CN117952841B - Remote sensing image self-adaptive enhancement method based on artificial intelligence - Google Patents

Remote sensing image self-adaptive enhancement method based on artificial intelligence Download PDF

Info

Publication number
CN117952841B
CN117952841B CN202410345596.2A CN202410345596A CN117952841B CN 117952841 B CN117952841 B CN 117952841B CN 202410345596 A CN202410345596 A CN 202410345596A CN 117952841 B CN117952841 B CN 117952841B
Authority
CN
China
Prior art keywords
pixel
remote sensing
gray
sensing image
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202410345596.2A
Other languages
Chinese (zh)
Other versions
CN117952841A (en
Inventor
姚磊
朱司宏
马宁宁
薛立明
于广婷
刘同文
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Institute of Geological Surveying and Mapping
Original Assignee
Shandong Institute of Geological Surveying and Mapping
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Institute of Geological Surveying and Mapping filed Critical Shandong Institute of Geological Surveying and Mapping
Priority to CN202410345596.2A priority Critical patent/CN117952841B/en
Publication of CN117952841A publication Critical patent/CN117952841A/en
Application granted granted Critical
Publication of CN117952841B publication Critical patent/CN117952841B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/40Image enhancement or restoration using histogram techniques
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/0464Convolutional networks [CNN, ConvNet]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/048Activation functions
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10032Satellite or aerial image; Remote sensing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20004Adaptive image processing
    • G06T2207/20012Locally adaptive
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20081Training; Learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20084Artificial neural networks [ANN]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Image Processing (AREA)

Abstract

The invention relates to the technical field of image data processing, in particular to an artificial intelligence-based remote sensing image self-adaptive enhancement method, which comprises the following steps: obtaining mapping pixel points of pixel points in a frequency domain image in a gray remote sensing image, obtaining a frequency domain detail coefficient of a pixel block according to the position distribution and gray values of the mapping pixel points, and obtaining a airspace edge coefficient of the pixel block according to the number of edge pixel points in the pixel block and the gradient amplitude of the pixel points in a neighborhood range; and according to the fusion of the frequency domain detail coefficient and the airspace edge coefficient, a first pixel block and a second pixel block are obtained, the first pixel block is combined through a neural network, and the new pixel block and the second pixel block are respectively subjected to image enhancement, so that an enhanced gray remote sensing image is obtained. The invention improves the self-adaptive enhancement capability of the gray remote sensing image and effectively improves the enhancement effect of the gray remote sensing image.

