CN112651469A - Infrared and visible light image fusion method and system - Google Patents
Infrared and visible light image fusion method and system Download PDFInfo
- Publication number
- CN112651469A CN112651469A CN202110089292.0A CN202110089292A CN112651469A CN 112651469 A CN112651469 A CN 112651469A CN 202110089292 A CN202110089292 A CN 202110089292A CN 112651469 A CN112651469 A CN 112651469A
- Authority
- CN
- China
- Prior art keywords
- frequency coefficient
- low
- coefficient information
- image
- frequency
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000007500 overflow downdraw method Methods 0.000 title claims abstract description 41
- 230000004927 fusion Effects 0.000 claims abstract description 68
- 238000000034 method Methods 0.000 claims abstract description 42
- 230000009466 transformation Effects 0.000 claims abstract description 41
- 238000004590 computer program Methods 0.000 claims description 13
- 239000000126 substance Substances 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000000354 decomposition reaction Methods 0.000 abstract description 23
- 238000013527 convolutional neural network Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 238000011426 transformation method Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06V10/52—Scale-space analysis, e.g. wavelet analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
- Studio Devices (AREA)
Abstract
The invention discloses an infrared and visible light image fusion method and a system, which are used for acquiring an infrared image and a visible light image; respectively carrying out multi-scale transformation on the infrared image and the visible light image to obtain first low-frequency coefficient information, first high-frequency coefficient information, second low-frequency coefficient information and second high-frequency coefficient information; determining third low-frequency coefficient information according to the first low-frequency coefficient information and the second low-frequency coefficient information based on a global contrast sequence maximum principle; determining third high-frequency coefficient information according to the first high-frequency coefficient information and the second high-frequency coefficient information; generating a fusion image according to the third low-frequency coefficient information and the third high-frequency coefficient information; the method and the device provided by the invention have the advantages that the inter-pixel relative value information is fully utilized, and the inter-pixel relative value in the low-frequency coefficient information is obtained by directly carrying out multi-scale transformation decomposition on the infrared image and the visible light image and then is fused, so that the contrast of the fused image is improved, and more detailed information in the image is reserved.
Description
Technical Field
The invention belongs to the technical field of image fusion, and particularly relates to an infrared and visible light image fusion method and system.
Background
Image fusion is a technique for fusing information contained in images acquired by multiple imaging systems of the same scene into one image. The fused image is more suitable for human eyes and machine perception than the image acquired by a single imaging system, so the fused image is widely applied to the fields of industry, medicine, military affairs and the like. Because the images collected by the infrared imaging system and the visible light imaging system have the information complementary characteristic and the imaging equipment is relatively simple and easy to obtain, the fusion of the infrared image and the visible light image has more superiority, and the method is an important way for improving the application values of video monitoring, medical diagnosis, target detection and the like.
The fusion of infrared images and visible light images generally adopts multi-scale transformation consistent with the visual characteristics of human eyes to represent image information, and generally comprises three steps: 1) carrying out multi-scale transformation decomposition on the input infrared image and the input visible light image respectively to obtain respective multi-scale representation coefficients; 2) fusing the multi-scale representation coefficients of the infrared image and the visible light image to obtain the multi-scale representation coefficient of the fused image; 3) and applying multi-scale inverse transformation to the multi-scale representation coefficient of the fused image to obtain the fused image. The transformation coefficient fusion in the step 2), especially the fusion of low-frequency coefficients, is the key of image fusion, because the low-frequency coefficients contain more information of the source image.
In the early low-frequency coefficient fusion method, a method of selecting the maximum value of the low-frequency coefficients of the infrared image and the visible light image or a simple equal-coefficient weighted average method of the low-frequency coefficients of the infrared image and the visible light image is adopted corresponding to each pixel position, and the methods are easy to introduce artifacts or lose image detail information in the fused image. Therefore, how to optimize the fusion weight value for assigning the weighted average of the low frequency coefficients becomes the focus of recent discussion of image fusion researchers. For example, chen and others propose to allocate a fusion weight based on the energy ratio of the adjacent regions of the infrared image and the low-frequency coefficient of the visible light image of each pixel, and even some researchers propose to allocate a fusion weight of the low-frequency coefficient of the infrared image and the low-frequency coefficient of the visible light image by using a CNN (convolutional neural network) network.
The image fusion method based on the adjacent region energy ratio and the CNN considers the mutual influence among the pixels when fusing the low-frequency coefficients obtained by the multi-scale decomposition of the source image, and distributes the fusion weight of the weighted average of the low-frequency coefficients by referring to the information of the adjacent pixels. Therefore, the contrast of the obtained fusion image is improved relative to the early fusion algorithm. However, the use of relative value information between adjacent pixels is indirect, and the fusion is still an operation on the absolute values of the low-frequency coefficients from different source images at a single pixel position.
In fact, the above method still has the problem of loss of detail information, so that the contrast of the fused image still needs to be further improved. In addition, although the CNN-based image fusion method can utilize relative information between adjacent pixels through a large amount of learning depth in a network, the method is limited in that fusion image target data is difficult to obtain during network model learning, fusion of low-frequency coefficients of the whole region is difficult to achieve, and the contrast of the fusion image is low.
Disclosure of Invention
The invention aims to provide an infrared and visible light image fusion method and system, which can improve the contrast of a fusion image by fusing relative values between low-frequency coefficient pixels after multi-scale transformation and decomposition of an infrared image and a visible light image.
The invention adopts the following technical scheme: an infrared and visible light image fusion method comprises the following steps:
acquiring an infrared image and a visible light image;
respectively carrying out multi-scale transformation on the infrared image and the visible light image to obtain first low-frequency coefficient information, first high-frequency coefficient information, second low-frequency coefficient information and second high-frequency coefficient information;
determining third low-frequency coefficient information according to the first low-frequency coefficient information and the second low-frequency coefficient information based on a global contrast sequence maximum principle;
determining third high-frequency coefficient information according to the first high-frequency coefficient information and the second high-frequency coefficient information;
and determining a fused image according to the third low-frequency coefficient information and the third high-frequency coefficient information.
Further, determining the third low frequency coefficient information according to the first low frequency coefficient information and the second low frequency coefficient information includes:
generating a first global contrast sequence according to the first low-frequency coefficient information, and generating a second global contrast sequence according to the second low-frequency coefficient information;
determining a third global contrast sequence according to the first global contrast sequence and the second global contrast sequence;
and determining third low-frequency coefficient information according to the third global contrast sequence.
Further, determining third low frequency coefficient information from the third global contrast sequence comprises:
calculating third low-frequency coefficient intermediate information according to the third global contrast sequence;
determining third low-frequency coefficient information according to the third low-frequency coefficient intermediate information;
the calculation method of the third low-frequency coefficient intermediate information comprises the following steps:
setting the coefficient of at least one pixel in the third low-frequency coefficient intermediate information as a preset value, and minimizingCalculating to obtain third low-frequency coefficient intermediate information; (ii) a
Wherein the content of the first and second substances,is the third low-frequency coefficient intermediate information, J is the total number of layers of multi-scale transformation, M is the total number of pixels in the fused image, x and y are pixel ordinal numbers in the image, and x is less than y,is the coefficient value of the xth pixel in the third low frequency coefficient intermediate information,the coefficient value of the y-th pixel in the third low-frequency coefficient intermediate information, dF(x, y) is the contrast value between the x-th pixel and the y-th pixel in the fused image, and F is the fused image.
Further, determining the third low-frequency coefficient information according to the third low-frequency coefficient intermediate coefficient includes:
to pairThe coefficient value of each pixel in the third low-frequency coefficient information is compensated to obtain the coefficient value of each pixel in the third low-frequency coefficient information.
Further, determining the third high frequency coefficient information from the first high frequency coefficient information and the second high frequency coefficient information includes:
Wherein the content of the first and second substances,for the coefficient value of the mth pixel in the high-frequency subgraph of the l-th directional subband of the fusion image j-th layer multi-scale transformation,for the coefficient value, w of the m-th pixel in the high-frequency subgraph of the jth layer multi-scale transformation ith direction sub-band of the infrared imagemIs composed ofThe fusion weight of (a) is calculated,the coefficient value of the mth pixel in the high-frequency subgraph of the ith directional subband is subjected to multi-scale transformation on the jth layer of the visible light image;
According to a calculation model andis fused with the weight wmDetermining a high frequency coefficient of each pixel in the third high frequency coefficient informationCounting and obtaining third high-frequency coefficient information.
Wherein R ismIs a local region of the m-th pixel, RnIs a local region of the nth pixel, n belongs to {1, 2.., M },is a local region RnThe fusion weight of (2).
respectively calculating the local area variance of the nth pixel in the infrared high-frequency subgraph corresponding to the first high-frequency coefficient information and the visible light high-frequency subgraph corresponding to the second high-frequency coefficient informationAnd
The other technical scheme of the invention is as follows: an infrared and visible image fusion system comprising the steps of:
the acquisition module is used for acquiring an infrared image and a visible light image;
the multi-scale transformation module is used for respectively carrying out multi-scale transformation on the infrared image and the visible light image to obtain first low-frequency coefficient information, first high-frequency coefficient information, second low-frequency coefficient information and second high-frequency coefficient information;
the first determining module is used for determining third low-frequency coefficient information according to the first low-frequency coefficient information and the second low-frequency coefficient information based on a global contrast sequence maximum principle;
the second determining module is used for determining third high-frequency coefficient information according to the first high-frequency coefficient information and the second high-frequency coefficient information;
and the generating module is used for generating a fusion image according to the third low-frequency coefficient information and the third high-frequency coefficient information.
The other technical scheme of the invention is as follows: an infrared and visible image fusion system comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing an infrared and visible image fusion method according to any one of the preceding claims when executing the computer program.
The other technical scheme of the invention is as follows: a computer-readable storage medium, in which a computer program is stored, wherein the computer program, when executed by a processor, implements any of the above-mentioned infrared and visible image fusion methods.
The invention has the beneficial effects that: the method and the device provided by the invention have the advantages that the inter-pixel relative value information is fully utilized, the inter-pixel relative value in the low-frequency coefficient information is obtained by directly carrying out multi-scale transformation decomposition on the infrared image and the visible light image, the fusion is carried out, the contrast of the fused image is improved, more detailed information in the image is reserved, meanwhile, the implementation is simple compared with a CNN-based fusion method, and the application scene is not constrained.
Drawings
FIG. 1 is a flowchart of a method for fusing infrared and visible light images according to an embodiment of the present invention;
FIG. 2 is a flow chart illustrating the detailed steps in an embodiment of the present invention;
FIG. 3 is a diagram illustrating an image decomposition process by NSCT according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating an index sequence of 5 rows and 5 columns of pixels according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of an infrared and visible light input source image sample 1 in a validation embodiment of the present invention;
FIG. 6 is a fused image output from the source image of FIG. 5 after passing through different fusion parties;
FIG. 7 is a schematic illustration of an infrared and visible light input source image sample 2 in a validation embodiment of the present invention;
FIG. 8 is a fused image obtained by different fusion methods of the source image in FIG. 7;
FIG. 9 is a schematic structural diagram of an infrared and visible light image fusion system according to an embodiment of the present invention;
fig. 10 is a schematic structural diagram of an infrared and visible light image fusion system according to another embodiment of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
In the image fusion process, the human eye is visually sensitive not to the absolute value of a single pixel but to the relative value between pixels. Therefore, the inter-pixel relative value information is not fully utilized, and the problem of loss of detail information still exists, so that the contrast of the fused image needs to be further improved. In addition, although the image fusion method based on the CNN can utilize relative information between adjacent pixels through a great amount of learning depth of a network, the method is limited in that fusion image target data is difficult to obtain during network model learning, and can only be applied to fusion of low-frequency coefficients of partial local regions where an infrared image and a visible light image are very similar at present.
The purpose of image fusion is to obtain new information which is helpful for subsequent application decision by improving the sensitivity of human vision to fused images. Since the image fusion technology is still a new research field at present, it is still necessary to continuously experiment and prove how to improve the sensitivity of human vision to the fused image by using the relative value information between pixels.
The invention aims to provide an infrared and visible light image fusion method and system, which are used for directly fusing the relative values among the pixels of the low-frequency coefficient obtained by carrying out multi-scale transformation decomposition on an infrared image and a visible light image, so that the contrast of the fused image is improved.
Specifically, the embodiment of the invention discloses an infrared and visible light image fusion method, as shown in fig. 1, comprising the following steps: step S110, acquiring an infrared image and a visible light image; step S120, respectively carrying out multi-scale transformation on the infrared image and the visible light image to obtain first low-frequency coefficient information, first high-frequency coefficient information, second low-frequency coefficient information and second high-frequency coefficient information; step S130, determining third low-frequency coefficient information according to the first low-frequency coefficient information and the second low-frequency coefficient information based on a global contrast sequence maximum principle; step S140, determining third high-frequency coefficient information according to the first high-frequency coefficient information and the second high-frequency coefficient information; and S150, generating a fusion image according to the third low-frequency coefficient information and the third high-frequency coefficient information.
The method and the device provided by the invention have the advantages that the inter-pixel relative value information is fully utilized, the inter-pixel relative value in the low-frequency coefficient information is obtained by directly carrying out multi-scale transformation decomposition on the infrared image and the visible light image, the fusion is carried out, the contrast of the fused image is improved, more detailed information in the image is reserved, meanwhile, the implementation is simple compared with a CNN-based fusion method, and the application scene is not constrained.
As a specific implementation manner, in the more detailed method flowchart of this embodiment, as shown in fig. 2, the infrared image and the visible light image acquired in step S110 may be a pair of images corresponding to the same scene respectively acquired by the infrared acquisition device and the visible light acquisition device, which are used as input images. Therefore, the size and the content of the collected image can be ensured to be corresponding, and the subsequent image fusion process can be facilitated. Preferably, the two types of image capturing devices capture images simultaneously at the same time to avoid different changes of image information at different times.
In step S120, the input infrared image I is processedAAnd a visible light image IBAnd respectively carrying out multi-scale transformation to obtain a low-frequency coefficient subgraph and a plurality of high-frequency coefficient subgraphs which are respectively decomposed in a multi-scale mode, wherein the coefficient value corresponding to each pixel in the low-frequency coefficient subgraph is the low-frequency coefficient, and all the low-frequency coefficients form low-frequency coefficient information corresponding to the low-frequency coefficient subgraph. And the coefficient value corresponding to each pixel in the high-frequency coefficient subgraph is a high-frequency coefficient, and each high-frequency coefficient subgraph corresponds to high-frequency coefficient information.
As for the multi-scale transform method, any one of pyramid transform, CT (curve transform), WT (wavelet transform), NSCT (non-subsampled contourlet transform), and other multi-scale transform methods may be selected, and in the present embodiment, the NSCT method is selected.
The NSCT is a multi-scale and multi-direction image decomposition method, and consists of two core modules, namely NSPFB (non-subsampled pyramid filter bank) and NSDFB (non-subsampled directional filter bank).
The image decomposition process by NSCT is shown in fig. 3, in which NSPFB is a set of cascaded NSPFs (non-subsampled pyramid filters) for completing the decomposition of the image in multiple scales. Each NSPF is a two-channel non-down sampling two-dimensional filter composed of a low-pass decomposition module and a high-pass decomposition module, and two sub-images with the same size as the input image are output after the input image is subjected to NSPF filtering decomposition and respectively correspond to the low-frequency part and the high-frequency part of the input image. And the low-frequency part output by each NSPF in the cascaded group of NSPFs is further decomposed as an input image of the next scale NSPF, and the high-frequency part obtains high-frequency coefficient subgraphs in multiple directions corresponding to the current scale through one NSDFB. The number of cascades of the NSPFBs, i.e., the number of layers J of the image multi-scale decomposition in the NSPFB, is called a scale parameter of multi-scale transformation, and may be configured as any integer greater than or equal to 1, and is preferably 4 in this embodiment.
NSDFB is a groupAnd the multi-directional non-down-sampling directional filter is used for completing the multi-directional decomposition of the high-frequency image output by each stage of NSPF. The NSDFB effectively divides the signal into L-2 through the decomposition of a k-layer tree structurekThe frequency band of each directional sub-band is divided into wedges. The direction parameter k of the NSDFB can be configured as any integer greater than or equal to 1, and values can be different under different scales, preferably between 2 and 4.
The non-down sampling contour transformation is adopted, the transformed scale and direction parameter adopt J-4 scales, and the direction parameter k of each scalejThe example illustrates the multi-scale transform decomposition results with j being 1,2,3, and 4. Infrared image IAWhen J is 4, kjNSCT decomposition of 3, j 1,2,3,4 yields a low-frequency subgraph and 4 × 2332 high frequency subgraphsWherein the content of the first and second substances, representing an infrared image IAIn a high-frequency subgraph in the direction l of the dimension j (i.e. the j-th layer), the value corresponding to each pixel position in the subgraph is a high-frequency coefficient. Low-frequency subgraph of infrared imageAnd the value corresponding to each pixel position in the subgraph is the infrared image low-frequency coefficient. Infrared image IAThe subgraph set obtained by multi-scale transformation decomposition can be represented as
Likewise, visible light image IBThe set of subgraphs decomposed by the multi-scale transformation can be represented asWherein the content of the first and second substances,as a visible light image IBThe low-frequency sub-graph of (a),as a visible light image IBHigh frequency subgraph in the direction l of the dimension j.
In order to facilitate subsequent calculation processing, the image pixel indexes are ordered line by line from top to bottom, and each line is numbered sequentially from left to right. At this timeCan be expressed as Can be expressed asM denotes the total number of pixels of the image and x is the ordinal number of the pixel. In the same way as above, the first and second,can be expressed as Can be expressed as
In step S130, determining the third low frequency coefficient information from the first low frequency coefficient information and the second low frequency coefficient information includes:
generating a first global contrast sequence according to the first low-frequency coefficient information, and generating a second global contrast sequence according to the second low-frequency coefficient information; and determining a third global contrast sequence according to the first global contrast sequence and the second global contrast sequence.
The global contrast sequence of the infrared image low-frequency subgraph is obtained in step 120The contrast value between each pixel in the sequence and the subsequent pixels is given by { d }A(x,y)}x=1,...,M-1;y=x+1,x+2,...,MWherein the contrast value between two pixels is the difference between the low frequency coefficient corresponding to the previous pixel and the low frequency coefficient corresponding to the next pixel, i.e. the contrast value isAccordingly, the global contrast sequence of the low-frequency subgraph of the visible-light image is { d }B(x,y)}x=1,...,M-1;y=x+1,x+2,...,MWherein, in the step (A),
when determining the third global contrast sequence, the embodiment obtains the global contrast sequence of the desired low-frequency subgraph of the fused image based on the maximum principle of the global contrast sequence of the fused image, { dF(x,y)}x=1,...,M-1;y=x+1,x+2,...,MWherein, in the step (A),
that is to say, the contrast value between two pixels in the global contrast sequence of the low-frequency subgraph of the fused image is determined by the contrast value of the low-frequency subgraph of the infrared image corresponding to the two pixels and the contrast value of the low-frequency subgraph of the visible image, and the contrast value with the maximum absolute value in the contrast values of the low-frequency subgraph of the infrared image and the low-frequency subgraph of the visible image is selected as the contrast value corresponding to the two pixels in the global contrast sequence of the low-frequency subgraph of the fused image.
And determining third low-frequency coefficient information according to the third global contrast sequence. In this embodiment, the low-frequency coefficient of the fused image, i.e. the third low-frequency coefficient, is calculated according to the global contrast sequence of the low-frequency subgraph of the fused image, and the specific solution is to make the objective functionMinimal fused image low-frequency sub-image sequenceWherein the value of each pixel xI.e. the low frequency coefficients of the fused image. The optimization solving process with the minimum objective function of the above formula can be preferably realized by adopting a Gaussian elimination method.
More specifically, the method for determining the third low-frequency coefficient information by using the intermediate information includes the following steps:
since the global contrast sequence defines the relative value between pixels and does not define the reference value, there are an infinite number of possible solutions to satisfy the above equation. To facilitate the solution, it may be assumed that the coefficient of at least one pixel is a predetermined value, in this embodiment, the following assumptions may be made:orOr assume thatOrOr assume that the low frequency coefficients for any other pixel location are arbitrary.
