CN112053279A - Dimension reduction method and device for hyperspectral image and storage medium - Google Patents
Dimension reduction method and device for hyperspectral image and storage medium Download PDFInfo
- Publication number
- CN112053279A CN112053279A CN202010916439.4A CN202010916439A CN112053279A CN 112053279 A CN112053279 A CN 112053279A CN 202010916439 A CN202010916439 A CN 202010916439A CN 112053279 A CN112053279 A CN 112053279A
- Authority
- CN
- China
- Prior art keywords
- image
- gaussian
- layer
- pyramid
- hyperspectral
- 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
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000009467 reduction Effects 0.000 title claims abstract description 44
- 230000015654 memory Effects 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 16
- 238000005070 sampling Methods 0.000 claims description 11
- 230000003321 amplification Effects 0.000 claims description 8
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 7
- 238000004891 communication Methods 0.000 claims description 2
- 238000000605 extraction Methods 0.000 abstract description 8
- 238000007499 fusion processing Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000002904 solvent Substances 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/06—Topological mapping of higher dimensional structures onto lower dimensional surfaces
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral 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
- G06T2207/20016—Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a dimension reduction method of a hyperspectral image, which comprises the following steps: acquiring a hyperspectral image, and constructing a Gaussian pyramid by using the hyperspectral image; constructing a Laplacian pyramid according to the information difference between the Gaussian images of the two adjacent layers; fusing the Gaussian image and the Laplacian image to construct a characteristic pyramid; and fusing the plurality of characteristic images to obtain a dimension reduction image. The Gaussian pyramid can sample and compress the hyperspectral image step by step, so that the dimensionality of the hyperspectral image is effectively reduced; the Laplacian pyramid can extract the information difference between two adjacent layers of Gaussian images, so that the loss of detail characteristic information is prevented; the feature pyramid can enhance the detail feature information of the hyperspectral image to obtain the dimension-reduced image with lower dimension and enhanced detail feature information, so that the speed and the accuracy of feature information extraction of the dimension-reduced image in practical application are improved, and the application range of the hyperspectral image is further improved.
Description
Technical Field
The invention relates to the field of image processing, in particular to a dimension reduction method and device for a hyperspectral image and a storage medium.
Background
The accurate classification of the hyperspectral images plays an important role in the fields of industrial, agricultural and aerospace application, and can be applied to many practical application fields such as precision agriculture, environmental mapping, social security, mineral exploration, biological and chemical detection and the like. However, the hyperspectral images have high spectrum dimensionality and strong statistical correlation among spectral bands, so that information redundancy and calculation complexity are high, the classification accuracy is low, and the application of the hyperspectral remote sensing images is restricted.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art. Therefore, the invention provides a dimension reduction method, a dimension reduction device and a storage medium for a hyperspectral image, which can perform dimension reduction processing on the hyperspectral image, thereby improving the speed and accuracy of extraction of the hyperspectral image characteristic information and further improving the application range of the hyperspectral image.
According to the embodiment of the first aspect of the invention, the dimension reduction method of the hyperspectral image comprises the following steps:
acquiring a hyperspectral image, and constructing a Gaussian pyramid by using the hyperspectral image; the Gaussian pyramid comprises a plurality of layers of Gaussian images;
constructing a Laplacian pyramid according to the information difference between the Gaussian images of the two adjacent layers; the Laplace pyramid contains a plurality of layers of Laplace images;
fusing the Gaussian image and the Laplacian image to construct a characteristic pyramid; the characteristic pyramid comprises a plurality of layers of characteristic images;
and fusing a plurality of characteristic images to obtain a dimension reduction image.
The dimension reduction method of the hyperspectral image, provided by the embodiment of the invention, has the following beneficial effects: the Gaussian pyramid can sample and compress the hyperspectral image step by step, so that the dimensionality of the hyperspectral image is effectively reduced; the Laplacian pyramid can extract the information difference between two adjacent layers of Gaussian images, so that the loss of detail characteristic information is prevented; the feature pyramid can effectively enhance the detail feature information of the hyperspectral image, and the dimension-reduced image with lower dimension and enhanced detail feature information can be obtained by fusing a plurality of feature images, so that the speed and the accuracy of feature information extraction of the dimension-reduced image in actual application are improved, and the application range of the hyperspectral image is further improved.
According to some embodiments of the invention, the gaussian image of layer 0 of the gaussian pyramid is a hyperspectral image.
According to some embodiments of the invention, the gaussian pyramid comprises N layers of gaussian images, and the l-th layer of gaussian image is obtained by subjecting the l-1 layer of gaussian image to gaussian low-pass filtering and interval sampling; wherein l is more than or equal to 1 and less than or equal to N, and N is a positive integer.
According to some embodiments of the invention, the gaussian image of layer i is obtained by the following formula:
wherein R islNumber of lines of said Gaussian image for layer l, ClFor the number of columns of the Gaussian image of layer l, Gl(i, j) is a pixel point of the ith row and j column of the Gaussian image on the ith layer, and omega is a Gaussian kernel; gl-1(2i + m,2j + n) is the pixel point of the 2i + m row 2j + n column of the Gaussian image of the l-1 layer.
According to some embodiments of the present invention, the constructing the laplacian pyramid according to the information difference between the gaussian images of the two adjacent layers includes:
acquiring the Gaussian image of the Nth layer, and setting the Gaussian image of the Nth layer as the Laplace image of the Nth layer; the Nth layer of the Gaussian image is the top layer of the Gaussian pyramid;
performing double interpolation amplification processing on the l +1 layer Gaussian image to obtain an l +1 layer amplified image; wherein l is more than or equal to 0 and less than N, and N is a positive integer;
performing difference operation on the Gaussian image of the l-th layer and the amplified image of the l + 1-th layer to obtain a Laplace image of the l-th layer;
and connecting all the Laplacian images in sequence to obtain a Laplacian pyramid.
According to some embodiments of the invention, the magnified image of layer l +1 is calculated by the following formula:
wherein R isl+1Number of lines, C, of said Gaussian image for layer l +1l+1The number of columns of the gaussian image at layer l +1,
the pixel points in the ith row and the j column of the amplified image on the l +1 th layer are represented by omega, which is a Gaussian kernel;
is the l +1 th layer of the Gaussian image
Line of
Interpolation points of the columns.
According to some embodiments of the present invention, the constructing a feature pyramid by fusing the gaussian image and the laplacian image comprises:
acquiring an Nth layer of Laplacian image, and setting the Nth layer of Laplacian image as an Nth layer of feature image;
acquiring the Laplace image of the l layer and the magnified image of the l +1 layer, and fusing the Laplace image and the magnified image to obtain a characteristic image of the l layer; wherein l is more than or equal to 0 and less than N;
and connecting all the characteristic images in sequence to obtain a characteristic pyramid.
A dimension reduction apparatus for hyperspectral images according to an embodiment of a second aspect of the invention comprises at least one control processor and a memory for communicative connection with the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform a method of dimensionality reduction of a hyperspectral image as defined in any of the above.
A computer-readable storage medium according to an embodiment of the third aspect of the present invention stores computer-executable instructions for causing a computer to perform a dimension reduction method for hyperspectral images as described in any of the above.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flowchart of a dimension reduction method for hyperspectral images according to an embodiment of the invention.
Detailed Description
The embodiments of the present invention will be further explained with reference to the drawings.
As shown in fig. 1, a dimension reduction method for a hyperspectral image according to an embodiment of the invention includes the following steps:
step S100: acquiring a hyperspectral image, and constructing a Gaussian pyramid by using the hyperspectral image; the Gaussian pyramid comprises a plurality of layers of Gaussian images;
step S200: constructing a Laplacian pyramid according to the information difference between the Gaussian images of the two adjacent layers; the Laplace pyramid contains a plurality of layers of Laplace images;
step S300: fusing the Gaussian image and the Laplacian image to construct a characteristic pyramid; the characteristic pyramid comprises a plurality of layers of characteristic images;
step S400: and fusing the plurality of characteristic images to obtain a dimension reduction image.
For example, as shown in fig. 1, in step S100, the gaussian pyramid is arranged in a pyramid shape with gradually decreasing resolution, and is derived from the image set of the hyperspectral image. Gaussian images are obtained by down-sampling in steps, which is not stopped until some termination condition is reached, so the higher the level, the smaller the image, and the lower the resolution. Therefore, the hyperspectral image is sampled step by step through the Gaussian pyramid, the dimensionality of the hyperspectral image can be effectively reduced, and the Gaussian image with a lower dimensionality is obtained.
In step S200, the laplacian pyramid can extract an information difference between two adjacent layers of gaussian images with different dimensions, and set the information difference as a laplacian image, so as to prevent the loss of relevant detail feature information in the hyperspectral image.
In step S300 and step S400, the detail feature information of the laplacian image is fused with the gaussian image, so that the feature image realizes the enhancement of the detail feature; through fusing the characteristic images, the detail characteristics of each layer can be fused, so that the detail characteristics of the hyperspectral image can be more effectively highlighted in the dimension-reduced image. The dimensionalities of the Gaussian image, the Laplace image and the characteristic image are lower, so that the dimensionality of the fused dimensionality-reduced image is lower, the speed and the accuracy of characteristic information extraction of the dimensionality-reduced image in actual application are improved, and the application range of the hyperspectral image is further improved.
In some embodiments of the invention, the layer 0 gaussian image of the gaussian pyramid is a hyperspectral image.
Specifically, the hyperspectral images are set to be the 0 th-layer Gaussian images, so that the Gaussian images on each layer are sampled step by step on the basis of the hyperspectral images, the characteristic information of the hyperspectral images can be effectively extracted, and the dimensionality of the hyperspectral images can be effectively reduced.
In some embodiments of the invention, the gaussian pyramid comprises N layers of gaussian images, and the l-th layer of gaussian image is obtained by subjecting the l-1 layer of gaussian image to gaussian low-pass filtering and interval sampling; wherein l is more than or equal to 1 and less than or equal to N, and N is a positive integer.
Specifically, the gaussian low-pass filtering is used for eliminating high-frequency signals and gaussian noise of the l-1 layer gaussian image, so that the noise of the gaussian image is eliminated step by step, and the effectiveness of extracting the feature information of the gaussian image is improved.
The interval sampling is that interlaced and spaced sampling is carried out on the Gaussian images of the l-1 layer, so that the Gaussian images of the l layer are obtained, the difference between the sizes of the two adjacent layers of Gaussian images is four times, and the dimensionality of the Gaussian images is effectively reduced.
For example, the pixel points of the l-1 st layer gaussian image are distributed in 4 rows by 4 columns, and the interval sampling may be to collect intersection pixels of the 1 st column, the 3 rd column, the 2 nd row and the 4 th row of the layer; intersection pixels of the modified 2 nd column, 4 th column, 2 nd row and 4 th row can also be acquired; it can be seen that the number of intersection pixels is four, and the four intersections constitute the 2 × 2 th layer gaussian image. Therefore, the specific number of rows and columns for sampling at intervals is not limited as long as the number of rows and the number of columns are set at intervals.
In some embodiments of the present invention, the gaussian image of layer i is obtained by the following formula:
wherein R islNumber of lines of said Gaussian image for layer l, ClFor the number of columns of the Gaussian image of layer l, Gl(i, j) is a pixel point of the ith row and j column of the Gaussian image on the ith layer, and omega is a Gaussian kernel; gl-1(2i + m,2j + n) is the pixel point of the 2i + m row 2j + n column of the Gaussian image of the l-1 layer.
Specifically, Gl-1The (2i + m,2j + n) can realize interval sampling processing of the l-1 layer Gaussian image, and the Gaussian kernel can realize Gaussian low-pass filtering of the l-1 layer Gaussian image. In addition, of the Gaussian nucleusThe values can be:
the value of the Gaussian kernel is obtained through multiple experiments, so that noise can be effectively removed, and the extraction efficiency of the characteristic information of the Gaussian image is improved.
In some embodiments of the present invention, constructing the laplacian pyramid according to an information difference between gaussian images of two adjacent layers includes the following steps:
step S210: acquiring an Nth-layer Gaussian image, and setting the Nth-layer Gaussian image as an Nth-layer Laplacian image; the N-th layer of Gaussian image is the top layer of the Gaussian pyramid;
step S220: performing double interpolation amplification processing on the l +1 layer Gaussian image to obtain an amplified image of the l +1 layer; wherein l is more than or equal to 0 and less than N, and N is a positive integer;
step S230: performing difference operation on the first layer Gaussian image and the (l + 1) th layer amplified image to obtain a first layer Laplace image;
step S240: and connecting all the Laplacian images in sequence to obtain the Laplacian pyramid.
Specifically, in step S210, since the nth layer gaussian image is the last layer of the feature information extraction layer, the feature information of the nth layer gaussian image is not lost; by setting the nth layer gaussian image as the nth layer laplacian image, the nth layer laplacian image can effectively save the feature information of the N layer gaussian image.
In step S220, since the number of rows and the number of columns between the i-th layer gaussian image and the i + 1-th layer gaussian image are both two times different, the i + 1-th layer gaussian image is subjected to two-time interpolation amplification processing, so that the size of the i + 1-th layer gaussian image can be consistent with that of the i-th layer gaussian image, and comparison of the characteristic information between the i-th layer gaussian image and the i + 1-th layer amplified image is facilitated.
In step S230, the i +1 th layer gaussian image is obtained by sampling the i +1 th layer gaussian image at intervals, that is, the i +1 th layer gaussian image only contains part of feature information in the i th layer gaussian image. By performing difference operation on the l +1 layer gaussian image and the l +1 layer enlarged image, the detail feature information which is not extracted from the l +1 layer gaussian image in the l layer gaussian image, that is, the l layer laplacian image is the detail feature information lost from the l +1 layer gaussian image, can be obtained.
In step S240, all laplacian images are connected in sequence, so that the detail feature information lost by each layer of corresponding gaussian image can be obtained, and the laplacian pyramid can effectively store the detail feature information lost by the hyperspectral images.
Wherein, the steps S210 and S240 can be expressed by the following formulas:
middle LP of the above formulalAs the first layer Laplace image, GlIs the first layer of the gaussian image,
for magnified images of layer l +1, GNIs the N-th layer Gaussian image.
In some embodiments of the present invention, the magnified image of layer l +1 is calculated by the following formula:
wherein R isl+1Number of lines of Gaussian image of l +1 th layer, Cl+1The number of columns of the l +1 th layer gaussian image,
pixel points in the ith row and the j column of the l + 1-layer amplified image are represented, and omega is a Gaussian kernel;
is the first +1 layer of Gaussian image
Line of
Interpolation points of the columns.
In particular, the amount of the solvent to be used,
the method can realize double interpolation amplification processing on the l + 1-th layer Gaussian image, the Gaussian kernel can realize Gaussian low-pass filtering on the l + 1-th layer Gaussian image, and the noise of the l + 1-th layer Gaussian image is reduced.
Wherein,
the following conditions are satisfied:
the above conditions mean: when it comes to
Line of
After the interpolation points of the rows are subjected to twice interpolation amplification processing, obtaining numerical values which are integers, and then taking the numerical values obtained; if the obtained value is a non-integer, the value is made 0. The condition can improve the accuracy of double interpolation amplification of each interpolation point and reduce the influence of noise.
In some embodiments of the present invention, the constructing the feature pyramid by fusing the gaussian image and the laplacian image comprises the following steps:
step S310: acquiring a Laplacian image of an Nth layer, and setting the Laplacian image of the Nth layer as an Nth layer characteristic image;
step S320: acquiring a Laplace image of a layer I and an enlarged image of a layer I +1, and fusing the Laplace image and the enlarged image to obtain a characteristic image of the layer I; wherein l is more than or equal to 0 and less than N;
step S330: and connecting all the characteristic images in sequence to obtain a characteristic pyramid.
Specifically, in step S310, since the laplacian image of the nth layer completely stores the feature information of the gaussian image of the nth layer, the laplacian image of the nth layer is set as the feature image of the nth layer, so that the laplacian image of the nth layer also effectively stores the feature information of the gaussian image of the nth layer.
In step S320, the size of the enlarged image of the l +1 th layer is the same as that of the laplacian image of the l +1 th layer, and the laplacian image of the l +1 th layer stores the information of the detail features lost by the gaussian image of the l +1 th layer, and the detail features are fused with the enlarged image, and the required detail features are enhanced in the fusion process, so that the detail features in the feature image of the l +1 th layer are more obvious.
Wherein, the steps S310 and S320 can be expressed by the following formulas:
wherein, G'lCharacteristic images of the first layer, LPlFor the first layer Laplace image, LPNFor the nth layer laplacian image,
is a magnified image of the l +1 th layer. In that
Not only is the superposition of two layers of images in the calculation process, but also the LP is includedlAnd performing enhancement processing on the layer detail characteristics.
By comparing the fusion formula of the feature image with the fusion formula of the laplacian image, it can be known that the fusion process of the feature image is similar to the inverse process of the fusion process of the laplacian image, and the difference is only that: the fusion process of the feature images also comprises enhancement processing on the required detail features, so the feature images of the ith layer can be regarded as: and (5) performing detail feature enhancement processing on the Gaussian image of the ith layer.
In step S330, since each layer of feature image includes the required detail feature information, the feature pyramid is a process of enhancing the detail feature information, compared with the gaussian pyramid.
Other components and operations of the dimension reduction method for hyperspectral images according to the embodiment of the invention are known to those skilled in the art and will not be described in detail here.
The dimension reduction method of the hyperspectral image according to the embodiment of the invention is described in detail in a specific embodiment with reference to fig. 1, and it is to be understood that the following description is only illustrative and not a specific limitation to the invention.
As shown in fig. 1, the dimension reduction method for the hyperspectral image comprises the following steps:
step S100: acquiring a hyperspectral image, and constructing a Gaussian pyramid by using the hyperspectral image; the Gaussian pyramid comprises N layers of Gaussian images; the Gaussian image of the 0 th layer of the Gaussian pyramid is a hyperspectral image; any layer of Gaussian image in the hyperspectral image is obtained by performing Gaussian low-pass filtering and interval sampling processing on the adjacent lower layer Gaussian image of the Gaussian image; wherein N is a positive integer;
step S210: acquiring an Nth-layer Gaussian image, and setting the Nth-layer Gaussian image as an Nth-layer Laplacian image;
step S220: performing double interpolation amplification processing on the l +1 layer Gaussian image to obtain an amplified image of the l +1 layer; wherein l is more than or equal to 0 and less than N;
step S230: performing difference operation on the first layer Gaussian image and the (l + 1) th layer amplified image to obtain a first layer Laplace image;
step S240: connecting all the Laplacian images in sequence to obtain a Laplacian pyramid;
step S310: acquiring a Laplacian image of an Nth layer, and setting the Laplacian image of the Nth layer as an Nth layer characteristic image;
step S320: acquiring a Laplace image of a layer I and an enlarged image of a layer I +1, and fusing the Laplace image and the enlarged image to obtain a characteristic image of the layer I; wherein l is more than or equal to 0 and less than N;
step S330: connecting all the characteristic images in sequence to obtain a characteristic pyramid;
step S400: and fusing the plurality of characteristic images to obtain the dimension reduction image.
According to the dimension reduction method of the hyperspectral image, at least the following effects can be achieved through the arrangement, the Gaussian pyramid can sample and reduce noise of the hyperspectral image step by step, and therefore the dimension and the noise of the hyperspectral image are effectively reduced; the Laplace pyramid can extract information difference between two adjacent layers of Gaussian images with different dimensionalities and set the information difference as the Laplace image, so that loss of relevant detail feature information in the hyperspectral image is prevented.
The detail feature information of the Laplace image is fused with the Gaussian image, and the required detail feature is enhanced in the fusion process, so that the detail feature in the feature image is more obvious. Through fusing the plurality of characteristic images, the detail characteristics of each layer can be fused, so that the detail characteristics of the hyperspectral image can be more effectively highlighted in the dimension-reduced image, and the extraction of the characteristic information of the dimension-reduced image is facilitated.
The dimensionalities of the Gaussian image, the Laplace image and the characteristic image are lower, so that the dimensionality of the fused dimensionality-reduced image is lower, the speed and the accuracy of characteristic information extraction of the dimensionality-reduced image in actual application are improved, and the application range of the hyperspectral image is further improved.
In addition, another embodiment of the invention also provides a dimension reduction device for hyperspectral images, which comprises at least one control processor and a memory which is in communication connection with the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform a method of dimensionality reduction of a hyperspectral image as defined in any of the above.
In this embodiment, the dimension reduction apparatus includes: one or more control processors and memory, which may be connected by a bus or otherwise.
The memory, which is a non-transitory computer readable storage medium, may be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules corresponding to the dimension reduction method in the embodiments of the present invention. The control processor executes various functional applications and data processing of the dimension reduction device by running non-transitory software programs, instructions and modules stored in the memory, namely, implements the dimension reduction method of the above method embodiment.
The memory may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to use of the dimension reduction device, and the like. Further, the memory may include high speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory located remotely from the control processor, and these remote memories may be connected to the dimension reduction device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The one or more modules are stored in the memory and, when executed by the one or more control processors, perform the dimension reduction method of the above-described method embodiments, e.g., perform the functions of the above-described dimension reduction method steps S100 to S400, steps S210 to S240, and steps S310 to S330.
Embodiments of the present invention also provide a computer-readable storage medium, which stores computer-executable instructions, which are executed by one or more control processors, for example, a control processor, and can enable the one or more control processors to execute the dimension reduction method in the above method embodiments, for example, execute the functions of the above method steps S100 to S400, steps S210 to S240, and steps S310 to S330.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, may be located in one place, or may be distributed over a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
Through the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by software plus a general hardware platform. Those skilled in the art will appreciate that all or part of the processes of the methods of the above embodiments may be implemented by hardware related to instructions of a computer program, which may be stored in a computer readable storage medium, and when executed, may include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read Only Memory (ROM), a Random Access Memory (RAM), or the like.
In the description herein, references to the description of "one embodiment," "some embodiments," or "the embodiment" or the like are intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
Claims (9)
1. A dimension reduction method for a hyperspectral image is characterized by comprising the following steps:
acquiring a hyperspectral image, and constructing a Gaussian pyramid by using the hyperspectral image; the Gaussian pyramid comprises a plurality of layers of Gaussian images;
constructing a Laplacian pyramid according to the information difference between the Gaussian images of the two adjacent layers; the Laplace pyramid contains a plurality of layers of Laplace images;
fusing the Gaussian image and the Laplacian image to construct a characteristic pyramid; the characteristic pyramid comprises a plurality of layers of characteristic images;
and fusing a plurality of characteristic images to obtain a dimension reduction image.
2. The dimension reduction method for the hyperspectral image according to claim 1 is characterized in that: and the Gaussian image on the 0 th layer of the Gaussian pyramid is a hyperspectral image.
3. The dimension reduction method for the hyperspectral image according to claim 2 is characterized in that: the Gaussian pyramid comprises N layers of Gaussian images, and the first layer of Gaussian image is obtained by performing Gaussian low-pass filtering and interval sampling processing on the first-1 layer of Gaussian image; wherein l is more than or equal to 1 and less than or equal to N, and N is a positive integer.
4. The dimension reduction method for the hyperspectral image according to claim 3 is characterized in that: the gaussian image of the ith layer is obtained by the following formula:
wherein R islNumber of lines of said Gaussian image for layer l, ClFor the number of columns of the Gaussian image of layer l, Gl(i, j) is a pixel point of the ith row and j column of the Gaussian image on the ith layer, and omega is a Gaussian kernel; gl-1(2i + m,2j + n) is the pixel point of the 2i + m row 2j + n column of the Gaussian image of the l-1 layer.
5. The dimension reduction method for the hyperspectral image according to claim 1 is characterized in that: the method for constructing the Laplacian pyramid according to the information difference between the two adjacent layers of Gaussian images comprises the following steps:
acquiring the Gaussian image of the Nth layer, and setting the Gaussian image of the Nth layer as the Laplace image of the Nth layer; the Nth layer of the Gaussian image is the top layer of the Gaussian pyramid;
performing double interpolation amplification processing on the l +1 layer Gaussian image to obtain an l +1 layer amplified image; wherein l is more than or equal to 0 and less than N, and N is a positive integer;
performing difference operation on the Gaussian image of the l-th layer and the amplified image of the l + 1-th layer to obtain a Laplace image of the l-th layer;
and connecting all the Laplacian images in sequence to obtain a Laplacian pyramid.
6. The dimension reduction method for the hyperspectral image according to claim 5 is characterized in that: the magnified image of layer l +1 is calculated by the following formula:
wherein R isl+1Number of lines, C, of said Gaussian image for layer l +1l+1The number of columns of the gaussian image at layer l +1,
enlargement for layer l +1Pixel points of ith row and j column of the image, wherein omega is a Gaussian kernel;
is the l +1 th layer of the Gaussian image
Line of
Interpolation points of the columns.
7. The dimension reduction method for the hyperspectral image according to claim 5 is characterized in that: the method for fusing the Gaussian image and the Laplacian image to construct the feature pyramid comprises the following steps:
acquiring an Nth layer of Laplacian image, and setting the Nth layer of Laplacian image as an Nth layer of feature image;
acquiring the Laplace image of the l layer and the magnified image of the l +1 layer, and fusing the Laplace image and the magnified image to obtain a characteristic image of the l layer; wherein l is more than or equal to 0 and less than N;
and connecting all the characteristic images in sequence to obtain a characteristic pyramid.
8. A dimension reduction device for hyperspectral images comprises at least one control processor and a memory which is in communication connection with the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform a method of dimensionality reduction of a hyperspectral image according to any of claims 1-7.
9. A computer-readable storage medium storing computer-executable instructions for causing a computer to perform the method for dimensionality reduction of a hyperspectral image according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010916439.4A CN112053279A (en) | 2020-09-03 | 2020-09-03 | Dimension reduction method and device for hyperspectral image and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010916439.4A CN112053279A (en) | 2020-09-03 | 2020-09-03 | Dimension reduction method and device for hyperspectral image and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112053279A true CN112053279A (en) | 2020-12-08 |
Family
ID=73607312
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010916439.4A Pending CN112053279A (en) | 2020-09-03 | 2020-09-03 | Dimension reduction method and device for hyperspectral image and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112053279A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1889125A (en) * | 2006-07-26 | 2007-01-03 | 深圳市嘉易通医疗科技有限公司 | Medical radiation image detail enhancing method |
| KR20120076122A (en) * | 2010-12-29 | 2012-07-09 | 주식회사 코어웨어 | Digital image processing apparatus and method for enhancing image contrast |
| US20160148359A1 (en) * | 2014-11-20 | 2016-05-26 | Siemens Medical Solutions Usa, Inc. | Fast Computation of a Laplacian Pyramid in a Parallel Computing Environment |
| CN109345499A (en) * | 2018-10-23 | 2019-02-15 | 太原理工大学 | A kind of infrared image integration technology |
| CN110021031A (en) * | 2019-03-29 | 2019-07-16 | 中广核贝谷科技有限公司 | A kind of radioscopic image Enhancement Method based on image pyramid |
-
2020
- 2020-09-03 CN CN202010916439.4A patent/CN112053279A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1889125A (en) * | 2006-07-26 | 2007-01-03 | 深圳市嘉易通医疗科技有限公司 | Medical radiation image detail enhancing method |
| KR20120076122A (en) * | 2010-12-29 | 2012-07-09 | 주식회사 코어웨어 | Digital image processing apparatus and method for enhancing image contrast |
| US20160148359A1 (en) * | 2014-11-20 | 2016-05-26 | Siemens Medical Solutions Usa, Inc. | Fast Computation of a Laplacian Pyramid in a Parallel Computing Environment |
| CN109345499A (en) * | 2018-10-23 | 2019-02-15 | 太原理工大学 | A kind of infrared image integration technology |
| CN110021031A (en) * | 2019-03-29 | 2019-07-16 | 中广核贝谷科技有限公司 | A kind of radioscopic image Enhancement Method based on image pyramid |
Non-Patent Citations (5)
| Title |
|---|
| 傅军栋 等: "实景图像拼接及其漫游控制技术", vol. 1, 西南交通大学出版社, pages: 108 - 113 * |
| 曾明 等: "视觉注意机制在图像增强中的应用研究", 光子学报, vol. 38, no. 5, 31 May 2009 (2009-05-31), pages 1283 - 1287 * |
| 朱伟 等: "基于高斯-拉普拉斯金字塔的DR图像增强改进算法研究", 中国医疗器械杂志, vol. 43, no. 1, pages 108 - 113 * |
| 杨致中 等: "动态规划立体匹配的状态空间分析和性能改进", 应用科技, vol. 42, no. 5, 31 October 2015 (2015-10-31), pages 19 - 23 * |
| 蒋萍 等: "一种新的移动机器人气体泄漏源视觉搜寻方法", 机器人, vol. 31, no. 5, 30 September 2009 (2009-09-30), pages 397 - 403 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108229497B (en) | Image processing method, image processing apparatus, storage medium, computer program, and electronic device | |
| Wang et al. | Understanding convolution for semantic segmentation | |
| CN111369440B (en) | Model training and image super-resolution processing method, device, terminal and storage medium | |
| US20210398252A1 (en) | Image denoising method and apparatus | |
| CN111860398B (en) | Remote sensing image target detection method and system and terminal equipment | |
| CN109784372B (en) | Target classification method based on convolutional neural network | |
| CN110415250B (en) | Overlapped chromosome segmentation method and device based on deep learning | |
| KR102188035B1 (en) | Learning method and apparatus for improved resolution of satellite images | |
| CN112348116B (en) | Target detection method and device using space context and computer equipment | |
| Ding et al. | Improved object detection algorithm for drone-captured dataset based on yolov5 | |
| CN109064402B (en) | Single image super-resolution reconstruction method based on enhanced non-local total variation model prior | |
| CN115063336B (en) | Full-color and multispectral image fusion method and device and medium thereof | |
| JP2014501980A (en) | Image parallel processing method and apparatus | |
| US20130243330A1 (en) | Method and apparatus for constructing image blur pyramid, and an image feature extracting circuit | |
| CN111709415B (en) | Target detection method, device, computer equipment and storage medium | |
| CN114626503A (en) | Model training method, target detection method, device, electronic device and medium | |
| CN115565045A (en) | Hyperspectral and multispectral image fusion method based on multi-scale space-spectral transformation | |
| CN110782398B (en) | Image processing method, generative countermeasure network system and electronic device | |
| CN114758145A (en) | Image desensitization method and device, electronic equipment and storage medium | |
| CN112529897A (en) | Image detection method and device, computer equipment and storage medium | |
| CN112053279A (en) | Dimension reduction method and device for hyperspectral image and storage medium | |
| EP3591608A1 (en) | Method and system for image reconstruction using deep dictionary learning (ddl) | |
| CN115620017A (en) | Image feature extraction method, device, equipment and storage medium | |
| CN115565034A (en) | Infrared small target detection method based on double-current enhanced network | |
| CN113077389B (en) | Infrared thermal imaging method based on information distillation structure |
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 |