Description

Remote sensing image self-adaptive enhancement method based on artificial intelligence
Technical Field
The invention relates to the technical field of image data processing, in particular to a remote sensing image self-adaptive enhancement method based on artificial intelligence.
Background
Remote sensing technology is a technology that acquires information by a location outside the earth's surface (e.g., an airplane or satellite) and is capable of collecting data from the earth's surface and the atmosphere without direct contact. Remote sensing images find application in many fields, including agriculture, forestry, and urban planning.
The remote sensing image is affected by various factors including atmospheric scattering, cloud layer shielding and the like, the quality and usability of the image are reduced, the traditional remote sensing image enhancement method improves the image quality by using histogram equalization operation to be enhanced, but when the overall contrast quality of the remote sensing image is adjusted by adopting a conventional histogram equalization method, the contrast difference of different areas in the image is not considered, the situation that the contrast is poor in the local area of the remote sensing image after histogram equalization often exists, and further, local details in the image are lost, so that the enhancement effect of the remote sensing image is not ideal.
Disclosure of Invention
The invention provides a remote sensing image self-adaptive enhancement method based on artificial intelligence, which aims to solve the existing problems.
The remote sensing image self-adaptive enhancement method based on artificial intelligence adopts the following technical scheme:
An embodiment of the invention provides a remote sensing image self-adaptation enhancement method based on artificial intelligence, which comprises the following steps:
Acquiring a frequency domain image of the gray remote sensing image;
Dividing a gray remote sensing image into a plurality of pixel blocks with the same size, acquiring mapping pixel points of pixel points in the gray remote sensing image in a frequency domain image according to the transformation relation between the gray remote sensing image and the frequency domain image, and acquiring the frequency domain detail coefficient of each pixel block according to the position distribution of all the pixel points in each pixel block in the mapping pixel points of the frequency domain image and the gray value of the mapping pixel points;
performing edge detection on the gray remote sensing image to obtain gradient amplitude values of pixel points in the gray remote sensing image and edge pixel points of the gray remote sensing image, and obtaining airspace edge coefficients of the pixel blocks according to the number of the edge pixel points in the pixel blocks and the gradient amplitude values of the pixel points in the neighborhood range of the edge pixel points; obtaining detail parameters of the pixel block according to the fusion result of the frequency domain detail coefficient and the airspace edge coefficient of the pixel block;
Combining the first pixel blocks according to the detail parameters of the first pixel blocks in the gray remote sensing image and combining the neural network to obtain a plurality of new pixel blocks, and respectively carrying out image enhancement on the new pixel blocks and the second pixel blocks to obtain the enhanced gray remote sensing image.
Further, the specific acquisition method of the frequency domain image of the gray remote sensing image comprises the following steps:
and processing the gray remote sensing image by utilizing Fourier transformation to obtain a frequency domain image of the gray remote sensing image.
Further, the method for obtaining the mapping pixel point of the pixel point in the gray remote sensing image in the frequency domain image according to the transformation relation between the gray remote sensing image and the frequency domain image comprises the following specific steps:
And marking any pixel point in the gray remote sensing image as a target pixel point, acquiring a pixel point with a Fourier transform relation of the target pixel in the gray remote sensing image in a frequency domain space, and marking the pixel point as a mapping pixel point of the target pixel point in the gray remote sensing image in the frequency domain image.
Further, the specific method for obtaining the frequency domain detail coefficient of each pixel block according to the position distribution of all the pixel points in each pixel block in the mapping pixel points of the frequency domain image and the gray value of the mapping pixel points includes the following steps:
acquiring a center point of a frequency domain image of a gray remote sensing image; acquiring a mapping point set of each pixel block of the gray remote sensing image;
the specific calculation method of the frequency domain detail coefficient of the pixel block comprises the following steps:
Wherein, First/>, representing gray scale remote sensing imageFrequency domain detail coefficients for the individual pixel blocks; /(I)First/>, representing gray scale remote sensing imageNumber of pixel points in a pixel block,/>First/>, representing gray scale remote sensing imageMapping Point set of individual Pixel blocks/>The Euclidean distance between each mapping pixel point and the central point of the frequency domain image; /(I)First/>, representing gray scale remote sensing imageMapping Point set of individual Pixel blocks/>The gray values of the pixels are mapped.
Further, the specific acquisition method of the mapping point set comprises the following steps:
And marking any pixel block in the gray remote sensing image as a target pixel block, and marking a set formed by mapping pixel points of all pixel points in the target pixel block in the frequency domain image as a mapping point set of the target pixel block.
Further, the method for performing edge detection on the gray remote sensing image to obtain the gradient amplitude of the pixel point in the gray remote sensing image and the edge pixel point of the gray remote sensing image comprises the following specific steps:
And acquiring the gradient amplitude of each pixel point in the gray remote sensing image by utilizing a Sobel operator, and performing Sobel edge detection on the gray remote sensing image to obtain the edge pixel points of the gray remote sensing image.
Further, the method for obtaining the airspace edge coefficient of the pixel block according to the number of the edge pixel points in the pixel block and the gradient amplitude of the pixel points in the neighborhood range of the edge pixel points comprises the following specific steps:
The first step in the gray remote sensing image The average gradient amplitude of all pixel points in each pixel block is recorded as the first/>Gradient parameters of individual pixel blocks/>
Gray scale remote sensing imageThe number of strong edge pixels within a pixel block and the/>The ratio of the number of pixel points in each pixel block is recorded as the space factor/>
Will beIs marked as/>Spatial edge coefficients of the pixel blocks.
Further, the method for obtaining the detail parameters of the pixel block according to the fusion result of the frequency domain detail coefficient and the spatial domain edge coefficient of the pixel block comprises the following specific steps:
The first step of the gray remote sensing image The product of the frequency domain detail coefficients and the spatial domain edge coefficients of each pixel block is recorded as the/>First value of the pixel block/>
Will beIs marked as/>Detail parameters of individual pixel blocks, wherein/>Representation/>Normalizing the function.
Further, the combining the first pixel blocks according to the detail parameters of the first pixel blocks in the gray remote sensing image and the neural network to obtain a plurality of new pixel blocks comprises the following specific steps:
combining the first pixel blocks in the gray remote sensing image by using the trained convolutional neural network to obtain a plurality of new pixel blocks;
The specific training process of the convolutional neural network is as follows: firstly, acquiring a plurality of gray remote sensing images and first pixel blocks in the gray remote sensing images, taking natural numbers as labels, marking the first pixel blocks to be combined in the gray remote sensing images as the same numbers, and taking all sets formed by the gray remote sensing images which comprise the first pixel blocks and are provided with the labels as data sets for training a convolutional neural network; and then inputting the data set into a convolutional neural network, taking the cross entropy loss function as the loss function of the convolutional neural network, outputting the number marked by each first pixel block, merging the first pixel blocks with the same number into one pixel block, and marking the pixel block obtained after merging as a new pixel block.
Further, the image enhancement is performed on the new pixel block and the second pixel block respectively to obtain an enhanced gray remote sensing image, which comprises the following specific methods:
And respectively carrying out histogram equalization processing on each new pixel block and each second pixel block to obtain an image formed by all new pixel blocks and all second pixel blocks subjected to the histogram equalization processing, and marking the image as an enhanced gray remote sensing image.
The technical scheme of the invention has the beneficial effects that: the detail parameters of the pixel blocks in the gray remote sensing image are obtained through the frequency domain detail coefficients and the airspace edge coefficients which are obtained through calculation, then the detail parameter differences of adjacent pixel blocks are combined, and then the histogram equalization processing is carried out, so that the targeted enhancement processing of the local area of the gray remote sensing image is realized, the adaptive enhancement processing of different areas is realized, the detail information in the gray remote sensing image can be reserved, the contrast is improved, the problem of detail loss caused by integral enhancement is avoided, and the enhancement effect of the gray remote sensing image is effectively improved while the adaptive enhancement capability of the gray remote sensing image is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of steps of an artificial intelligence based remote sensing image adaptive enhancement method of the present invention;
FIG. 2 is a schematic diagram of a gray scale remote sensing image provided by the present invention;
Fig. 3 is a schematic diagram of a frequency domain image of a gray remote sensing image according to the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description is given below of the remote sensing image self-adaptive enhancement method based on artificial intelligence according to the invention, which is provided by combining the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the remote sensing image self-adaptive enhancement method based on artificial intelligence provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of steps of an artificial intelligence based remote sensing image adaptive enhancement method according to an embodiment of the present invention is shown, where the method includes the following steps:
Step S001: and acquiring the gray remote sensing image and a frequency domain image of the gray remote sensing image.
It should be noted that, since there are various image information in the remote sensing image, in order to enhance the remote sensing image, there are various processing methods in the existing method, the embodiment enhances the remote sensing image based on the image information of the detail part in the remote sensing image, such as edges, textures, etc., while in order to enhance the remote sensing image according to the edge texture information of the remote sensing image, the embodiment obtains the frequency domain image of the remote sensing image by performing frequency domain conversion on the remote sensing image, so as to obtain the relevant features, and improve the subsequent enhancement effect on the remote sensing image.
Specifically, in order to implement the remote sensing image self-adaptive enhancement method based on artificial intelligence provided in this embodiment, firstly, a gray remote sensing image needs to be collected, and the specific process is as follows:
And carrying a space remote sensing camera by using an unmanned aerial vehicle, acquiring a high-resolution remote sensing image of a target area, preprocessing an original remote sensing image, and obtaining a gray remote sensing image by denoising, graying and other operations.
And obtaining a frequency domain image of the gray remote sensing image by utilizing Fourier transformation.
As shown in fig. 2 and 3, the acquired gray remote sensing image of a certain region and the acquired frequency domain image thereof are respectively.
Thus, the gray remote sensing image and the frequency domain image thereof are obtained by the method.
Step S002: according to the transformation relation between the gray remote sensing image and the frequency domain image, mapping pixel points of pixel points in the gray remote sensing image in the frequency domain image are obtained, and according to the position distribution of all pixel points in each pixel block in the mapping pixel points of the frequency domain image and the gray value of the mapping pixel points, the frequency domain detail coefficient of each pixel block is obtained.
Specifically, in step 2.1, firstly, the gray remote sensing image is divided into a plurality of gray remote sensing images with the size ofIs denoted as pixel block, where/>Is a preset first parameter.
It should be noted that, in this embodiment, the first parameter is preset empiricallyThe first parameter may be adjusted according to the actual situation, which is not specifically limited in this embodiment; in addition, in the process of dividing the pixel blocks, when the size of the pixel blocks obtained by dividing the gray remote sensing image is insufficient/>In this embodiment, the actual division result is the result of insufficient size/>Also as a block of pixels.
And then, marking any pixel point in the gray remote sensing image as a target pixel point, acquiring a pixel point, in which a Fourier transform relation exists in a frequency domain space, of a target pixel in the gray remote sensing image, marking the pixel point as a mapping pixel point, in the frequency domain image, of the target pixel point in the gray remote sensing image, acquiring coordinates and gray values of the mapping pixel point, in the frequency domain image, of each pixel point in the gray remote sensing image, marking any pixel block in the gray remote sensing image as a target pixel block, and marking a set formed by mapping pixel points, in the frequency domain image, of all the pixel points in the target pixel block as a mapping point set of the target pixel block.
Step 2.2, obtaining a center point of a frequency domain image of the gray remote sensing image, and obtaining a frequency domain detail coefficient of each pixel block in the gray remote sensing image according to a gray value of a mapping pixel point of a pixel point in a pixel block in the gray remote sensing image in the frequency domain image and an Euclidean distance between the mapping pixel point and the center point of the frequency domain image, wherein the specific calculation method comprises the following steps of:
Wherein, First/>, representing gray scale remote sensing imageFrequency domain detail coefficients for the individual pixel blocks; /(I)First/>, representing gray scale remote sensing imageNumber of pixel points in a pixel block,/>First/>, representing gray scale remote sensing imageMapping Point set of individual Pixel blocks/>The Euclidean distance between each mapping pixel point and the central point of the frequency domain image; /(I)First/>, representing gray scale remote sensing imageMapping Point set of individual Pixel blocks/>The gray values of the pixels are mapped.
The larger the Euclidean distance between a pixel point in the frequency domain image and the central point of the frequency domain image is, the larger the corresponding frequency of the pixel point is, namely, the higher the appearance frequency of the corresponding pixel point in the gray remote sensing image is, so that the image information of the edge detail part in the gray remote sensing image is more obvious and prominent; the larger the gray value of the mapping pixel point of the pixel point in the frequency domain image in the gray remote sensing image, the larger the corresponding amplitude in the frequency domain image, the larger the gray value of the corresponding pixel point in the gray remote sensing image. Thus the frequency domain factorReflecting the image characteristics of all pixel points in the pixel block in the frequency domain space, and describing the degree of edge detail information in the pixel block, and the frequency domain factorThe greater the value, the greater the degree of edge detail information contained in the pixel block, and the greater the frequency domain detail coefficient of each pixel block in the gray remote sensing image, otherwise, the lesser the degree of edge detail information contained in the pixel block, and the lesser the frequency domain detail coefficient of each pixel block in the gray remote sensing image.
In addition, since the frequency domain image obtained after the gray remote sensing image serving as the space domain image is subjected to fourier transform reflects the frequency domain information of the pixel points in the gray remote sensing image, and since the mapping relationship exists between the space domain image before and after the fourier transform and the pixel points in the frequency domain image, the embodiment obtains the frequency domain detail coefficient of the pixel block in the gray remote sensing image in the space domain according to the frequency and the gray value of the pixel points in the frequency domain image by combining the mapping relationship between the gray remote sensing image and the frequency domain image of the gray remote sensing image so as to describe the detail characteristics of the local area in the gray remote sensing image.
So far, the frequency domain detail coefficient of each pixel block in the gray remote sensing image is obtained through the method.
Step S003: performing edge detection on the gray remote sensing image to obtain gradient amplitude values of pixel points in the gray remote sensing image and an edge image of the gray remote sensing image, and obtaining airspace edge coefficients of the pixel blocks according to the number of the edge pixel points in the pixel blocks and the gradient amplitude values of the pixel points in the neighborhood range of the edge pixel points; and obtaining the detail parameters of the pixel block according to the fusion result of the frequency domain detail coefficient and the airspace edge coefficient of the pixel block.
Specifically, in step 3.1, firstly, a gradient amplitude of each pixel point in the gray remote sensing image is obtained by using a Sobel operator, and Sobel edge detection is performed on the gray remote sensing image to obtain an edge image of the gray remote sensing image.
It should be noted that, the Sobel operator and Sobel edge detection are known techniques, so this embodiment is not repeated.
Then, according to the number of strong edge pixel points in the pixel block and the gradient amplitude of the pixel points in the pixel block in the gray remote sensing image, the airspace edge coefficient of the pixel block is obtained, and as an embodiment, the specific calculation method is as follows:
Wherein, Representing the/>, in a gray scale remote sensing imageSpatial edge coefficients of the pixel blocks; /(I)Representing the/>, in a gray scale remote sensing imageThe number of strong edge pixels within a block of pixels; /(I)Representing the/>, in a gray scale remote sensing imageThe number of pixel points within a block of pixels; /(I)Representing the/>, in a gray scale remote sensing imageGradient parameters for individual pixel blocks.
The gradient parameter obtaining method is that the average gradient amplitude of all pixel points in any pixel block in the gray remote sensing image is recorded as the gradient parameter of the pixel block.
The method for acquiring the strong edge pixel point comprises the following steps: and obtaining average gradient amplitude values of all pixel points of any edge pixel point in an edge image of the gray remote sensing image within an 8 neighborhood range, marking the average gradient amplitude values as gradient amplitude coefficients of the edge pixel points, and marking the edge pixel points with gradient amplitude values larger than or equal to the gradient amplitude coefficients as strong edge pixel points.
In the method for calculating the spatial domain edge coefficient, the spatial domain factorThe larger the value is, the more the number of the strong edge pixel points in the pixel block is, namely the higher the density of the strong edge pixel points in the pixel block is, the more the edge detail information is contained in the pixel block, and further the airspace edge coefficient of the pixel block is larger; the larger gradient parameter indicates that the gradient fluctuation of pixel points in the pixel block is larger, and the information contained in the pixel block is mostly edge information generated by influence factors such as noise interference and the like, so/>The smaller the value of the detail area probability degree parameter is, the larger the value is, which indicates the/>The smaller the average gradient value of the region inside each pixel block and the more strong edges remain, the greater the possibility that the pixel block contains a large amount of detail information.
In general, the region with more detail information has more edges, so in order to protect the corresponding region from being lost after histogram equalization, the embodiment obtains local regions containing detail information with different degrees by obtaining detail parameters of pixel blocks, and performs enhancement processing respectively.
Step 3.2, firstly, obtaining detail parameters of a pixel block according to the frequency domain detail coefficients and the airspace edge coefficients of the pixel block in the gray remote sensing image, wherein the detail parameters are used as an embodiment, the specific calculation method comprises the following steps:
Wherein, Represents the/>Detail parameters of the individual pixel blocks; /(I)First/>, representing gray scale remote sensing imageFrequency domain detail coefficients for the individual pixel blocks; /(I)Representing the/>, in a gray scale remote sensing imageSpatial edge coefficients of the pixel blocks; /(I)Representation/>Normalizing the function.
It should be noted that, the detail parameter is used to describe the degree of the detail information contained in the pixel block, that is, the greater the detail parameter, the more edge detail information contained in the pixel block, and conversely, the smaller the detail parameter, the less edge detail information contained in the pixel block.
And then presetting a detail parameter threshold, marking a pixel block with the detail parameter smaller than the detail parameter threshold as a first pixel block, and marking a pixel block with the detail parameter larger than or equal to the detail parameter threshold as a second pixel block.
It should be noted that, the threshold value of the detail parameter is preset to be 0.4 according to experience, and the threshold value of the sequence parameter can be adjusted according to actual situations, which is not particularly limited in this embodiment.
So far, the detail parameters of each pixel block in the gray remote sensing image are obtained through the method.
Step S004: combining the first pixel blocks according to the detail parameters of the first pixel blocks in the gray remote sensing image and combining the neural network to obtain a plurality of new pixel blocks, and respectively carrying out image enhancement on the new pixel blocks and the second pixel blocks to obtain the enhanced gray remote sensing image.
Combining the first pixel blocks according to the detail parameters of the first pixel blocks in the gray remote sensing image and the neural network to obtain a plurality of new pixel blocks, and specifically:
and combining the first pixel blocks in the gray remote sensing image by using a trained convolutional neural network (CNN, convolutional Neural Networks) to obtain a plurality of new pixel blocks.
It should be noted that, the specific training process of the convolutional neural network is as follows: firstly, acquiring a plurality of gray remote sensing images and first pixel blocks in the gray remote sensing images, taking natural numbers as artificial labels, marking the first pixel blocks to be combined in the gray remote sensing images as the same numbers in a manual marking mode, and taking a set formed by all the gray remote sensing images which comprise the first pixel blocks and are provided with the artificial labels as a data set for training a convolutional neural network; and then inputting the data set into a convolutional neural network, taking the cross entropy loss function as the loss function of the convolutional neural network, outputting the number marked by each first pixel block, merging the first pixel blocks with the same number into one pixel block, and marking the pixel block obtained after merging as a new pixel block.
And 4.2, respectively carrying out histogram equalization processing on each new pixel block and each second pixel block to obtain an image formed by all new pixel blocks and second pixel blocks subjected to the histogram equalization processing, and marking the image as an enhanced gray remote sensing image.
It should be noted that, in this embodiment, by acquiring the detail features of the image in different regions in the remote sensing image, the remote sensing image is divided into a plurality of regions with different sizes, that is, pixel blocks, so that histogram equalization can be performed on the regions with different detail degrees of the remote sensing image, the situation that details are lost when the remote sensing image is integrally enhanced is avoided, the self-adaptive enhancement effect on the remote sensing image is improved, and the contrast in the image is improved while the details of the image are maintained.
This embodiment is completed.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the invention, but any modifications, equivalent substitutions, improvements, etc. within the principles of the present invention should be included in the scope of the present invention.

Claims (6)

1. The remote sensing image self-adaptive enhancement method based on artificial intelligence is characterized by comprising the following steps of:
Acquiring a frequency domain image of the gray remote sensing image;
Dividing a gray remote sensing image into a plurality of pixel blocks with the same size, acquiring mapping pixel points of pixel points in the gray remote sensing image in a frequency domain image according to the transformation relation between the gray remote sensing image and the frequency domain image, and acquiring the frequency domain detail coefficient of each pixel block according to the position distribution of all the pixel points in each pixel block in the mapping pixel points of the frequency domain image and the gray value of the mapping pixel points;
performing edge detection on the gray remote sensing image to obtain gradient amplitude values of pixel points in the gray remote sensing image and edge pixel points of the gray remote sensing image, and obtaining airspace edge coefficients of the pixel blocks according to the number of the edge pixel points in the pixel blocks and the gradient amplitude values of the pixel points in the neighborhood range of the edge pixel points; obtaining detail parameters of the pixel blocks according to fusion results of the frequency domain detail coefficients and the airspace edge coefficients of the pixel blocks, and dividing the pixel blocks into a first pixel block and a second pixel block according to the size of the detail parameters;
combining the first pixel blocks according to the detail parameters of the first pixel blocks in the gray remote sensing image and combining a neural network to obtain a plurality of new pixel blocks, and respectively carrying out image enhancement on the new pixel blocks and the second pixel blocks to obtain an enhanced gray remote sensing image;
The specific method for obtaining the frequency domain detail coefficient of each pixel block according to the position distribution of all the pixel points in each pixel block in the mapping pixel points of the frequency domain image and the gray values of the mapping pixel points comprises the following steps:
acquiring a center point of a frequency domain image of a gray remote sensing image; acquiring a mapping point set of each pixel block of the gray remote sensing image;
the specific calculation method of the frequency domain detail coefficient of the pixel block comprises the following steps:
Wherein, First/>, representing gray scale remote sensing imageFrequency domain detail coefficients for the individual pixel blocks; /(I)First step of representing gray remote sensing imageNumber of pixel points in a pixel block,/>First/>, representing gray scale remote sensing imageMapping Point set of individual Pixel blocks/>The Euclidean distance between each mapping pixel point and the central point of the frequency domain image; /(I)First/>, representing gray scale remote sensing imageMapping Point set of individual Pixel blocks/>Gray values of the individual mapping pixel points;
the method for obtaining the airspace edge coefficient of the pixel block according to the number of the edge pixel points in the pixel block and the gradient amplitude of the pixel points in the neighborhood range of the edge pixel points comprises the following specific steps:
The first step in the gray remote sensing image The average gradient amplitude of all pixel points in each pixel block is recorded as the first/>Gradient parameters of individual pixel blocks/>
Gray scale remote sensing imageThe number of strong edge pixels within a pixel block and the/>The ratio of the number of pixel points in each pixel block is recorded as the space factor/>
Will beIs marked as/>Spatial edge coefficients of the pixel blocks;
The method comprises the following steps of:
The first step of the gray remote sensing image The product of the frequency domain detail coefficients and the spatial domain edge coefficients of each pixel block is recorded as the/>First value of the pixel block/>
Will beIs marked as/>Detail parameters of individual pixel blocks, wherein/>Representation/>Normalizing the function;
presetting a detail parameter threshold, marking a pixel block with a detail parameter smaller than the detail parameter threshold as a first pixel block, and marking a pixel block with a detail parameter larger than or equal to the detail parameter threshold as a second pixel block;
combining the first pixel blocks according to the detail parameters of the first pixel blocks in the gray remote sensing image and a neural network to obtain a plurality of new pixel blocks, wherein the specific method comprises the following steps:
combining the first pixel blocks in the gray remote sensing image by using the trained convolutional neural network to obtain a plurality of new pixel blocks;
The specific training process of the convolutional neural network is as follows: firstly, acquiring a plurality of gray remote sensing images and first pixel blocks in the gray remote sensing images, taking natural numbers as labels, marking the first pixel blocks to be combined in the gray remote sensing images as the same numbers, and taking all sets formed by the gray remote sensing images which comprise the first pixel blocks and are provided with the labels as data sets for training a convolutional neural network; and then inputting the data set into a convolutional neural network, taking the cross entropy loss function as the loss function of the convolutional neural network, outputting the number marked by each first pixel block, merging the first pixel blocks with the same number into one pixel block, and marking the pixel block obtained after merging as a new pixel block.
2. The remote sensing image self-adaptive enhancement method based on artificial intelligence according to claim 1, wherein the specific acquisition method of the frequency domain image of the gray remote sensing image is as follows:
and processing the gray remote sensing image by utilizing Fourier transformation to obtain a frequency domain image of the gray remote sensing image.
3. The method for adaptively enhancing a remote sensing image based on artificial intelligence according to claim 1, wherein the specific method for obtaining the mapping pixel point of the pixel point in the gray remote sensing image in the frequency domain image according to the transformation relation between the gray remote sensing image and the frequency domain image comprises the following steps:
And marking any pixel point in the gray remote sensing image as a target pixel point, acquiring a pixel point with a Fourier transform relation of the target pixel in the gray remote sensing image in a frequency domain space, and marking the pixel point as a mapping pixel point of the target pixel point in the gray remote sensing image in the frequency domain image.
4. The remote sensing image self-adaptive enhancement method based on artificial intelligence according to claim 1, wherein the specific acquisition method of the mapping point set is as follows:
And marking any pixel block in the gray remote sensing image as a target pixel block, and marking a set formed by mapping pixel points of all pixel points in the target pixel block in the frequency domain image as a mapping point set of the target pixel block.
5. The method for adaptively enhancing a remote sensing image based on artificial intelligence according to claim 1, wherein the step of performing edge detection on the gray remote sensing image to obtain a gradient amplitude of a pixel point in the gray remote sensing image and an edge pixel point of the gray remote sensing image comprises the following specific steps:
And acquiring the gradient amplitude of each pixel point in the gray remote sensing image by utilizing a Sobel operator, and performing Sobel edge detection on the gray remote sensing image to obtain the edge pixel points of the gray remote sensing image.
6. The adaptive enhancement method of remote sensing image based on artificial intelligence according to claim 1, wherein the image enhancement is performed on the new pixel block and the second pixel block respectively to obtain an enhanced gray remote sensing image, and the specific method comprises the following steps:
And respectively carrying out histogram equalization processing on each new pixel block and each second pixel block to obtain an image formed by all new pixel blocks and all second pixel blocks subjected to the histogram equalization processing, and marking the image as an enhanced gray remote sensing image.
CN202410345596.2A 2024-03-26 2024-03-26 Remote sensing image self-adaptive enhancement method based on artificial intelligence Active CN117952841B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410345596.2A CN117952841B (en) 2024-03-26 2024-03-26 Remote sensing image self-adaptive enhancement method based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410345596.2A CN117952841B (en) 2024-03-26 2024-03-26 Remote sensing image self-adaptive enhancement method based on artificial intelligence

Publications (2)

Publication Number Publication Date
CN117952841A CN117952841A (en) 2024-04-30
CN117952841B true CN117952841B (en) 2024-06-04

Family

ID=90794492

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410345596.2A Active CN117952841B (en) 2024-03-26 2024-03-26 Remote sensing image self-adaptive enhancement method based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN117952841B (en)

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103606135A (en) * 2013-12-03 2014-02-26 山东中医药大学 Medical image enhancement processing method
CN104766297A (en) * 2014-10-08 2015-07-08 南京音视软件有限公司 Power video image stripe fault detection method based on combination of space-domain analysis and time-domain analysis
CN109345491A (en) * 2018-09-26 2019-02-15 中国科学院西安光学精密机械研究所 Remote sensing image enhancement method fusing gradient and gray scale information
CN109711284A (en) * 2018-12-11 2019-05-03 江苏博墨教育科技有限公司 A kind of test answer sheet system intelligent recognition analysis method
CN113344810A (en) * 2021-05-31 2021-09-03 新相微电子(上海)有限公司 Image enhancement method based on dynamic data distribution
CN114049283A (en) * 2021-11-16 2022-02-15 上海无线电设备研究所 Self-adaptive gray gradient histogram equalization remote sensing image enhancement method
CN114118144A (en) * 2021-11-11 2022-03-01 陈稷峰 Anti-interference accurate aerial remote sensing image shadow detection method
WO2022151589A1 (en) * 2021-01-18 2022-07-21 平安科技(深圳)有限公司 Image enhancement method, apparatus and device, and storage medium
CN115797798A (en) * 2023-02-10 2023-03-14 山东省地质矿产勘查开发局八〇一水文地质工程地质大队(山东省地矿工程勘察院) Ecological restoration effect evaluation method based on abandoned mine remote sensing image
CN115830459A (en) * 2023-02-14 2023-03-21 山东省国土空间生态修复中心(山东省地质灾害防治技术指导中心、山东省土地储备中心) Method for detecting damage degree of mountain forest and grass life community based on neural network
CN115984134A (en) * 2022-12-30 2023-04-18 黄河水利职业技术学院 Intelligent enhancing method for remote sensing mapping image
CN116128781A (en) * 2023-02-16 2023-05-16 周津同 Infrared image processing method and device
CN116342440A (en) * 2023-05-26 2023-06-27 山东广汇安通物联科技有限公司 Vehicle-mounted video monitoring management system based on artificial intelligence
WO2023134103A1 (en) * 2022-01-14 2023-07-20 无锡英菲感知技术有限公司 Image fusion method, device, and storage medium
CN117314801A (en) * 2023-09-27 2023-12-29 南京邮电大学 Fuzzy image optimization enhancement method based on artificial intelligence

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9858495B2 (en) * 2015-06-23 2018-01-02 Hong Kong Applied Science And Technology Research Wavelet-based image decolorization and enhancement
US12106334B2 (en) * 2021-05-07 2024-10-01 Tina Anne Sebastian Artificial intelligence-based system and method for grading collectible trading cards

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103606135A (en) * 2013-12-03 2014-02-26 山东中医药大学 Medical image enhancement processing method
CN104766297A (en) * 2014-10-08 2015-07-08 南京音视软件有限公司 Power video image stripe fault detection method based on combination of space-domain analysis and time-domain analysis
CN109345491A (en) * 2018-09-26 2019-02-15 中国科学院西安光学精密机械研究所 Remote sensing image enhancement method fusing gradient and gray scale information
CN109711284A (en) * 2018-12-11 2019-05-03 江苏博墨教育科技有限公司 A kind of test answer sheet system intelligent recognition analysis method
WO2022151589A1 (en) * 2021-01-18 2022-07-21 平安科技(深圳)有限公司 Image enhancement method, apparatus and device, and storage medium
CN113344810A (en) * 2021-05-31 2021-09-03 新相微电子(上海)有限公司 Image enhancement method based on dynamic data distribution
CN114118144A (en) * 2021-11-11 2022-03-01 陈稷峰 Anti-interference accurate aerial remote sensing image shadow detection method
CN114049283A (en) * 2021-11-16 2022-02-15 上海无线电设备研究所 Self-adaptive gray gradient histogram equalization remote sensing image enhancement method
WO2023134103A1 (en) * 2022-01-14 2023-07-20 无锡英菲感知技术有限公司 Image fusion method, device, and storage medium
CN115984134A (en) * 2022-12-30 2023-04-18 黄河水利职业技术学院 Intelligent enhancing method for remote sensing mapping image
CN115797798A (en) * 2023-02-10 2023-03-14 山东省地质矿产勘查开发局八〇一水文地质工程地质大队(山东省地矿工程勘察院) Ecological restoration effect evaluation method based on abandoned mine remote sensing image
CN115830459A (en) * 2023-02-14 2023-03-21 山东省国土空间生态修复中心(山东省地质灾害防治技术指导中心、山东省土地储备中心) Method for detecting damage degree of mountain forest and grass life community based on neural network
CN116128781A (en) * 2023-02-16 2023-05-16 周津同 Infrared image processing method and device
CN116342440A (en) * 2023-05-26 2023-06-27 山东广汇安通物联科技有限公司 Vehicle-mounted video monitoring management system based on artificial intelligence
CN117314801A (en) * 2023-09-27 2023-12-29 南京邮电大学 Fuzzy image optimization enhancement method based on artificial intelligence

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
Bio-Medical Image Enhancement Using Adaptive Multi-Resolution Technique;Lalit Mohan Satapathy;《IEEE access》;20200210;全文 *
一种改进的直方图均衡化图像增强方法;扈佃海;吕绪良;文刘强;;光电技术应用;20120615(第03期);全文 *
亮度保持和细节增强的红外图像增强方法;凡遵林;毕笃彦;马时平;何林远;;中南大学学报(自然科学版);20160626(第06期);全文 *
低质遥感图像压缩域细节特征高效增强仿真;耿艳萍;赵丽;耿春艳;;计算机仿真;20180715(第07期);全文 *
基于空频结合的图像增强的脑肿瘤分割;黄靖;《光子学报》;20120731;全文 *

Also Published As

Publication number Publication date
CN117952841A (en) 2024-04-30

Similar Documents

Publication Publication Date Title
CN115830459B (en) Mountain forest grass life community damage degree detection method based on neural network
CN108122213A (en) A kind of soft image Enhancement Method based on YCrCb
CN107341837B (en) Grid-vector data conversion and continuous scale expression method based on image pyramid
CN114049283A (en) Self-adaptive gray gradient histogram equalization remote sensing image enhancement method
Khan et al. Localization of radiance transformation for image dehazing in wavelet domain
CN103295204B (en) A kind of image self-adapting enhancement method based on non-down sampling contourlet transform
CN108389163A (en) A kind of self-adapting enhancement method based on half-light coloured image
CN107481210B (en) Infrared image enhancement method based on detail local selective mapping
CN104463814A (en) Image enhancement method based on local texture directionality
CN104268833B (en) Image interfusion method based on translation invariant shearing wave conversion
CN113077486B (en) Method and system for monitoring vegetation coverage rate in mountainous area
CN104616308A (en) Multiscale level set image segmenting method based on kernel fuzzy clustering
CN102243711A (en) Neighbor embedding-based image super-resolution reconstruction method
CN115984134A (en) Intelligent enhancing method for remote sensing mapping image
CN109712149B (en) Image segmentation method based on wavelet energy and fuzzy C-means
CN112070717A (en) Power transmission line icing thickness detection method based on image processing
CN108492268A (en) Enhancement algorithm for low-illumination image based on wavelet coefficient fusion
CN114820401A (en) Method for enhancing marine backlight infrared image by combining histogram transformation and edge information
CN111340692A (en) Infrared image dynamic range compression and contrast enhancement algorithm
CN109635809B (en) Super-pixel segmentation method for visual degradation image
CN115994870A (en) Image processing method for enhancing denoising
CN108931825A (en) A kind of remote sensing image clouds thickness detecting method based on atural object clarity
Chen et al. Attentive generative adversarial network for removing thin cloud from a single remote sensing image
CN109003247B (en) Method for removing color image mixed noise
CN110752013A (en) Data preprocessing system for cervical cancer laser image

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