After the above setting, the Gaussian elimination method is usedOver minimizationSolved to obtainWherein the content of the first and second substances,is the third low-frequency coefficient intermediate information, J is the total number of layers of multi-scale transformation, M is the total number of pixels in the fused image, x and y are pixel ordinal numbers in the image, and x is less than y,for the coefficient median of the xth pixel in the fused image,for the coefficient median of the y-th pixel in the fused image, dF(x, y) is the contrast value between the x-th pixel and the y-th pixel in the fused image, and F is the fused image.
However, the sequences obtained under the above assumptionsIt is not necessarily the best fused image low frequency coefficient, therefore, it is also necessary toFurther adjustments are made.
In the embodiment of the invention, a simple further adjustment mode is based on prior information that the average of the low-frequency coefficient mean value of the fusion image is consistent with the average of the low-frequency coefficient mean values of the infrared image and the visible light image, namelyTo pairMiddle pairCompensating the difference value of the mean value according to the coefficient of each pixel so as to obtain the low-frequency subgraph of the expected fused imageWherein
Namely toThe coefficient value of each pixel in the third low-frequency coefficient information is compensated, the compensation value is delta, and then the coefficient value of each pixel in the third low-frequency coefficient information is obtained.
As described above, the process of fusing the first low-frequency coefficient information and the second low-frequency coefficient information into the third low-frequency coefficient information is completed.
The process of determining the third high frequency coefficient information from the first high frequency coefficient information and the second high frequency coefficient information in step S140 will be described. The infrared image high-frequency subgraph obtained by the decomposition in the step S120And visible image high frequency subgraphAnd calculating by using a weighted average method to obtain a high-frequency subgraph corresponding to the fused image, wherein the distribution of the fused weight of the weighted average of the high-frequency coefficients is calculated based on the local variance maximum principle.
The specific implementation mode comprises the following steps:
constructing a third high-frequency coefficient information calculation modelM ∈ {1,2,.., M }; determiningThe fusion weight of (2); according to a calculation model anddetermining the high-frequency coefficient of each pixel in the third high-frequency coefficient information, and obtaining the third high-frequency coefficient information. Wherein the content of the first and second substances,for the coefficient value of the mth pixel in the high-frequency subgraph of the l-th directional subband of the fusion image j-th layer multi-scale transformation,for the coefficient value, w of the m-th pixel in the high-frequency subgraph of the jth layer multi-scale transformation ith direction sub-band of the infrared imagemIs composed ofThe fusion weight of (1) is not less than 0 and not more than wm≤1,The coefficient value of the mth pixel in the high-frequency subgraph of the ith directional subband is subjected to multi-scale transformation on the jth layer of the visible light image;
firstly, determining a pixel set covered by a local region corresponding to each pixel in the high-frequency subgraph, and determining a local region R of the mth pixel in the infrared imagem. The local area window parameter W is set, W being a positive integer larger than 0, preferably a number between 1 and 3. Local region R corresponding to pixel m in high frequency sub-imagemThe image area is an image area covered by a rectangle having a width of (2 × W +1) and a height of (2 × W +1) with the pixel m as the center. Taking W as 1 as an example, the width of the local region is (2 × 1+1) 3, and the height is (2 × 1+1) 3. Local region R corresponding to pixel m in high frequency sub-imagemThis means that 3 × 3 — 9 pixels centered on pixel m are covered. In particular, for pixels located at the sides and corners of the image, the number of pixels covered by the local area is smaller because there is no other pixel on the left or right side or on the top or bottom side.
Taking an example of a 5-row and 5-column image, as shown in fig. 4, the image pixel indices are numbered 1,2, …, 25 sequentially in a row-by-row serial order from top to bottom and left to right. Then, the local region R corresponding to the pixel m 13m=13Covering 9 pixels centered at 13 {7, 8, 9, 12, 13, 14, 17, 18, 19 }. Still taking the image with 5 rows and 5 columns and W equal to 1 as an example, the pixel m equal to 1 is located at the upper left corner of the image, and there are no other pixels at the upper edge and the left edge, so that the local region R corresponding to the pixel m equal to 1 corresponds tom=1Only 4 pixels 1,2, 6, 7 are covered.
The fusion weight of the local region of the pixel m is recorded asComputing fused image high frequency subgraph using computational modelLocal area variance of middle pixel mWherein the content of the first and second substances, |Rmi denotes a local region RmThe total number of pixels covered.
For the nth pixel, the local area variance thereof is maximizedObtaining the local region fusion weight of the nth pixelThe process is as follows: respectively calculating the local area variance of the nth pixel in the infrared high-frequency subgraph corresponding to the first high-frequency coefficient information and the visible light high-frequency subgraph corresponding to the second high-frequency coefficient informationAndwherein the content of the first and second substances,
And after the third low-frequency coefficient information and the third high-frequency coefficient information are obtained, the third low-frequency coefficient information and the third high-frequency coefficient information are used as a low-frequency coefficient and a high-frequency coefficient of the fused image, and then multi-scale inverse transformation is applied to obtain the fused image. The multi-scale inverse transformation method and the method adopted in step S120The multi-scale transformation method corresponds. That is, when the non-downsampling contour transformation is adopted in the corresponding step S120, the contour transformation is performed on the contour information obtained in the steps S130 and S140And applying non-downsampling contour inverse transformation to obtain a fused image.
In the embodiment of the invention, a global contrast sequence is defined for representing the relative value information of the inter-pixel low-frequency coefficient in the image multi-scale transformation low-frequency sub-image; the image fusion operation is the fusion of the relative value information of the low-frequency coefficient between pixels in the infrared image low-frequency subgraph and the visible image low-frequency subgraph, namely the fusion of the defined global contrast sequence of the infrared image low-frequency subgraph and the global contrast sequence of the visible image low-frequency subgraph. A simple method for solving the absolute value of the low-frequency coefficient of the fusion image by using the relative value information among the low-frequency coefficients of the fusion image is provided.
Verification of the examples:
FIG. 5 is an infrared and visible light input source image sample 1, (a) is an input infrared source image, and (b) is an input visible source image. FIG. 6 is a fused image obtained from the source image of FIG. 5 by different fusion methods, (a) is a low-frequency fusion method using early direct equal-coefficient weighted averaging of multi-scale decomposition low-frequency coefficients; (b) is a low-frequency fusion method based on adjacent region energy ratio proposed by Chen and the like; (c) adopting a low-frequency fusion method based on a CNN model; (d) the fusion method of the present invention is used. It can be seen that the fused image output by the low-frequency fusion method with equal coefficient weighted average has low contrast, and particularly, the sky and the branch region lose the details of a plurality of branches. The low-frequency fusion method based on the adjacent region energy ratio and the CNN has the advantages that the contrast of a fusion image is greatly improved compared with the early low-frequency fusion method based on the equal coefficient weighted average, but the problem of detail loss still exists, for example, the details of a fused image peaked house output by the low-frequency fusion method based on the adjacent region energy ratio are lost, and the details of a fused image cabin output by the low-frequency fusion method based on the CNN and a skirt-wearing person are lost. The fusion image detail output by the fusion method is relatively better preserved.
Fig. 7 is an infrared and visible light input source image sample 2, (a) is an input infrared source image, and (b) is an input visible light source image. FIG. 8 is a fused image obtained from the source image of FIG. 7 by different fusion methods, (a) is a low frequency fusion method using early direct equal coefficient weighted averaging of the multi-scale decomposition low frequency coefficients; (b) is a low-frequency fusion method based on adjacent region energy ratio proposed by Chen and the like; (c) adopting a low-frequency fusion method based on a CNN model; (d) the fusion method of the present invention is used. The fusion method can also be used for outputting the fusion image with the best fusion result in different fusion methods, and better accords with the visual effect of human eyes.
Another embodiment of the present invention discloses an infrared and visible light image fusion system, as shown in fig. 9, comprising the steps of:
an acquiring module 210, configured to acquire an infrared image and a visible light image; the multi-scale transformation module 220 is configured to perform multi-scale transformation on the infrared image and the visible light image respectively to obtain first low-frequency coefficient information, first high-frequency coefficient information, second low-frequency coefficient information, and second high-frequency coefficient information; a first determining module 230, configured to determine, based on a global contrast sequence maximization principle, third low-frequency coefficient information according to the first low-frequency coefficient information and the second low-frequency coefficient information; a second determining module 240, configured to determine third high frequency coefficient information according to the first high frequency coefficient information and the second high frequency coefficient information; and a generating module 250, configured to generate a fused image according to the third low-frequency coefficient information and the third high-frequency coefficient information.
It should be noted that, because the contents of information interaction, execution process, and the like between the modules are based on the same concept as the method embodiment of the present invention, specific functions and technical effects thereof may be referred to specifically in the method embodiment section, and are not described herein again.
It will be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely illustrated, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the apparatus is divided into different functional modules to perform all or part of the above described functions. Each functional module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, the specific names of the functional modules are only for convenience of distinguishing from each other and are not used for limiting the protection scope of the present invention. The specific working process of the modules in the system may refer to the corresponding process in the foregoing method embodiment, and is not described herein again.
Another embodiment of the present invention further discloses an infrared and visible light image fusion system, as shown in fig. 10, which includes a memory 31, a processor 32, and a computer program 33 stored in the memory 31 and executable on the processor 32, and when the computer program is executed by the processor, the processor implements an infrared and visible light image fusion method of any one of the above.
Another embodiment of the present invention further discloses a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program implements any one of the above-mentioned infrared and visible light image fusion methods.
The computer readable medium may include at least: any entity or device capable of carrying computer program code to a photographing apparatus/terminal apparatus, a recording medium, computer Memory, Read-Only Memory (ROM), random-access Memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In certain jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and patent practice.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
Claims (10)
1. An infrared and visible light image fusion method is characterized by comprising the following steps:
acquiring an infrared image and a visible light image;
respectively carrying out multi-scale transformation on the infrared image and the visible light image to obtain first low-frequency coefficient information, first high-frequency coefficient information, second low-frequency coefficient information and second high-frequency coefficient information;
determining third low-frequency coefficient information according to the first low-frequency coefficient information and the second low-frequency coefficient information based on a global contrast sequence maximum principle;
determining third high-frequency coefficient information according to the first high-frequency coefficient information and the second high-frequency coefficient information;
and generating a fusion image according to the third low-frequency coefficient information and the third high-frequency coefficient information.
2. The method of claim 1, wherein determining third low frequency coefficient information based on the first low frequency coefficient information and the second low frequency coefficient information comprises:
generating a first global contrast sequence according to the first low-frequency coefficient information, and generating a second global contrast sequence according to the second low-frequency coefficient information;
determining a third global contrast sequence according to the first global contrast sequence and the second global contrast sequence;
and determining third low-frequency coefficient information according to the third global contrast sequence.
3. The method of claim 2, wherein determining third low frequency coefficient information from the third global contrast sequence comprises:
calculating third low-frequency coefficient intermediate information according to the third global contrast sequence;
determining third low-frequency coefficient information according to the third low-frequency coefficient intermediate information;
the calculation method of the third low-frequency coefficient intermediate information comprises the following steps:
setting the coefficient of at least one pixel in the third low-frequency coefficient intermediate information as a preset value, and minimizingCalculating to obtain the third low-frequency coefficient intermediate information;
wherein the content of the first and second substances,is the third low-frequency coefficient intermediate information, J is the total number of layers of multi-scale transformation, M is the total number of pixels in the fused image, x and y are pixel ordinal numbers in the image, and x is less than y,is the coefficient value of the xth pixel in the third low frequency coefficient intermediate information,the coefficient value of the y-th pixel in the third low-frequency coefficient intermediate information, dF(x, y) is the contrast value between the x-th pixel and the y-th pixel in the fused image, and F is the fused image.
4. The method of claim 2 or 3, wherein determining the third low frequency coefficient information based on the third low frequency coefficient intermediate information comprises:
5. The method of claim 4, wherein determining the third high frequency coefficient information based on the first high frequency coefficient information and the second high frequency coefficient information comprises:
Wherein the content of the first and second substances,for the coefficient value of the mth pixel in the high-frequency subgraph of the l-th directional subband of the fusion image j-th layer multi-scale transformation,the coefficient value, w, of the mth pixel in the high-frequency subgraph of the jth layer multi-scale transformation ith direction sub-band of the infrared imagemIs composed ofThe fusion weight of (a) is calculated,performing multi-scale transformation on the coefficient value of the mth pixel in the high-frequency subgraph of the ith directional subband of the jth layer of the visible light image;
6. The method of claim 4, wherein determining comprises determining the image of the subject based on the image of the subjectThe fusion weight value comprises:
7. The infrared and visible light image fusion method of claim 6, wherein the local region fusion weightsThe calculating method comprises the following steps:
respectively calculating the local area variance of the nth pixel in the infrared high-frequency subgraph corresponding to the first high-frequency coefficient information and the visible light high-frequency subgraph corresponding to the second high-frequency coefficient informationAnd
8. An infrared and visible image fusion system comprising the steps of:
the acquisition module is used for acquiring an infrared image and a visible light image;
the multi-scale transformation module is used for respectively carrying out multi-scale transformation on the infrared image and the visible light image to obtain first low-frequency coefficient information, first high-frequency coefficient information, second low-frequency coefficient information and second high-frequency coefficient information;
the first determining module is used for determining third low-frequency coefficient information according to the first low-frequency coefficient information and the second low-frequency coefficient information based on a global contrast sequence maximum principle;
the second determining module is used for determining third high-frequency coefficient information according to the first high-frequency coefficient information and the second high-frequency coefficient information;
and the generating module is used for generating a fusion image according to the third low-frequency coefficient information and the third high-frequency coefficient information.
9. An infrared and visible image fusion system comprising a memory, a processor and a computer program stored in said memory and executable on said processor, characterized in that said processor implements an infrared and visible image fusion method according to any one of claims 1 to 7 when executing said computer program.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the infrared and visible image fusion method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110089292.0A CN112651469A (en) | 2021-01-22 | 2021-01-22 | Infrared and visible light image fusion method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110089292.0A CN112651469A (en) | 2021-01-22 | 2021-01-22 | Infrared and visible light image fusion method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112651469A true CN112651469A (en) | 2021-04-13 |
Family
ID=75370659
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110089292.0A Pending CN112651469A (en) | 2021-01-22 | 2021-01-22 | Infrared and visible light image fusion method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112651469A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113470055A (en) * | 2021-07-16 | 2021-10-01 | 南京信息工程大学 | Image fusion processing method based on FPGA acceleration |
CN113628151A (en) * | 2021-08-06 | 2021-11-09 | 苏州东方克洛托光电技术有限公司 | Infrared and visible light image fusion method |
CN116503454A (en) * | 2023-06-27 | 2023-07-28 | 季华实验室 | Infrared and visible light image fusion method and device, electronic equipment and storage medium |
CN116580062A (en) * | 2023-07-12 | 2023-08-11 | 南京诺源医疗器械有限公司 | Data processing method of infrared laser diagnostic device suitable for infrared excitation light source |
CN117218048A (en) * | 2023-11-07 | 2023-12-12 | 天津市测绘院有限公司 | Infrared and visible light image fusion method based on three-layer sparse smooth model |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109242888A (en) * | 2018-09-03 | 2019-01-18 | 中国科学院光电技术研究所 | Infrared and visible light image fusion method combining image significance and non-subsampled contourlet transformation |
CN109308691A (en) * | 2017-07-28 | 2019-02-05 | 南京理工大学 | Infrared and visible light image fusion method based on image enhancement and NSCT |
US20190318463A1 (en) * | 2016-12-27 | 2019-10-17 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for fusing infrared image and visible light image |
CN111899209A (en) * | 2020-08-11 | 2020-11-06 | 四川警察学院 | Visible light infrared image fusion method based on convolution matching pursuit dictionary learning |
-
2021
- 2021-01-22 CN CN202110089292.0A patent/CN112651469A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190318463A1 (en) * | 2016-12-27 | 2019-10-17 | Zhejiang Dahua Technology Co., Ltd. | Systems and methods for fusing infrared image and visible light image |
CN109308691A (en) * | 2017-07-28 | 2019-02-05 | 南京理工大学 | Infrared and visible light image fusion method based on image enhancement and NSCT |
CN109242888A (en) * | 2018-09-03 | 2019-01-18 | 中国科学院光电技术研究所 | Infrared and visible light image fusion method combining image significance and non-subsampled contourlet transformation |
CN111899209A (en) * | 2020-08-11 | 2020-11-06 | 四川警察学院 | Visible light infrared image fusion method based on convolution matching pursuit dictionary learning |
Non-Patent Citations (2)
Title |
---|
沈瑜;党建武;王阳萍;: "基于NSCT和Bilateral滤波器的含噪声图像融合", 兰州交通大学学报, no. 04, 15 August 2017 (2017-08-15) * |
罗萍;黄良学;黎羡飞;马强;: "基于NSCT的自适应多判决图像融合方法", 激光与红外, no. 03, 20 March 2016 (2016-03-20) * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113470055A (en) * | 2021-07-16 | 2021-10-01 | 南京信息工程大学 | Image fusion processing method based on FPGA acceleration |
CN113628151A (en) * | 2021-08-06 | 2021-11-09 | 苏州东方克洛托光电技术有限公司 | Infrared and visible light image fusion method |
CN113628151B (en) * | 2021-08-06 | 2024-04-26 | 苏州东方克洛托光电技术有限公司 | Infrared and visible light image fusion method |
CN116503454A (en) * | 2023-06-27 | 2023-07-28 | 季华实验室 | Infrared and visible light image fusion method and device, electronic equipment and storage medium |
CN116503454B (en) * | 2023-06-27 | 2023-10-20 | 季华实验室 | Infrared and visible light image fusion method and device, electronic equipment and storage medium |
CN116580062A (en) * | 2023-07-12 | 2023-08-11 | 南京诺源医疗器械有限公司 | Data processing method of infrared laser diagnostic device suitable for infrared excitation light source |
CN116580062B (en) * | 2023-07-12 | 2024-04-12 | 南京诺源医疗器械有限公司 | Data processing method of infrared laser diagnostic device suitable for infrared excitation light source |
CN117218048A (en) * | 2023-11-07 | 2023-12-12 | 天津市测绘院有限公司 | Infrared and visible light image fusion method based on three-layer sparse smooth model |
CN117218048B (en) * | 2023-11-07 | 2024-03-08 | 天津市测绘院有限公司 | Infrared and visible light image fusion method based on three-layer sparse smooth model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112651469A (en) | Infrared and visible light image fusion method and system | |
CN105096280B (en) | Handle the method and device of picture noise | |
CN111402146B (en) | Image processing method and image processing apparatus | |
Ancuti et al. | Single-scale fusion: An effective approach to merging images | |
CN108399611B (en) | Multi-focus image fusion method based on gradient regularization | |
CN109993707B (en) | Image denoising method and device | |
CN114041161A (en) | Method and device for training neural network model for enhancing image details | |
CN103020933B (en) | A kind of multisource image anastomosing method based on bionic visual mechanism | |
CN111951195A (en) | Image enhancement method and device | |
CN111914997A (en) | Method for training neural network, image processing method and device | |
CN114120176A (en) | Behavior analysis method for fusion of far infrared and visible light video images | |
CN109658354A (en) | A kind of image enchancing method and system | |
CN113362338B (en) | Rail segmentation method, device, computer equipment and rail segmentation processing system | |
CN115311186B (en) | Cross-scale attention confrontation fusion method and terminal for infrared and visible light images | |
CN112446835A (en) | Image recovery method, image recovery network training method, device and storage medium | |
CN114187214A (en) | Infrared and visible light image fusion system and method | |
Luo et al. | Deep wavelet network with domain adaptation for single image demoireing | |
CN115131256A (en) | Image processing model, and training method and device of image processing model | |
CN116757986A (en) | Infrared and visible light image fusion method and device | |
CN109064402A (en) | Based on the single image super resolution ratio reconstruction method for enhancing non local total variation model priori | |
CN116847209A (en) | Log-Gabor and wavelet-based light field full-focusing image generation method and system | |
Guo et al. | Multi-scale multi-attention network for moiré document image binarization | |
CN111353982B (en) | Depth camera image sequence screening method and device | |
CN111462004B (en) | Image enhancement method and device, computer equipment and storage medium | |
CN109614976A (en) | A kind of heterologous image interfusion method based on Gabor characteristic |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |