CN111445430A - Bimodal infrared image fusion algorithm selection method based on difference characteristic amplitude interval fusion validity distribution - Google Patents
Bimodal infrared image fusion algorithm selection method based on difference characteristic amplitude interval fusion validity distribution Download PDFInfo
- Publication number
- CN111445430A CN111445430A CN202010241703.9A CN202010241703A CN111445430A CN 111445430 A CN111445430 A CN 111445430A CN 202010241703 A CN202010241703 A CN 202010241703A CN 111445430 A CN111445430 A CN 111445430A
- Authority
- CN
- China
- Prior art keywords
- fusion
- amplitude
- interval
- image
- difference characteristic
- 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
- 230000004927 fusion Effects 0.000 title claims abstract description 298
- 238000009826 distribution Methods 0.000 title claims abstract description 98
- 230000002902 bimodal effect Effects 0.000 title claims abstract description 21
- 238000010187 selection method Methods 0.000 title claims abstract description 13
- 230000010287 polarization Effects 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims description 16
- 230000000903 blocking effect Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 238000000034 method Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 43
- 230000015572 biosynthetic process Effects 0.000 abstract description 5
- 238000004451 qualitative analysis Methods 0.000 abstract description 4
- 238000003786 synthesis reaction Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 8
- 238000011160 research Methods 0.000 description 6
- 238000012795 verification Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000007500 overflow downdraw method Methods 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000012418 validation experiment Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
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 by the use of more than one image, e.g. averaging, subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted 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
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Abstract
The invention relates to the fusion of bimodal infrared images, in particular to a bimodal infrared image fusion algorithm selection method based on fusion effectiveness distribution of difference characteristic amplitude intervals, which comprises the following steps of (1) extracting the amplitude of a difference characteristic by selecting the main difference characteristic of an infrared polarization image and a light intensity image; (2) determining the fusion effectiveness of the algorithm in different amplitude intervals by dividing the intervals of the difference characteristic amplitudes, and constructing the fusion effectiveness distribution of the difference characteristic amplitudes of the fusion algorithm; (3) according to the difference characteristic amplitude fusion effectiveness distribution of the constructed fusion algorithm, the algorithm with the relatively best fusion effect can be optimally selected from a plurality of fusion algorithms, the limitation of selecting the fusion algorithm according to the relation between the general subjective qualitative analysis algorithm and the image difference characteristic is overcome, a new objective basis is provided for the optimal selection of the fusion algorithm, and a theoretical basis is provided for researching the multi-characteristic amplitude fusion effectiveness distribution synthesis.
Description
Technical Field
The invention relates to bimodal infrared image fusion, in particular to a bimodal infrared image fusion algorithm selection method based on fusion effectiveness distribution of difference characteristic amplitude intervals.
Background
The infrared polarization and the light intensity imaging respectively obtain different attributes of the target by detecting infrared polarization and intensity information, so that the two types of images have strong complementarity, the two types of images are fused, different imaging advantages can be integrated, and the practical application requirements of target detection, safety detection and the like can be met. In order to obtain a fused image with higher quality, how to reasonably select a fusion algorithm gradually becomes one of the hot spots of the current image fusion research.
At present, as for the selection of a fusion algorithm, scholars mostly realize image fusion by qualitatively analyzing a fusion effect selection algorithm of the fusion algorithm on image difference characteristics, for example, a Top-Hat transform (Top-Hat) algorithm has the advantage of keeping the brightness difference characteristics among images, and the Top-Hat transform (Top-Hat) algorithm is combined with a Support Value Transform (SVT) fusion algorithm to improve the contrast of a fusion image; robust Principal Component Analysis (RPCA) can effectively represent sparse characteristics of an image, introduce the sparse characteristics into a non-subsampled contourlet transform (NSCT) domain for fusion, highlight a target in a source image and keep background information of the target; the multi-scale edge preserving decomposition can acquire image edge characteristics under multi-scale by using a nonlinear edge preserving filter and avoid the occurrence of edge artifacts, and can improve the overall visual effect of a fused image by combining the edge characteristics with Guide Filtering (GF). However, in an actual detection environment, it is impossible to determine the difference features of the image in advance, and the types, amplitudes, and other attributes of the difference features change with the change of the detection environment, so that the preselected fusion algorithm may not always maintain a good fusion performance, and a situation of poor or unstable fusion effect may be unavoidable. The reason for this is that the relationship between the fusion algorithm and the fusion effectiveness degree of the changed difference features is not studied in depth, so that the static fusion algorithm cannot be dynamically changed and adjusted according to the dynamic difference features, and thus the fusion algorithm effective for the fusion of the difference features cannot be selected.
Currently, researchers have developed researches on the relationship between the fusion algorithm and the image difference features, for example, the flexible fusion of multiple types of difference information theoretically analyzes the correspondence between the difference information and the fusion method. The applicant analyzes the performance of different fusion algorithms through earlier research on the formation mechanism of the difference characteristics and the characteristic-driven fusion method, and researches the corresponding relation between the type of the difference characteristics and the fusion algorithms. Research finds that the amplitude of the difference characteristic has great influence on the selection of the fusion algorithm besides the attribute of type, and the related documents do not research the amplitude deeply. Therefore, it is necessary to search the relationship between the fusion algorithm and the fusion validity degree of the varied difference feature amplitudes and establish the fusion validity distribution of the difference feature amplitudes of the fusion algorithm.
Disclosure of Invention
The invention provides a bimodal infrared image fusion algorithm selection method based on fusion effectiveness distribution of difference characteristic amplitude intervals, and aims to establish a corresponding relation between a fusion algorithm and difference characteristic amplitudes by constructing the fusion effectiveness distribution of the difference characteristic amplitudes of the fusion algorithm and realize the optimal selection of the fusion algorithm.
The invention is realized by adopting the following technical scheme: the bimodal infrared image fusion algorithm selection method based on the fusion validity distribution of the difference characteristic amplitude interval comprises the following steps:
(1) extracting the difference characteristic amplitude: simultaneously performing block processing on the infrared polarization and infrared light intensity images to obtain m image blocks
And calculating the image block P corresponding to each i positioni、IiThree kinds of characteristic values of
T ═ M, EI, SF }; then according to formula
Calculating Pi、IiIs a difference characteristic value
Finally, counting the difference characteristic values of all image blocks
Obtaining the amplitude A of the difference characteristic of the whole imageT；
(2) Calculating the proximity of the source image and the fused image difference characteristic value: fusing the infrared polarization and infrared light intensity images by adopting a fusion algorithm, and obtaining a fused image block by adopting a blocking processing mode on the fused image F
By the formula
The closeness of the fused image to the infrared polarization and infrared light intensity image on three special types is calculated,
representing each i-position corresponding image block Pi、IiAnd the fused image block FiAt a particular difference feature amplitude
A proximity value of where
Image block F for representing fused image at position iiA difference feature value of (c); proximity S1Closer to 0 indicates better fusion of the fusion algorithm to the difference feature, however, proximity S1The specific range in which this occurs may indicate that the higher fusion effect is still unknown, and therefore, the determination of the effective fusion threshold C is required1I.e. when in proximity S1Less than C1It can be shown that there is a higher fusion effect, thereby obtaining an effective fusion interval [0, C ] within the proximity1]；
(3) By the difference characteristic amplitude ATMaximum value as standard, for widthDividing the values into equally spaced intervals, and equally dividing the amplitude into n intervals
(4) Discrete probability distribution construction by first arbitrarily taking L different values of C close to 0kK is 1,2, … L is greater than or equal to 2, and the interval R is calculatedjThe inner proximity value is in the interval [0, Ck]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CkDiscrete probability distribution with amplitude interval R on the horizontal axisjThe vertical axis is the probability ratio Pj；
(5) Selecting an effective fusion threshold value: from different values of CkSelecting numerical values with the same distribution trend from the lower discrete probability distribution;
(6) determination of effective fusion interval: carrying out weighted average on the selected numerical values with the same distribution trend to obtain C1Thereby obtaining an effective fusion interval [0, C1]；
(7) Calculating fusion validity of the difference characteristic amplitude interval: by using
Is calculated at RiWithin, proximity is located at [0, C1]Discrete number of points and RjInner total scatter QjRatio P ofjAnd with PjAs fusion validity of fusion algorithm in different difference characteristic amplitude intervals;
(8) and selecting a fusion algorithm with the fusion validity values in different amplitude intervals kept highest for a plurality of times to fuse the bimodal infrared images according to the fusion validity of different fusion algorithms in the three different characteristic amplitude intervals.
The bimodal infrared image fusion algorithm selection method based on the fusion validity distribution of the difference characteristic amplitude interval comprises the following steps:
(1) extracting the difference characteristic amplitude: simultaneously performing block processing on the infrared polarization and infrared light intensity images to obtain m image blocks
And calculating the image block P corresponding to each i positioni、IiThree kinds of characteristic values of
T ═ M, EI, SF }; then according to formula
Calculating Pi、IiIs a difference characteristic value
Finally, counting the difference characteristic values of all image blocks
Obtaining the amplitude A of the difference characteristic of the whole imageT；
(2) Calculating the proximity of the source image and the fused image difference characteristic value: fusing the infrared polarization and infrared light intensity images by adopting a fusion algorithm, and obtaining a fused image block by adopting a blocking processing mode on the fused image F
By the formula
The closeness of the fused image to the infrared polarization and infrared light intensity image on three special types is calculated,
representing each i-position corresponding image block Pi、IiAnd the fused image block FiAt a particular difference feature amplitude
Proximity value ofIn
Image block F for representing fused image at position iiA difference feature value of (c); proximity S2Closer to 1 indicates that the fusion algorithm has better fusion effect on the difference feature, however, the proximity S2The specific range in which this occurs may indicate that the higher fusion effect is still unknown, and therefore, the determination of the effective fusion threshold C is required2I.e. when in proximity S2Greater than C2Can indicate a higher fusion effect, thereby obtaining an effective fusion interval [ C ] within the proximity2,1]；
(3) By the difference characteristic amplitude ATDividing the amplitude into n intervals equally by taking the maximum value as a standard
(4) Discrete probability distribution construction: first, arbitrarily take S different values C close to 1tT is 1,2, … S, S is 2 or more; calculating the interval RjThe inner proximity value is in the interval [ Ct,1]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CtDiscrete probability distribution with amplitude interval R on the horizontal axisjThe vertical axis is the probability ratio Pj；
(5) Selecting an effective fusion threshold value: from different values of CtSelecting numerical values with the same distribution trend from the lower discrete probability distribution;
(6) determination of effective fusion interval: carrying out weighted average on the selected numerical values with the same distribution trend to obtain C2Thereby obtaining an effective fusion interval [ C2,1]；
(7) Calculating fusion validity of the difference characteristic amplitude interval: by using
Is calculated at RiWithin, proximity is located at [ C2,1]Discrete number of points and RjInner total scatter QjRatio P ofjAnd with PjAs fusion validity of fusion algorithm in different difference characteristic amplitude intervals;
(8) and selecting a fusion algorithm with the fusion validity values in different amplitude intervals kept highest for a plurality of times to fuse the bimodal infrared images according to the fusion validity of different fusion algorithms in the three different characteristic amplitude intervals.
The bimodal infrared image fusion algorithm selection method based on the fusion validity distribution of the difference characteristic amplitude interval comprises the following steps:
(1) extracting the difference characteristic amplitude: simultaneously performing block processing on the infrared polarization and infrared light intensity images to obtain m image blocks
And calculating the image block P corresponding to each i positioni、IiThree kinds of characteristic values of
T ═ M, EI, SF }; then according to formula
Calculating Pi、IiIs a difference characteristic value
Finally, counting the difference characteristic values of all image blocks
Obtaining the amplitude A of the difference characteristic of the whole imageT；
(2) Calculating the proximity of the source image and the fused image difference characteristic value: fusing the infrared polarization and infrared light intensity images by adopting a fusion algorithm, and obtaining a fused image block by adopting a blocking processing mode on the fused image F
By the formula
And
the closeness of the fused image to the infrared polarization and infrared light intensity image in three special modes is respectively calculated,
and
representing each i-position corresponding image block Pi、IiAnd the fused image block FiAt a particular difference feature amplitude
A proximity value of where
Image block F for representing fused image at position iiA difference feature value of (c); proximity S1Closer to 0 indicates better fusion of the fusion algorithm to the difference feature, however, proximity S1The specific range in which this occurs may indicate that the higher fusion effect is still unknown, and therefore, the determination of the effective fusion threshold C is required1I.e. when in proximity S1A lower than C1 may indicate a higher fusion effect, thereby resulting in an effective fusion interval [0, C ] within proximity1](ii) a Proximity S2Closer to 1 indicates that the fusion algorithm has better fusion effect on the difference feature, however, the proximity S2The specific range in which this occurs may indicate that the higher fusion effect is still unknown, and therefore, the determination of the effective fusion threshold C is required2I.e. when in proximity S2Greater than C2Can indicate a higher fusion effect, thereby obtaining an effective fusion interval [ C ] within the proximity2,1]；
(3) By difference characteristic breadthValue ATDividing the amplitude into n intervals equally by taking the maximum value as a standard
(4) Discrete probability distribution construction for closeness S1, first take L different values C close to 0kK is 1,2, … L is greater than or equal to 2, and the interval R is calculatedjThe inner proximity value is in the interval [0, Ck]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CkA lower discrete probability distribution; for the proximity S2, first, S different values C close to 1 are arbitrarily takentT is 1,2, … S, S is 2 or more; calculating the interval RjThe inner proximity value is in the interval [ Ct,1]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CtDiscrete probability distribution with amplitude interval R on the horizontal axisjThe vertical axis is the probability ratio Pj；
(5) Selecting an effective fusion threshold value: from different values of CkAnd CtSelecting numerical values with the same distribution trend from the lower discrete probability distribution;
(6) determination of effective fusion interval: c obtained by carrying out weighted average on the selected numerical values with the same distribution trend1And C2Thereby obtaining an effective fusion interval [0, C1]And [ C2,1]；
(7) Calculating fusion validity of the difference characteristic amplitude interval: by using
Is calculated at RiWithin, proximity is located at [0, C1]And [ C2,1]Discrete number of points and RjInner total scatter QjRatio P ofjAnd with PjAs fusion validity of the fusion algorithm in different difference characteristic amplitude intervals, performing weighted average processing on the fusion validity under two proximity degrees to obtain final fusion validity;
(8) and selecting a fusion algorithm which keeps the highest fusion validity value in different amplitude intervals for a plurality of times to fuse the bimodal infrared images according to the final fusion validity of different fusion algorithms in the three different characteristic amplitude intervals.
The selection method of the bimodal infrared image fusion algorithm based on the fusion validity distribution of the difference characteristic amplitude interval is L different values C which are close to 0kAnd selecting from small to large in sequence.
The selection method of the bimodal infrared image fusion algorithm based on the fusion validity distribution of the difference characteristic amplitude interval has S different values C close to 1tAnd selecting from small to large in sequence.
The invention can optimize and select the algorithm with the relatively best fusion effect from a plurality of fusion algorithms, overcomes the limitation of selecting the fusion algorithm according to the relation between the common subjective qualitative analysis algorithm and the image difference characteristic, provides a new objective basis for the optimization and selection of the fusion algorithm, and provides a theoretical basis for researching the distribution and synthesis of the fusion effectiveness of the multi-characteristic amplitude.
Drawings
Fig. 1 is an experimental image of 5 sets of infrared polarization and light intensity.
FIG. 2 is a graph of the difference feature amplitude distribution of source images P1 and I1.
FIG. 3 is a discrete point diagram of spatial frequency feature value proximity between a source image and a fused image.
FIG. 4 shows a variant C1(or C)2) The proximity value of the lower discrete point is [0, C ]1](or [ C)2,1]) The discrete probability distribution diagram of the number of the scattered points and the total number of the scattered points in the amplitude interval.
Fig. 5 is a spatial frequency amplitude range fusion validity distribution diagram of the GFF algorithm.
Fig. 6 is a difference feature amplitude region fusion validity distribution diagram of different algorithms and a difference feature amplitude fusion validity distribution diagram of a synthesized fusion algorithm.
Fig. 7 is a first experimental graph for validity verification.
Fig. 8 is a second experimental graph for validity verification.
Detailed Description
1 fusion algorithm difference characteristic amplitude range fusion validity
1.1 extraction of differential feature amplitude
The difference characteristics T selected in the present application includes Mean gray scale values (MeFreen, M), edge intensity (EdgeDenity, Intensity) and spatial frequency (SpAlice, SF) since the Mean gray scale values of the images reflect the brightness information of the images, the edge intensity reflects the edge information of the images and the spatial frequency reflects the detail information of the images, and the difference characteristics T selected in the present application includes the Mean gray scale values (MeFreen, M), the edge intensity (EdgeDenity, Intensity) and the spatial frequency (SpAlice, SF) since the infrared polarization and the intensity images are different from each other as seen from the 5 sets of strictly registered experimental images (size: 256 × 256) of the infrared polarization and the intensity of FIG. 1.
In order to obtain the amplitude of the difference characteristic of the infrared polarization and the light intensity image, firstly, a blocking processing mode with the same size and no overlap is adopted for the infrared polarization and the infrared light intensity image at the same time to obtain m image blocks
(m is 256 in the present invention) and calculates the image block P corresponding to each i positioni、IiThree characteristics ofValue of
(T ═ { M, EI, SF }); then P is calculated according to equation 1i、IiIs a difference characteristic value
Finally, counting the difference characteristic values of all image blocks
Obtaining the amplitude A of the difference characteristic of the whole imageTAs fig. 2 shows the difference feature amplitude distribution of the source images P1 and I1, it can be seen that the variation of the difference feature amplitude is random and uncertain.
1.2 proximity of Source images to fused image Difference feature values
For infrared polarization and light intensity images, the fused image with better fusion effect is represented as follows: (1) the brightness characteristic value of the fused image is very close to the brightness characteristic value of the infrared light intensity image; (2) the edge and texture feature values of the fused image are also as close as possible to those of the infrared polarization image. In order to convert the qualitative analysis into the quantitative analysis of the fusion effect of the fusion image, firstly, the method of 1.1 blocks is adopted for the fusion image F to obtain a fusion image block
Then on each corresponding image block, F for quantitative analysisiAnd IiProximity and F on the mean of gray featureiAnd PiThe closeness degree of the edge intensity feature and the spatial frequency feature needs to correspondingly calculate F under the three difference featuresiCharacteristic value of and source image Pi、IiThe closeness between the largest eigenvalues. In particular, the present application takes two different formulas to calculate proximity
As shown in formula (2) and formula (3), wherein
Representing the fused image in the image block FiThe difference characteristic value of (a).
1.3 fusion significance of differential feature amplitude interval
As can be seen from the analysis in the combination of formula (2) and formula (3) in combination with 1.2, the proximity S1The closer to 0 (or S)2Closer to 1), indicating that the fusion algorithm has better fusion effect on the difference feature. However, the proximity S1(or S)2) The specific range in which this occurs may indicate that the higher fusion effect is still unknown, and therefore, the determination of the effective fusion threshold C is required1(or C)2) I.e. when in proximity S1Less than C1(or S)2Greater than C2) It can be shown that there is a higher fusion effect, thereby obtaining an effective fusion interval [0, C ] within the proximity1](or [ C)2,1]). Considering that the overall range of the difference characteristic amplitude values of different source images is different, in order to be able to synthesize the difference characteristic amplitude value distribution of different source images, the difference characteristic amplitude values of different source images need to be divided into the same interval. Thus, within different amplitude intervals, the closeness S is calculated1(or S)2) In the effective fusion interval [0, C1](or [ C)2,1]) The number of image blocks of (a) is a proportion of the total number of blocks, and thus the difference feature amplitude interval of the present application is fused with the significance.
The specific steps are as follows, wherein a GFF algorithm with edge preserving property is used as a fusion algorithm, a spatial frequency is used as a difference feature, and the first group of images P1 and I1 in fig. 1 are used as source images (fig. 3 to 6 all use the first group of images P1 and I1 in fig. 1 as source images):
step 1: and establishing a discrete point diagram of the proximity of the difference characteristic values between the source image and the fused image. Firstly, the difference characteristic amplitude is obtained by the step of 1.1
The difference feature proximity is then calculated from 1.2
Finally, obtaining a discrete point diagram of the proximity of the spatial frequency characteristic value between the source image and the fused image as shown in FIG. 3. Wherein, each scatter point in the figure represents each i-position corresponding block Pi、IiAnd fusion image FiAt a particular difference feature amplitude
Lower proximity value
(or
)。
Step 2: and dividing equal interval intervals of the difference characteristic amplitude. Dividing the amplitude (horizontal axis of the discrete point diagram in FIG. 3) into equal interval intervals by taking the maximum value of the difference characteristic amplitude as a standard, and equally dividing the amplitude into n intervals
(in this application n-8).
Step 3: and constructing discrete probability distribution. First, respectively to C1(or C)2) Any number very close to 0 (or 1), e.g. 0.01, 0.02, 0.04, 0.06, 0.08, 0.1 (or 0.9, 0.92, 0.94, 0.96, 0.98, 0.99); however, the device is not suitable for use in a kitchenThen, the interval R is calculated according to FIG. 3 and using equation 4jInner discrete point when its proximity value is in the interval [0,0.01 ]]、[0,0.02]、[0,0.04]、[0,0.06]、[0,0.08]And [0,0.1](or [0.9,1 ]]、[0.92,1]、[0.94,1]、[0.96,1]、[0.98,1]And [0.99,1]) Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofjConstructed as C shown in FIG. 41(or C)2) Taking discrete probability distribution under different values, wherein the horizontal axis is an amplitude interval RjThe vertical axis is the probability ratio Pj。
Step 4: and selecting an effective fusion threshold value. For the amplitude interval RjTo avoid C1(or C)2) The variation of the value generates larger fluctuation to the value, so the distribution trends in the graph 4(a) and (b) are the same (namely in the amplitude interval R) from the different values arbitrarily selected in Step3jThe ascending or descending trend of the inner distribution is the same), such as selecting the effective fusion threshold C in FIG. 4(a)1Is 0.06, 0.08, 0.1, and the effective fusion threshold C is selected from FIG. 4(b)2Are 0.98 and 0.99.
Step 5: and determining the effective fusion interval. For a plurality of effective fusion threshold values C selected in Step41(or C)2) C obtained by weighted average1(or C)2) Value, thereby obtaining the final effective fusion interval [0, C1](or [ C)2,1]). Therefore, for spatial frequency features, the final effective fusion interval is [0,0.08 ], respectively]、[0.985,1]。
Step 6: and calculating the fusion validity of the difference characteristic amplitude interval. The final effective fusion interval [0, C ] is obtained from Step51](or [ C)2,1]) Calculated at R using equation 4iWithin, proximity is located at [0, C1](or [ C)2,1]) Discrete number of points and RjInner total scatter QjRatio P ofjAnd with PjAnd the fusion effectiveness of the fusion algorithm in different difference characteristic amplitude intervals is obtained. Table 1 shows the difference characteristics as spaceIn frequency, in two valid fusion regions [0,0.08 ]]And [0.985,1 ]]Next, the GFF algorithm operates at different amplitude intervals RjThe fusion efficacy of the following.
TABLE 1 fusion effectiveness of GFF algorithm in different spatial frequency amplitude intervals
Table 1 Fusion validity in different amplitude intervals of spatialfrequency feature for GFF algorithm
2 fusion algorithm difference feature amplitude fusion significance distribution
Based on the fusion validity obtained in 1.3, a fusion validity in a valid fusion interval [0, C ] can be constructed1](or [ C)2,1]) The two GFF algorithm fusion effectiveness distributions of (1) shown in fig. 5(a) are represented by the solid line and the broken line respectively, where the bold black line Average represents the distribution after weighted Average processing of the two fusion effectiveness distributions, and the two fusion effectiveness distributions are obtained from the data of table 1. In order to comprehensively consider the information of the two distributions, the application takes the distribution Average as the fusion effectiveness distribution of the GFF algorithm at the spatial frequency.
For other 4 groups of images in fig. 1, a fusion validity distribution similar to the GFF algorithm in fig. 5(a) may also be obtained, and in order to reflect the fusion effect of the fusion algorithm in different source images, the fusion validity distribution in the 5 groups of source images is weighted-averaged, so that the fusion validity distribution in the spatial frequency amplitude interval of the GFF algorithm after weighted averaging in the 5 groups of source images shown in fig. 5(b) can be obtained.
In addition to the GFF algorithm, the method selects (1) Laplace pyramid transformation fusion (L P) with better fusion effect on image edges and texture details, (2) non-downsampling contourlet transformation fusion (NSST) with multi-scale, multi-resolution and multi-direction characteristics, and (3) Gradient Transfer Fusion (GTF) with better brightness and detail information, so that the spatial frequency amplitude fusion effectiveness distribution of L P, NSST, GFF and GTF algorithms under the average of five groups of images can be obtained as shown in fig. 6(a), and the amplitude fusion effectiveness distribution of the fusion algorithm can be made for edge intensity and gray level mean values as shown in fig. 6(b) and (c). As can be seen from fig. 6, when the amplitude interval of the difference features changes, the fusion effectiveness value of the fusion algorithm fluctuates, which shows that the change of the amplitude of the difference features can cause important influence on the fusion effect of the fusion algorithm on the difference features.
In order to optimize and select the fusion algorithm along with the variation of the difference characteristic amplitude value and realize the purpose of selecting the relatively optimal fusion algorithm from the multiple algorithms, the fusion validity distribution of the difference characteristic amplitude value intervals of the multiple algorithms constructed above needs to be further processed: calculating the fusion effectiveness of the 4 algorithms in each difference characteristic amplitude interval, selecting the algorithm with the highest fusion effectiveness value as the algorithm with the relatively best fusion effect in the amplitude interval, i.e. taking the large synthesis of the fusion effectiveness distribution of the 4 algorithms, and finally obtaining the difference characteristic amplitude fusion effectiveness distribution of the fusion algorithm, as shown by the bold black line M in fig. 6, which reflects the corresponding relationship between the optimal fusion algorithm and the difference characteristic amplitude interval.
Particularly, the more times of the maximum fusion validity value of a certain algorithm occurs in different amplitude intervals, the more the algorithm shows the better and more stable overall fusion effect on the difference feature, (1) according to fig. 6(a), the GFF algorithm with the best fusion effect when the spatial frequency amplitude is in the 1 st interval, and the L P algorithm with the best fusion effect in the 2 nd to 8 th intervals, so that the fusion effect of the L P algorithm is better for the edge intensity feature, (2) according to fig. 6(b), the GFF algorithm with the best fusion effect when the edge intensity amplitude is in the 1 st, 5, 7 th intervals, and the L P algorithm with the best fusion effect in the 2 nd, 3, 4, 6, 7, and 8 th intervals, so that for the spatial frequency feature, the L P, GFF algorithm has the better fusion effect, (3) according to fig. 6(c), the NSST algorithm with the best fusion effect when the gray mean amplitude is in the 1 st interval, the GTF algorithm with the best fusion effect in the 3 rd, 4, 5 th interval, the GTF algorithm is the GTF algorithm with the best effect when the gray mean amplitude is in the 1 st interval, the GTF algorithm, and the GTF algorithm is the best in the GTF 6, and the GFF 8, so that the best fusion effect is the gray mean algorithm is the best.
3 verification experiment
3.1 validation experiment of the Experimental images
Since the more times an algorithm appears in the fusion validity distribution M of fig. 6, that is, the more times the algorithm maintains the highest fusion validity value in different amplitude intervals, the better the fusion effect of the fusion algorithm on the difference features can be reflected. Therefore, based on the difference feature amplitude fusion validity distribution M of the fusion algorithm in section 2, it is possible to select, from among a plurality of algorithms, whether the algorithm having the key of the difference feature fusion effect optimization algorithm that is more frequent in the occurrence of the fusion validity distribution can be consistent with the algorithm having the highest average fusion validity.
Considering that fig. 6 is the fusion validity distribution under the average of the five groups of images in fig. 1, it can be seen from the average value of 5 groups of images (Group 1-5 indicates 5 groups of images in fig. 1) of three difference features in different amplitude intervals, that the proportion of the number of scatter points in 7 th and 8 th intervals of the difference feature amplitude to the total scatter points is the lowest, and the influence degree of the two amplitude intervals on the selection of the algorithm for fusing the difference features relatively best is lower than that of the other intervals, so for fig. 6, only the case where the difference feature amplitude is in the 1 st to 6 th intervals is considered herein.
TABLE 2 ratio of scatter point number to total scatter point number in different gray level mean amplitude intervals
Table2The ratios of scattered numbers of different amplitudeintervals on the mean to total scatter numbers
TABLE 3 ratio of scatter point number to total scatter point number in different edge intensity amplitude intervals
Table 3 The ratios of scattered numbers of different amplitudeintervals on edge intensity to total scatter numbers
TABLE 4 ratio of the number of scattered points to the total number of scattered points in different spatial frequency amplitude intervals
Table 4 The ratios of scattered numbers of different amplitudeintervals on spatial frequency to total scatter numbers
According to fig. 6, the average fusion validity values of 4 algorithms in 6 amplitude intervals under different difference characteristics can be calculated, as shown in table 5. the detailed analysis is as follows, (1) from fig. 6(a) M, L P algorithm appears 5 times in 1 st to 6 th amplitude intervals, and the occurrence frequency is the largest, from table 5, it can be seen that L P algorithm has the highest average fusion validity value for spatial frequency characteristics, so that L P algorithm with the largest occurrence frequency of fusion validity distribution M in fig. 6(a) is consistent with L P algorithm with the highest average fusion validity value, (2) from fig. 6(b) M, L P algorithm appears 4 times in 1 st to 6 th amplitude intervals, and the largest occurrence frequency, from table 5, it can be seen that L P algorithm has the highest average fusion validity value for edge intensity characteristics, so that L P algorithm with the largest occurrence frequency of validity distribution M in fig. 6(b) is consistent with L P algorithm with the highest average fusion validity distribution, so that GTF algorithm with the largest occurrence frequency of fusion efficiency distribution M in fig. 6(b) is consistent with the largest average fusion validity distribution f distribution, so that GTF algorithm is selected from the largest average fusion validity distribution, and the most effective distribution of GTF algorithm (GTF) is shown by the most effective distribution.
TABLE 5 average fusion validity values of 4 algorithms in 6 amplitude intervals under different difference characteristics
Table 5 The average fusion validity of six amplitude intervals ondifference features for 4 kinds of fusion algorithms
3.2 verification experiment of test images
In order to verify the validity of the fusion validity distribution constructed in section 2, in the application, MAT L AB 2014b is used as an experimental software platform, two groups of infrared polarization and light intensity test images which are different in size from the 5 groups of images in fig. 1 and subjected to strict registration are selected as source images (marked as Group 6 and 7) in different scenes, as shown in (a) and (b) in fig. 7 and 8, and the following validity verification experiments are performed:
firstly, three more remarkable infrared polarization and light intensity image difference characteristics of spatial frequency, edge intensity and gray level mean value are selected, then, according to the fusion effectiveness distribution of the fusion algorithm and the difference characteristic amplitude value in the figure 6, the GTF algorithm effect is better for the gray level mean value, the L P algorithm effect is better for the edge intensity and the spatial frequency, and finally, the two groups of images are respectively fused according to L P and GTF algorithms, as shown in (c) and (f) of figures 7 and 8.
TABLE 6 difference eigenvalues of different algorithms under Group 6
Table 6 Difference feature values of different algorithms on Group 6
TABLE 7 difference eigenvalues of different algorithms under Group 7
Table 7Difference feature values of different algorithms on Group 7
It can be seen from the difference feature values of the 4 algorithm fusion images shown in tables 6 and 7 that, under two new sets of infrared polarization and light intensity source images, the gray level mean value of the GTF algorithm is the highest, and the edge intensity and spatial frequency value of the L P algorithm are also the highest, therefore, according to the fusion validity distribution M constructed in section 2 of the present application, the fusion algorithm that makes the difference feature value of the fusion image the highest (the best fusion effect on the difference feature) can be effectively selected.
4 conclusion
In order to establish the relationship between the fusion algorithm and the fusion effectiveness degree of the varied difference characteristic amplitudes, the difference characteristic amplitude fusion effectiveness degree distribution of the fusion algorithm is constructed (1) the amplitude of the difference characteristic is extracted by selecting the main difference characteristic of the infrared polarization and light intensity image, and the change of the amplitude is analyzed to influence the fusion effect of the fusion algorithm on the difference characteristic, so that the fluctuation of the fusion effect of the algorithm is caused; (2) the fusion effectiveness of the algorithm in different amplitude intervals is determined by dividing the intervals of the difference characteristic amplitudes, and the difference characteristic amplitude fusion effectiveness distribution of the fusion algorithm is constructed according to the fusion effectiveness, so that the corresponding relation between the algorithm with the best fusion effect and the difference characteristic amplitudes is reflected; (3) according to the difference characteristic amplitude fusion effectiveness distribution of the constructed fusion algorithm, the algorithm with the relatively best fusion effect can be optimally selected from a plurality of fusion algorithms, the limitation of selecting the fusion algorithm according to the relation between the general subjective qualitative analysis algorithm and the image difference characteristic is overcome, a new objective basis is provided for the optimal selection of the fusion algorithm, and a theoretical basis is provided for researching the multi-characteristic amplitude fusion effectiveness distribution synthesis.
Claims (5)
1. The bimodal infrared image fusion algorithm selection method based on the fusion validity distribution of the difference characteristic amplitude interval is characterized by comprising the following steps of:
(1) extracting the difference characteristic amplitude: simultaneously performing block processing on the infrared polarization and infrared light intensity images to obtain m image blocks
And calculating each i-position corresponding imageBlock Pi、IiThree kinds of characteristic values of
T ═ M, EI, SF }; then according to formula
Calculating Pi、IiIs a difference characteristic value
Finally, counting the difference characteristic values of all image blocks
Obtaining the amplitude A of the difference characteristic of the whole imageT；
(2) Calculating the proximity of the source image and the fused image difference characteristic value: fusing the infrared polarization and infrared light intensity images by adopting a fusion algorithm, and obtaining a fused image block by adopting a blocking processing mode on the fused image F
By the formula
The closeness of the fused image to the infrared polarization and infrared light intensity image on three special types is calculated,
representing each i-position corresponding image block Pi、IiAnd the fused image block FiAt a particular difference feature amplitude
A proximity value of where
Image block F for representing fused image at position iiA difference feature value of (c);
(3) by the difference characteristic amplitude ATDividing the amplitude into n intervals equally by taking the maximum value as a standard
(4) Discrete probability distribution construction by first arbitrarily taking L different values of C close to 0kK is 1,2, … L is greater than or equal to 2, and the interval R is calculatedjThe inner proximity value is in the interval [0, Ck]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CkDiscrete probability distribution with amplitude interval R on the horizontal axisjThe vertical axis is the probability ratio Pj；
(5) Selecting an effective fusion threshold value: from different values of CkSelecting numerical values with the same distribution trend from the lower discrete probability distribution;
(6) determination of effective fusion interval: carrying out weighted average on the selected numerical values with the same distribution trend to obtain C1Thereby obtaining an effective fusion interval [0, C1]；
(7) Calculating fusion validity of the difference characteristic amplitude interval: by using
Is calculated at RiWithin, proximity is located at [0, C1]Discrete number of points and RjInner total scatter QjRatio P ofjAnd with PjAs fusion validity of fusion algorithm in different difference characteristic amplitude intervals;
(8) and selecting a fusion algorithm with the fusion validity values in different amplitude intervals kept highest for a plurality of times to fuse the bimodal infrared images according to the fusion validity of different fusion algorithms in the three different characteristic amplitude intervals.
2. The bimodal infrared image fusion algorithm selection method based on the fusion validity distribution of the difference characteristic amplitude interval is characterized by comprising the following steps of:
(1) extracting the difference characteristic amplitude: simultaneously performing block processing on the infrared polarization and infrared light intensity images to obtain m image blocks
And calculating the image block P corresponding to each i positioni、IiThree kinds of characteristic values of
T ═ M, EI, SF }; then according to formula
Calculating Pi、IiIs a difference characteristic value
Finally, counting the difference characteristic values of all image blocks
Obtaining the amplitude A of the difference characteristic of the whole imageT；
(2) Calculating the proximity of the source image and the fused image difference characteristic value: fusing the infrared polarization and infrared light intensity images by adopting a fusion algorithm, and obtaining a fused image block by adopting a blocking processing mode on the fused image F
By the formula
The closeness of the fused image to the infrared polarization and infrared light intensity image on three special types is calculated,
representing each i-position corresponding image block Pi、IiAnd the fused image block FiAt a particular difference feature amplitude
A proximity value of where
Image block F for representing fused image at position iiA difference feature value of (c);
(3) by the difference characteristic amplitude ATDividing the amplitude into n intervals equally by taking the maximum value as a standard
(4) Discrete probability distribution construction: first, arbitrarily take S different values C close to 1tT is 1,2, … S, S is 2 or more; calculating the interval RjThe inner proximity value is in the interval [ Ct,1]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CtDiscrete probability distribution with amplitude interval R on the horizontal axisjThe vertical axis is the probability ratio Pj；
(5) Selecting an effective fusion threshold value: from different values of CtSelecting numerical values with the same distribution trend from the lower discrete probability distribution;
(6) determination of effective fusion interval: carrying out weighted average on the selected numerical values with the same distribution trend to obtain C2Thereby obtaining an effective fusion interval [ C2,1]；
(7) Calculating fusion validity of the difference characteristic amplitude interval: by using
Is calculated at RiWithin, proximity is located at [ C2,1]Discrete number of points and RjInner total scatter QjRatio P ofjAnd with PjAs fusion validity of fusion algorithm in different difference characteristic amplitude intervals;
(8) and selecting a fusion algorithm with the fusion validity values in different amplitude intervals kept highest for a plurality of times to fuse the bimodal infrared images according to the fusion validity of different fusion algorithms in the three different characteristic amplitude intervals.
3. The bimodal infrared image fusion algorithm selection method based on the fusion validity distribution of the difference characteristic amplitude interval is characterized by comprising the following steps of:
(1) extracting the difference characteristic amplitude: simultaneously performing block processing on the infrared polarization and infrared light intensity images to obtain m image blocks
And calculating the image block P corresponding to each i positioni、IiThree kinds of characteristic values of
T ═ M, EI, SF }; then according to formula
Calculating Pi、IiIs a difference characteristic value
Finally, counting the difference characteristic values of all image blocks
Obtaining the amplitude A of the difference characteristic of the whole imageT；
(2) Calculating the proximity of the source image and the fused image difference characteristic value: fusing the infrared polarization and the infrared light intensity image by adopting a fusion algorithm, and fusing the imageF, obtaining a fused image block by adopting a blocking processing mode
By the formula
And
the closeness of the fused image to the infrared polarization and infrared light intensity image in three special modes is respectively calculated,
and
representing each i-position corresponding image block Pi、IiAnd the fused image block FiAt a particular difference feature amplitude
A proximity value of where
Image block F for representing fused image at position iiA difference feature value of (c);
(3) by the difference characteristic amplitude ATDividing the amplitude into n intervals equally by taking the maximum value as a standard
(4) Discrete probability distribution construction for closeness S1, first take L different values C close to 0kK is 1,2, … L is greater than or equal to 2, and the interval R is calculatedjThe inner proximity value is in the interval [0, Ck]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CkA lower discrete probability distribution; for the proximity S2, first, S different values C close to 1 are arbitrarily takentT is 1,2, … S, S is 2 or more; calculating the interval RjThe inner proximity value is in the interval [ Ct,1]Number of scatter points q of hourjAnd the interval RjNumber of total scatter points QjProbability ratio P ofj，
Different values of structure CtDiscrete probability distribution with amplitude interval R on the horizontal axisjThe vertical axis is the probability ratio Pj；
(5) Selecting an effective fusion threshold value: from different values of CkAnd CtSelecting numerical values with the same distribution trend from the lower discrete probability distribution;
(6) determination of effective fusion interval: c obtained by carrying out weighted average on the selected numerical values with the same distribution trend1And C2Thereby obtaining an effective fusion interval [0, C1]And [ C2,1]；
(7) Calculating fusion validity of the difference characteristic amplitude interval: by using
Is calculated at RiWithin, proximity is located at [0, C1]And [ C2,1]Discrete number of points and RjInner total scatter QjRatio P ofjAnd with PjAs fusion validity of the fusion algorithm in different difference characteristic amplitude intervals, performing weighted average processing on the fusion validity under two proximity degrees to obtain final fusion validity;
(8) and selecting a fusion algorithm which keeps the highest fusion validity value in different amplitude intervals for a plurality of times to fuse the bimodal infrared images according to the final fusion validity of different fusion algorithms in the three different characteristic amplitude intervals.
4. The method of claim 1 or 3, wherein L different values C are close to 0kAnd selecting from small to large in sequence.
5. The method of claim 2 or 3, wherein the S different values C close to 1 are selected from the group consisting oftAnd selecting from small to large in sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010241703.9A CN111445430B (en) | 2020-03-31 | 2020-03-31 | Bimodal infrared image fusion algorithm selection method based on difference characteristic amplitude interval fusion validity distribution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010241703.9A CN111445430B (en) | 2020-03-31 | 2020-03-31 | Bimodal infrared image fusion algorithm selection method based on difference characteristic amplitude interval fusion validity distribution |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111445430A true CN111445430A (en) | 2020-07-24 |
| CN111445430B CN111445430B (en) | 2022-03-29 |
Family
ID=71650902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010241703.9A Expired - Fee Related CN111445430B (en) | 2020-03-31 | 2020-03-31 | Bimodal infrared image fusion algorithm selection method based on difference characteristic amplitude interval fusion validity distribution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111445430B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116503704A (en) * | 2023-06-27 | 2023-07-28 | 长春理工大学 | Target polarization detection system under strong background and detection method thereof |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5909251A (en) * | 1997-04-10 | 1999-06-01 | Cognitech, Inc. | Image frame fusion by velocity estimation using region merging |
| CN106846288A (en) * | 2017-01-17 | 2017-06-13 | 中北大学 | A kind of many algorithm fusion methods of bimodal infrared image difference characteristic Index |
| CN107909112A (en) * | 2017-11-27 | 2018-04-13 | 中北大学 | The fusion method that a kind of infrared light intensity is combined with polarization image multiclass argument |
| CN110223264A (en) * | 2019-04-26 | 2019-09-10 | 中北大学 | Image difference characteristic attribute fusion availability distributed structure and synthetic method based on intuition possibility collection |
-
2020
- 2020-03-31 CN CN202010241703.9A patent/CN111445430B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5909251A (en) * | 1997-04-10 | 1999-06-01 | Cognitech, Inc. | Image frame fusion by velocity estimation using region merging |
| CN106846288A (en) * | 2017-01-17 | 2017-06-13 | 中北大学 | A kind of many algorithm fusion methods of bimodal infrared image difference characteristic Index |
| CN107909112A (en) * | 2017-11-27 | 2018-04-13 | 中北大学 | The fusion method that a kind of infrared light intensity is combined with polarization image multiclass argument |
| CN110223264A (en) * | 2019-04-26 | 2019-09-10 | 中北大学 | Image difference characteristic attribute fusion availability distributed structure and synthetic method based on intuition possibility collection |
Non-Patent Citations (4)
| Title |
|---|
| LEIZHANG,LINNAJI: ""A categorization method of infrared polarization and intensity image fusion algorithm based on the transfer ability of difference features"", 《INFRARED PHYSICS & TECHNOLOGY》 * |
| 张雅玲,吉琳娜: ""基于余弦相似性的双模态红外图像融合性能表征"", 《光电工程》 * |
| 张雷: ""差异特征指数测度的红外偏振与光强图像多算法融合"", 《火力与指挥控制》 * |
| 郭喆: ""双模态红外图像融合有效度分布的合成研究"", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116503704A (en) * | 2023-06-27 | 2023-07-28 | 长春理工大学 | Target polarization detection system under strong background and detection method thereof |
| CN116503704B (en) * | 2023-06-27 | 2023-09-05 | 长春理工大学 | Target polarization detection system under strong background and detection method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111445430B (en) | 2022-03-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Dabov et al. | Image denoising by sparse 3-D transform-domain collaborative filtering | |
| WO2017197618A1 (en) | Method and system for removing stripe noise in infrared image | |
| Herranz et al. | Scale-adaptive filters for the detection/separation of compact sources | |
| US7940989B2 (en) | Apparatus and method for a generalized benford's law analysis of DCT and JPEG coefficients | |
| CN107590785A (en) | A kind of Brillouin spectrum image-recognizing method based on sobel operators | |
| CN103700109B (en) | SAR image change detection based on multiple-objection optimization MOEA/D and fuzzy clustering | |
| CN115825410B (en) | Method for estimating action noise level of tire road surface based on road surface structure | |
| Morard et al. | Parsimonious path openings and closings | |
| CN111445430B (en) | Bimodal infrared image fusion algorithm selection method based on difference characteristic amplitude interval fusion validity distribution | |
| Ferris et al. | Using ROC curves and AUC to evaluate performance of no-reference image fusion metrics | |
| Grigorescu et al. | Texture analysis using Renyi's generalized entropies | |
| Xu et al. | An enhanced infrared and visible image fusion method based on wavelet transform | |
| Ruchay et al. | Removal of impulse noise clusters from color images with local order statistics | |
| Navas et al. | Image fidelity metrics: future directions | |
| Wang et al. | Combined opportunity cost and image classification for non-linear image enhancement | |
| Liu et al. | Enhancing sensor pattern noise for source camera identification: An empirical evaluation | |
| CN110223264B (en) | Image difference characteristic attribute fusion validity distribution structure based on intuition possibility set and synthesis method | |
| EP2856429B1 (en) | Methods and apparatus for image processing, and laser scanning ophthalmoscope having an image processing apparatus | |
| Jeong et al. | Blind identification of image manipulation type using mixed statistical moments | |
| Qu et al. | A framework for identifying shifted double JPEG compression artifacts with application to non-intrusive digital image forensics | |
| Taher et al. | Image fusion approach with noise reduction using Genetic Algorithm | |
| Peng | Automatic Denoising and Unmixing in Hyperspectral image processing | |
| Yoo et al. | Optimizing the maximum reported cluster size for normal-based spatial scan statistics | |
| Bergstrom et al. | Image quality and object detection performance of convolutional neural networks | |
| Varadarajan et al. | A no-reference texture regularity metric based on visual saliency |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220329 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |