based on rarefaction representation to calculate
Transitional region is merged to improve the quality of final fusion results by method.
Summary of the invention
It is an object of the invention to, propose a kind of new multi-focus image fusion framework i.e. based on decision diagram and rarefaction representation
Multi-focus image fusing method, be possible not only to realize the fusion of several multiple focussing images, it is also possible to improve existing multi-focus figure
As the shortcoming that the blending algorithm suitability is low, and improve the quality of multi-focus image fusion image.
The technical scheme is that, first input the two width polies about a certain scene obtained from a certain digital camera
Burnt image, decompose input picture obtaining size by multi-scale wavelet transformation is input picture low scalogram picture the most half as large,
Complete sparse representation model was used it to be analyzed, by low scalogram with intelligibility measure method based on image block subsequently
As each pixel is divided into three classes: focusing, out of focus, uncertain, produce initial decision map and also it carried out up-sampling process so that
The size of initial decision map is in the same size with input picture.Focusing territory, the final decision figure in out of focus territory is only comprised in order to obtain,
Use spatial frequency method that the uncertain region in initial decision map is processed, determine to divide and focus on territory, out of focus territory
Demarcation line.According to final decision figure, extract the clear area of each input picture and they combinations are obtained fusion image.So
And, owing to being difficult to be divided into by the pixel of transitional region exactly two classes: focusing, out of focus, the final decision figure resulted in
There is error.Therefore, it is necessary to determine transitional region and it be analyzed and process, thus reduce the shadow that error is brought
Ring, improve fusion mass.The present invention is centered by detected boundary line, and the rectangular area of suitable radius is considered transition region
Territory, utilizes the fusion results that it is carried out merging this region of acquisition by Multi-focus image fusion based on rarefaction representation.Finally,
In conjunction with the fusion image obtained based on decision diagram, i.e. the fusion results of transitional region is merged knot with the decision diagram of respective regions
Fruit carries out mean operation, produces final fusion image.
Specifically comprise the following steps that
One, input two width multiple focussing image to be fused about a certain scene, the multiple focussing image of this two width input is carried out
Wavelet decomposition, it is thus achieved that,,,, four width sub-band images, wherein willLow frequency sub-band image is as the present invention's
Low scalogram picture.
Multiple focussing image to be fused can be that gray level image can also be coloured image but must be about Same Scene
Multiple focussing image；Wherein, through a wavelet decomposition, the size of the low scalogram picture obtained is that input multiple focussing image is big
Little half.
Two, use complete sparse representation model that low scalogram picture is carried out rarefaction representation, namely extract sparse spy
Levy composition two width corresponding degree of rarefication figure.
Three, compared by intelligibility measure method based on image block and weigh the definition of each point in degree of rarefication figure, obtaining
Obtain two width corresponding definition score value figure.
Four, according to above-mentioned obtained definition score value figure, it is carried out binarization operation and obtains corresponding bianry image.
Wherein, can there is disconnected small surfaces territory due to erroneous judgement in bianry image, needs to open fortune by Mathematical Morphology technology
Calculate and eliminate this region.Subsequently, the bianry image after processing two width according to the decision rule arranged makes decisions at the beginning of acquisition one width
Beginning decision diagram.Owing to the size of initial decision map is only the half of input picture, therefore it is carried out up-sampling process so that it is big
Little in the same size with input picture.
Wherein, binary conversion treatment can be described as: arranges suitable threshold value, and more than the point of threshold value, definition values is entered as 1；
The point being unsatisfactory for above-mentioned condition will be assigned 0.
By the process of above-mentioned steps three, initial decision map contains: focus on territory, out of focus territory, uncertain region these three
Region, in brief, each point in the multiple focussing image of input is divided into above-mentioned three classes by decision diagram.
Five, utilize smoothing windows technology that the uncertain region in each multiple focussing image of input is divided into polylith sub-block, meter
Calculate and compare the spatial frequency of relevant block, according to comparing result, each point in uncertain region is further divided into two classes: focus on
With out of focus, thus obtain final decision diagram.
Wherein, decision diagram becomes only comprising two regions: focus on territory, out of focus territory, the border simultaneously divided in this two region
Line is determined the most therewith.
Uncertain region not necessarily belongs to the rectangular area of rule, therefore to be divided into son of the same size
Block, is positioned at the pixel of edges of regions, and the content of its corresponding sub-block not comes under uncertain region.
Six, based on final decision figure, input multiple focussing image is extracted in the clear area of its position, zoning
Come, combine the region extracted subsequently according to the fusion rule arranged, it is thus achieved that fusion results based on decision diagram.
Seven, according to the boundary line in above-mentioned steps four obtained final decision figure, transitional region is determined.Use based on dilute
Transitional region is merged by the Multi-focus image fusion that relieving the exterior syndrome shows, it is thus achieved that the fusion results in this region.
Point centered by boundary line in final decision figure, arranging suitable parameter is radius, is formed around central point
Rectangular area be transitional region.
Eight, last, the fusion results of fusion results based on decision diagram Yu transitional region is carried out mean operation, it is thus achieved that
Whole fusion image.
The mode of mean operation: relevant position pixel is taken average as the final pixel value of this point.
In order to improve the effect of multi-focus image fusion, the characteristic being had according to human visual system, the present invention proposes
A kind of new multi-focus image fusion framework i.e. based on decision diagram Yu rarefaction representation multi-focus image fusing method, the method
By compression of images, image sparse represents, image space frequency contrast, Mathematical Morphology technology have been combined with image fusion technology
Come, analyze the characteristic of the low scalogram picture of multiple focussing image, produce decision diagram, obtain fusion results according to decision diagram；And consider
To the transitional region erroneous judgement impact on syncretizing effect based on decision diagram, determine transitional region according to decision diagram further and carry out
Process, reduce transitional region and judge the error brought by accident, improve the quality merging picture.
Accompanying drawing explanation
Fig. 1 present invention multi-focus image fusing method flow chart based on decision diagram Yu rarefaction representation.
Fig. 2 final decision of the present invention figure product process figure.
The operational flowchart that Fig. 3 degree of rarefication of the present invention figure generates with definition score value figure.
Two multiple focussing images to be fused of Fig. 4, 。
Fig. 5 present invention obtains two width low scalogram picture after two images to be fused are carried out wavelet decomposition, 's
Design sketch.
Fig. 6 present invention used complete sparse representation model and block-based intelligibility measure method respectively to low scalogram
Picture, Carry out definition analysis and obtain its corresponding definition score value figure, Design sketch.
Fig. 7 present invention is respectively to definition score value figure, Carry out binary conversion treatment and obtain its corresponding bianry image, Design sketch.
Fig. 8 present invention uses mathematical morphology open operator respectively to bianry image, After processing, acquisition processes
After bianry image, Design sketch.
Fig. 9 present invention is respectively to locating the bianry image after two width process, Make decisions acquisition initial decision map's
Design sketch.
Figure 10 present invention uses block-based spatial frequency measuring method to carry out uncertain region contained by initial decision map
Classification obtains final decision figureDesign sketch.
Figure 11 present invention is according to final decision figureCarry out input picture merging and obtain fusion resultsDesign sketch.
Figure 12 present invention is according to final decision figureTransitional region is determined in contained boundary lineDesign sketch.
Figure 13 present invention uses multi-focus image fusing method based on rarefaction representation to transitional regionCarry out merging and obtain
Fusion resultsDesign sketch.
Figure 14 present invention is by fusion resultsWith fusion resultsCarry out mean operation and obtain final fusion resultsEffect
Fruit figure.
Detailed description of the invention
Below the detailed description of the invention of the present invention is described in further detail.
As it is shown in figure 1, a kind of multi-focus image fusing method flow chart based on decision diagram Yu rarefaction representation of the present invention, first
First two multiple focussing images to be fused of input, , use daub1 small echo once to decompose, this two width image by institute
The low frequency sub-band image obtainedLow scalogram picture as corresponding every width multiple focussing image,。
Next step is the process that final decision figure generates.Shown in final decision figure product process figure as of the present invention in Fig. 2:
Input parameter: low scalogram picture, ；
Output result: final decision figure；
(1) respectively two width low scalogram picture is carried out rarefaction representation to obtain it corresponding initially with crossing complete sparse representation model
Degree of rarefication figure.Concrete mode is as follows: if low scalogram picture is,For its size,
The size crossing complete dictionary K-SVD is, use smoothing windows technology, from low scalogram as the upper left corner to the lower right corner, with one
Pixel is step-length, is divided into by low scalogram picture successivelyThe image block of size,Represent the quantity of image block.
The corresponding column vector of each image block, and these column vectors are combined sequentially into matrix.OMP sparse coding is utilized to calculate
Method is according to formulaCalculate the corresponding sparse coefficient of each column vector,
Wherein,It was the atom in complete dictionary K-SVD and had；And according to ruleTo each sparse vectorProcess.Subsequently, according to each sparse to
AmountIn the position of low scalogram picture, each sparse vector is polymerized and reconstructs with low scalogram as equal-sized sparse
Degree figure.Concrete operations are shown as shown in Figure 3.
(2) secondly, utilize intelligibility measure method based on image block to degree of rarefication figureCarry out intelligibility measure to obtain
Obtain corresponding definition score value figure.Concrete operations mode is: use smoothing windows technology, from degree of rarefication figureThe upper left corner is to right
Inferior horn, with a pixel as step-length, is divided into this figure successivelyThe image subblock of size,Represent image
The quantity of block.Degree of rarefication summation by pixel contained in every piece of image block, it is thus achieved that the definition of every piece of image block.Subsequently,
The equal-sized score value figure of three width sizes and degree of rarefication figure is set, wherein the value of score value figure each point is all 0.According to
The comparison rule arranged, compares the definition of respective image block in two width degree of rarefication figures, and by comparative result record at score value figure
In, wherein,Mainly it is responsible for recording the number of times that each pixel compares.Comparison rule is: for every two pieces of corresponding degree of rarefications
The sub-block of figure, it is assumed that the sub-block of one group of degree of rarefication figure, If, sub-blockDefinition more than sub-block, then by score value figure
Each point value of the position of corresponding sub block increases by 1；Otherwise at score value figureEach point value of the position of corresponding sub block increases by 1；Meanwhile, often
One group of sparse sub-block contrast terminates, by score value figureEach point value of the position of corresponding sub block increases by 1.Compare successively, until completing institute
The definition having sub-block contrasts.Finally, respectively by score value figure,The value of middle each point and score value figureThe value of middle each point is removed
Method computing, it is thus achieved that required degree of rarefication score value figure.Specific operation process is as shown in Figure 3.
(3) suitable threshold value is set, by each score value figureIn be entered as 1 more than the point value of this threshold value, be otherwise 0,
It is derived from bianry image.The opening operation utilizing Mathematical Morphology technology eliminates small surfaces territory discrete in binary map, and
And according to the bianry image after processingAnd judgement formula, generate
Initial decision diagram.Judgement formula may be interpreted as: if a certain pixelAt score value figureIn value be 1 and
Score value figureIn value be 0, then this point is referred to as focus point, and is entered as 1；If point is at score value figureIn value be 0
And at score value figureIn value be 1, then this point is referred to as focal point, and is entered as 0；If this point is all unsatisfactory for above-mentioned bar
Part, then be referred to as uncertainty node, and be entered as 0.5.
(4) last, use spatial frequency comparative approach further the point of uncertain region to be divided.Use smoothing windows
Technology, is divided into this region from uncertain region upper left corner the to the lower right corner successivelyThe image block of size, according to
FormulaCalculate every piece of sub-block of uncertain region in each input multiple focussing image
Spatial frequency, wherein,Represent with certain pointCentered by point sub-block region；Respectively represent some horizontal component with
The difference of vertical component.According to judgement formula, point corresponding to uncertain region
Make decisions, produce final decision figure.Decision rule may be interpreted as: if pointCorresponding sub-block is at multiple focussing imageIn spatial frequency values more than its at multiple focussing imageThe value of middle corresponding sub block, this point just belongs to focus point, and is entered as
1, it is otherwise focal point, and is entered as 0.
Next step is based on final decision figure again, according to fusion rule
The multiple focussing image of input is merged.Fusion rule may be interpreted as: extracts each region in decision diagram and inputs multi-focus
Clear area in image, and these clear areas are constituted fusion results according to corresponding position grouping。
Owing to the judgement of decision diagram is existed error, therefore next step is according to final decision figureThe boundary line comprised
Determine transitional region.Using each pixel of boundary line as central point, suitable value is set as upper and lower, left and right radius, draws
The fixed matrix area around central point is as transitional region.Use Multi-focus image fusion based on rarefaction representation to this district
Territory is merged, it is thus achieved that the fusion results of transitional region。
Finally, next step is by fusion resultsWith transitional region fusion resultsThe point of respective regions carries out average fortune
Calculate, it is thus achieved that final fusion results。
The above content detailed step based on decision diagram and the multi-focus image fusing method of rarefaction representation that is the present invention and
Implementation.This is broadly fallen into without departing from any change made on the premise of the design of the present invention for those skilled in the art
Within the protection domain of invention.
Claims (9)
1. multi-focus image fusing method based on decision diagram Yu rarefaction representation, the method comprises the following steps:
Step 1: first input two width multiple focussing images to be fused, respectively image is carried out a wavelet decomposition, by respective institute
The low-frequency image obtained is as the low scalogram picture of multiple focussing image；
Step 2: secondly, the low scalogram picture utilizing complete sparse representation model to be obtained above-mentioned steps carries out rarefaction representation
I.e. extract sparse features, constitute corresponding degree of rarefication figure；
Step 3: use smoothing windows technology and intelligibility measure method based on image block to above-mentioned obtained degree of rarefication figure
Carry out intelligibility measure and compare, it is thus achieved that corresponding definition score value figure；
Step 4: above-mentioned obtained definition score value figure is carried out binaryzation, mathematical morphology open operator operation, it is thus achieved that process
After binary map；According to process after binary map, according to set decision rule, image each point is divided three classes: focus on, from
Burnt, uncertain, generate initial decision map；Initial decision map is carried out up-sampling process, until its size and the poly inputted
Burnt image is in the same size；
Step 5: use smoothing windows technology, with a pixel as step-length, from the upper left corner in region to the lower right corner successively by every width
The uncertain region of multiple focussing image is divided into some sub-blocks；Calculate the spatial frequency of sub-block corresponding to each pixel, and compare
The result often organizing sub-block of respective pixel point, is defined as focus point or focal point according to comparative result by each pixel；Finally
Obtain and only comprise the final decision figure focusing on territory with out of focus territory；
Step 6: according to the final decision figure of above-mentioned acquisition, fusion based on decision diagram knot can be obtained according to the fusion rule arranged
Really；
Step 7: determine transitional region according to the boundary line in above-mentioned obtained final decision figure, and use based on sparse table
The fusion results that it is carried out merging acquisition transitional region by the multi-focus image fusion algorithm shown；
Step 8: fusion results based on decision diagram and transitional region fusion results to above-mentioned acquisition carry out mean operation acquisition
Final fusion image.
2. according to respectively two width multiple focussing images being carried out a wavelet decomposition, by each obtain described in claim 1.1
Low-frequency image is as its low scalogram picture, the half that size is input picture size of the lowest scalogram picture.
3., according to the method described in claim 1.2, use based on training dictionary K-SVD and OMP sparse coding algorithm the completeest
Standby sparse representation model carries out degree of rarefication expression to low scalogram picture, and constitutes two width corresponding degree of rarefication figure.
4., according to the method described in claim 1.3, first with smoothing windows technology, low for every width scalogram picture is divided into some
Size is the image subblock of a certain value；Secondly, use intelligibility measure method based on image block to weigh often and organize corresponding sub block
Definition is also compared to each other；Finally according to often organizing result of the comparison, each point value in sub-block is carried out assignment, it is thus achieved that corresponding two
Width definition score value figure.
5. according to the method described in claim 1.4, wherein, binary conversion treatment may be interpreted as: arranges suitable threshold value, by above-mentioned
In the definition score value figure obtained, the point more than this threshold value is entered as 1, is otherwise 0, is derived from corresponding bianry image；So
After according to set decision rule, bianry image is made further judgement, generate initial decision map.
6., according to the method described in claim 1.5, according to the position of the uncertain region that initial decision map is comprised, detect
Corresponding uncertain region in two width multiple focussing images of input, the uncertain region, draws further by each point in this region
It is divided into focusing and out of focus two class, generates final decision figure.
7., according to the method described in claim 1.6, according to above-mentioned obtained final decision figure, extract decision diagram Zhong Ge district
Clear area in the input image, territory, and these clear areas are constituted fusion results according to corresponding position grouping.
8., according to the method described in claim 1.7, arranging suitable value is upper and lower, left and right radius, is wrapped with final decision figure
Point centered by each point of the boundary line contained, delimiting out rectangular area is transitional region；And use multi-focus based on rarefaction representation
This region is carried out merging the fusion results obtaining transitional region by method.
9., according to the method described in claim 1.8, the fusion results based on decision diagram of above-mentioned acquisition is melted with transitional region
Close result and carry out the fusion image that mean operation acquisition is final.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610605703.6A CN106228528B (en) | 2016-07-29 | 2016-07-29 | A kind of multi-focus image fusing method based on decision diagram and rarefaction representation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610605703.6A CN106228528B (en) | 2016-07-29 | 2016-07-29 | A kind of multi-focus image fusing method based on decision diagram and rarefaction representation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106228528A true CN106228528A (en) | 2016-12-14 |
| CN106228528B CN106228528B (en) | 2019-09-27 |
Family
ID=57534529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610605703.6A Expired - Fee Related CN106228528B (en) | 2016-07-29 | 2016-07-29 | A kind of multi-focus image fusing method based on decision diagram and rarefaction representation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106228528B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680072A (en) * | 2017-11-01 | 2018-02-09 | 淮海工学院 | It is a kind of based on the positron emission fault image of depth rarefaction representation and the fusion method of MRI |
| CN108364273A (en) * | 2018-01-30 | 2018-08-03 | 中南大学 | A kind of method of multi-focus image fusion under spatial domain |
| CN108510465A (en) * | 2018-01-30 | 2018-09-07 | 西安电子科技大学 | The multi-focus image fusing method indicated based on consistency constraint non-negative sparse |
| CN109785282A (en) * | 2019-01-22 | 2019-05-21 | 厦门大学 | A kind of multi-focus image fusing method |
| CN111127375A (en) * | 2019-12-03 | 2020-05-08 | 重庆邮电大学 | Multi-focus image fusion method combining DSIFT and self-adaptive image blocking |
| CN111127380A (en) * | 2019-12-26 | 2020-05-08 | 云南大学 | Multi-focus image fusion method based on novel intuitionistic fuzzy similarity measurement technology |
| CN111582283A (en) * | 2020-04-27 | 2020-08-25 | 安创生态科技(深圳)有限公司 | Double-view image fusion method and device |
| CN112184646A (en) * | 2020-09-22 | 2021-01-05 | 西北工业大学 | Image fusion method based on gradient domain oriented filtering and improved PCNN |
| CN112233049A (en) * | 2020-12-14 | 2021-01-15 | 成都中轨轨道设备有限公司 | Image fusion method for improving image definition |
| CN112508828A (en) * | 2019-09-16 | 2021-03-16 | 四川大学 | Multi-focus image fusion method based on sparse representation and guided filtering |
| CN115393243A (en) * | 2022-10-28 | 2022-11-25 | 北京开运联合信息技术集团股份有限公司 | Multi-focus image fusion method and device and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120120221A1 (en) * | 2009-10-30 | 2012-05-17 | Siemens Aktiengesellschaft | Body Fluid Analyzing System and an Imaging Processing Device and Method for Analyzing Body Fluids |
| CN105678723A (en) * | 2015-12-29 | 2016-06-15 | 内蒙古科技大学 | Multi-focus image fusion method based on sparse decomposition and differential image |
-
2016
- 2016-07-29 CN CN201610605703.6A patent/CN106228528B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120120221A1 (en) * | 2009-10-30 | 2012-05-17 | Siemens Aktiengesellschaft | Body Fluid Analyzing System and an Imaging Processing Device and Method for Analyzing Body Fluids |
| CN105678723A (en) * | 2015-12-29 | 2016-06-15 | 内蒙古科技大学 | Multi-focus image fusion method based on sparse decomposition and differential image |
Non-Patent Citations (1)
| Title |
|---|
| 李华锋: "多聚焦图像像素级融合方法研究", 《中国博士学位论文全文数据库》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680072A (en) * | 2017-11-01 | 2018-02-09 | 淮海工学院 | It is a kind of based on the positron emission fault image of depth rarefaction representation and the fusion method of MRI |
| CN108364273A (en) * | 2018-01-30 | 2018-08-03 | 中南大学 | A kind of method of multi-focus image fusion under spatial domain |
| CN108510465A (en) * | 2018-01-30 | 2018-09-07 | 西安电子科技大学 | The multi-focus image fusing method indicated based on consistency constraint non-negative sparse |
| CN108510465B (en) * | 2018-01-30 | 2019-12-24 | 西安电子科技大学 | Multi-focus image fusion method based on consistency constraint non-negative sparse representation |
| CN108364273B (en) * | 2018-01-30 | 2022-02-25 | 中南大学 | Method for multi-focus image fusion in spatial domain |
| CN109785282A (en) * | 2019-01-22 | 2019-05-21 | 厦门大学 | A kind of multi-focus image fusing method |
| CN112508828A (en) * | 2019-09-16 | 2021-03-16 | 四川大学 | Multi-focus image fusion method based on sparse representation and guided filtering |
| CN111127375A (en) * | 2019-12-03 | 2020-05-08 | 重庆邮电大学 | Multi-focus image fusion method combining DSIFT and self-adaptive image blocking |
| CN111127380A (en) * | 2019-12-26 | 2020-05-08 | 云南大学 | Multi-focus image fusion method based on novel intuitionistic fuzzy similarity measurement technology |
| CN111127380B (en) * | 2019-12-26 | 2023-05-23 | 云南大学 | Multi-focus image fusion method based on intuitionistic fuzzy similarity measurement technology |
| CN111582283A (en) * | 2020-04-27 | 2020-08-25 | 安创生态科技(深圳)有限公司 | Double-view image fusion method and device |
| CN112184646A (en) * | 2020-09-22 | 2021-01-05 | 西北工业大学 | Image fusion method based on gradient domain oriented filtering and improved PCNN |
| CN112184646B (en) * | 2020-09-22 | 2022-07-29 | 西北工业大学 | Image fusion method based on gradient domain oriented filtering and improved PCNN |
| CN112233049A (en) * | 2020-12-14 | 2021-01-15 | 成都中轨轨道设备有限公司 | Image fusion method for improving image definition |
| CN115393243A (en) * | 2022-10-28 | 2022-11-25 | 北京开运联合信息技术集团股份有限公司 | Multi-focus image fusion method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106228528B (en) | 2019-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106228528B (en) | A kind of multi-focus image fusing method based on decision diagram and rarefaction representation | |
| CN109377448B (en) | Face image restoration method based on generation countermeasure network | |
| CN106339998B (en) | Multi-focus image fusing method based on contrast pyramid transformation | |
| CN110348445B (en) | Instance segmentation method fusing void convolution and edge information | |
| CN109299274B (en) | Natural scene text detection method based on full convolution neural network | |
| JP6866495B2 (en) | Image quality evaluation method and image quality evaluation system | |
| CN111242238B (en) | RGB-D image saliency target acquisition method | |
| CN111126202A (en) | Optical remote sensing image target detection method based on void feature pyramid network | |
| CN110490158B (en) | Robust face alignment method based on multistage model | |
| CN104077761B (en) | Multi-focus image fusion method based on self-adaption sparse representation | |
| CN113160062B (en) | Infrared image target detection method, device, equipment and storage medium | |
| CN114359526B (en) | Cross-domain image style migration method based on semantic GAN | |
| CN110689599A (en) | 3D visual saliency prediction method for generating countermeasure network based on non-local enhancement | |
| CN108573222A (en) | The pedestrian image occlusion detection method for generating network is fought based on cycle | |
| CN110472634A (en) | Change detecting method based on multiple dimensioned depth characteristic difference converged network | |
| CN106447640B (en) | Multi-focus image fusing method and device based on dictionary learning, rotation guiding filtering | |
| CN110457515A (en) | The method for searching three-dimension model of the multi-angle of view neural network of polymerization is captured based on global characteristics | |
| CN105488759B (en) | A kind of image super-resolution rebuilding method based on local regression model | |
| CN114972312A (en) | Improved insulator defect detection method based on YOLOv4-Tiny | |
| CN114241422A (en) | Student classroom behavior detection method based on ESRGAN and improved YOLOv5s | |
| Zhu et al. | Towards automatic wild animal detection in low quality camera-trap images using two-channeled perceiving residual pyramid networks | |
| Luo et al. | Bi-GANs-ST for perceptual image super-resolution | |
| CN109766918A (en) | Conspicuousness object detecting method based on the fusion of multi-level contextual information | |
| Zhao et al. | Detecting deepfake video by learning two-level features with two-stream convolutional neural network | |
| CN114972332A (en) | Bamboo laminated wood crack detection method based on image super-resolution reconstruction network |
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: 20190927 Termination date: 20200729 |