Search

CN-121981931-A - Remote sensing image defogging method and system based on multi-frequency dominant feature aggregation

CN121981931ACN 121981931 ACN121981931 ACN 121981931ACN-121981931-A

Abstract

The invention discloses a remote sensing image defogging method and system based on multi-frequency dominant feature aggregation. The method comprises the steps of performing multi-frequency exposure enhancement processing on an input foggy remote sensing image to obtain a global enhancement component and a local enhancement component, performing self-adaptive frequency compensation processing to generate a low-frequency compensation enhancement image, constructing multi-scale representation and calculating low-frequency components of each scale, determining high-frequency dominant features in a multi-scale space and fusing the high-frequency dominant features with the low-frequency components to form multi-scale fusion features, and completing reconstruction of the foggy remote sensing image based on the multi-scale fusion features. The system structure comprises an image input module, a multi-frequency exposure enhancement module, a self-adaptive frequency compensation module, a multi-scale feature processing module, a dominant feature aggregation module and an image reconstruction module. According to the invention, the brightness structure and detail texture can be enhanced and the residual haze can be restrained at the same time in a complex remote sensing scene, so that a clear, natural and fogless remote sensing image with good structural consistency can be obtained.

Inventors

  • CHENG JIEREN
  • WANG CHENGCHAO
  • TANG XIANGYAN
  • ZHANG ZIHUI

Assignees

  • 海南大学

Dates

Publication Date
20260505
Application Date
20251231

Claims (10)

  1. 1. The remote sensing image defogging method based on multi-frequency dominant feature aggregation is characterized by comprising the following steps of: (1) Performing multi-frequency exposure enhancement processing on an input foggy remote sensing image, generating a multi-exposure image sequence by adopting a plurality of different exposure adjustment parameters, and respectively extracting a global enhancement component and a local enhancement component based on the multi-exposure image sequence; (2) Performing self-adaptive frequency compensation processing on a brightness channel of the foggy remote sensing image, acquiring brightness statistical information according to a preset local block division mode, performing self-adaptive enhancement on a brightness difference value based on the brightness statistical information, and simultaneously setting an enhancement amplitude limiting condition to generate a low-frequency compensation enhancement image; (3) Constructing a multi-scale representation based on the global enhancement component and the low-frequency compensation enhancement image, acquiring a corresponding low-frequency component under each scale and determining high-frequency dominant features; (4) Fusing the low-frequency component and the high-frequency dominant feature under each scale to obtain a multi-scale fusion feature; (5) And performing image reconstruction based on the multi-scale fusion features to generate a haze-free remote sensing image.
  2. 2. The method of claim 1, wherein the multi-frequency exposure enhancement process in step (1) comprises: And respectively carrying out exposure adjustment on the foggy remote sensing image by adopting a plurality of groups of different exposure adjustment parameters, and generating a multi-exposure image sequence which simultaneously covers three exposure levels of dark brightness, moderate brightness and bright brightness by changing brightness gain, gamma mapping intensity or exposure offset so as to improve the structural information expression capability of the image in different brightness ranges.
  3. 3. The method of claim 1, wherein the obtaining of the global enhancement component and the local enhancement component in step (1) comprises: And (3) respectively carrying out structure component extraction processing on the multi-exposure image sequence generated in the step (1) to obtain global enhancement components reflecting the overall brightness variation trend, and carrying out differential processing on each exposure image and the corresponding structure components to obtain local enhancement components representing local detail variation.
  4. 4. The method of claim 1, wherein the fusing of the global enhancement component and the local enhancement component in step (1) comprises: And (3) constructing exposure weights based on the exposure characteristics of each exposure image, respectively weighting the global enhancement component and the local enhancement component obtained in the step (1), and fusing the weighted global enhancement component and the weighted local enhancement component to form a multi-frequency exposure enhancement result in the step (1).
  5. 5. The method of claim 1, wherein the adaptive frequency compensation process in step (2) comprises: And (3) carrying out block division processing on the brightness channel of the foggy remote sensing image according to a preset local block division mode, carrying out self-adaptive enhancement on the brightness difference value of each local block according to the brightness average value and the brightness variation amplitude of the local block, and setting an enhancement amplitude limiting condition to avoid excessive enhancement of the brightness of a local area, thereby obtaining the low-frequency compensation enhanced image in the step (2).
  6. 6. The method of claim 1, wherein the adaptive frequency compensation process in step (2) further comprises: And (3) conducting guide filtering processing on the brightness enhancement result obtained in the step (2), and inhibiting noise generated in the brightness enhancement process in a local linear modeling mode and keeping smooth transition of a brightness structure so as to improve the visual stability of the low-frequency compensation enhanced image in the step (2).
  7. 7. The method of claim 1, wherein the multi-scale feature processing in step (3) and step (4) comprises: In the multi-scale representation constructed in the step (3), low-frequency components reflecting the whole brightness structure are respectively extracted for each scale, high-frequency dominant features capable of highlighting detail areas are determined after the local change characteristics of the global enhancement components and the low-frequency compensation enhancement images are compared, and in the step (4), fusion processing is carried out on the low-frequency components of each scale and the corresponding high-frequency dominant features to form the multi-scale fusion features in the step (4).
  8. 8. Remote sensing image defogging system based on multifrequency dominant feature aggregation, characterized by comprising: (1) The multi-frequency exposure enhancement module is used for performing multi-frequency exposure enhancement on the input foggy remote sensing image, generating a multi-exposure image sequence, and extracting a global enhancement component and a local enhancement component from the multi-exposure image sequence; (2) The self-adaptive frequency compensation module is used for carrying out self-adaptive frequency compensation on the brightness channel of the foggy remote sensing image, enhancing the brightness difference value based on local brightness statistical information and generating a low-frequency compensation enhanced image according to the enhancing amplitude limiting condition; (3) The multi-scale representation construction module is used for constructing a multi-scale representation based on the global enhancement component and the low-frequency compensation enhancement image and acquiring low-frequency components and high-frequency dominant features of each scale; (4) The multi-scale fusion module is used for fusing the low-frequency components of each scale with the corresponding high-frequency dominant features to generate multi-scale fusion features; (5) And the image reconstruction module is used for reconstructing and obtaining a haze-free remote sensing image based on the multi-scale fusion characteristics.
  9. 9. A computer readable storage medium having stored thereon a computer program for implementing the method of any of claims 1 to 7 when executed by a processor.
  10. 10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program is for implementing the method of any one of claims 1 to 7 when the computer program is executed by the processor.

Description

Remote sensing image defogging method and system based on multi-frequency dominant feature aggregation Technical Field The invention relates to the technical field of remote sensing image processing, in particular to a remote sensing image defogging method and system based on multi-frequency dominant feature aggregation. Background The remote sensing image has important value in the fields of aerospace observation, environment monitoring, agriculture and forestry investigation, emergency rescue, homeland analysis and the like, and the imaging process is generally influenced by factors such as long distance, large scale, high imaging height, complex illumination and the like. When the remote sensing image is transmitted in the atmosphere, water vapor, smoke dust, pollutants and the like in the air can scatter and attenuate light, so that haze degradation problems such as brightness reduction, insufficient contrast, fuzzy details and the like occur in the acquired image, and the usability of the remote sensing image and the accuracy of subsequent analysis tasks are obviously affected. Therefore, the development of defogging and quality enhancement research of the remote sensing image is of great significance. The existing remote sensing image defogging means mainly comprise a physical model-based method and a data driving-based method. The physical model-based method generally relies on an atmospheric scattering model to recover a haze-free image by estimating the transmittance and the atmospheric light, but color cast, artifact and inaccurate transmittance estimation are easy to occur in complex ground objects, local overexposure or shadow areas. The deep learning-based method can learn more complex degradation characteristics, but has extremely strong dependence on training data, and is difficult to adapt to imaging differences of remote sensing images across regions, seasons and sensors. In an actual remote sensing scene, high-frequency textures (such as building edges and road structures) and low-frequency brightness information (such as regional fog layers and uniform fog distribution) are simultaneously degraded, and the existing model is difficult to achieve multi-scale structure recovery and local brightness compensation. In addition, the large-scale remote sensing image processing also faces the challenges of high efficiency requirement, large equipment isomerism, high image complexity and the like, so that the traditional method still has obvious defects in the aspects of robustness, generalization and processing precision. Under the background, the main problem to be solved in the remote sensing defogging technology is how to effectively recover the high-frequency texture structure and the low-frequency brightness detail in the image when facing complex atmospheric imaging conditions and multi-category ground object scenes, and to improve the definition and the structural consistency of the image and simultaneously maintain the stability and universality of the method, so that the usability of the remote sensing image under the multi-scene and multi-time-space conditions is improved. Disclosure of Invention In order to solve the problems in the prior art, the embodiment of the invention provides a remote sensing image defogging method and system based on multi-frequency dominant feature aggregation. The technical scheme is as follows: in one aspect, a remote sensing image defogging method based on multi-frequency dominant feature aggregation is provided, which comprises the following steps: (1) Performing multi-frequency exposure enhancement processing on an input foggy remote sensing image, generating a multi-exposure image sequence by adopting a plurality of different exposure adjustment parameters, and respectively extracting a global enhancement component and a local enhancement component based on the multi-exposure image sequence; (2) Performing self-adaptive frequency compensation processing on a brightness channel of the foggy remote sensing image, acquiring brightness statistical information according to a preset local block division mode, performing self-adaptive enhancement on a brightness difference value based on the brightness statistical information, and simultaneously setting an enhancement amplitude limiting condition to generate a low-frequency compensation enhancement image; (3) Constructing a multi-scale representation based on the global enhancement component and the low-frequency compensation enhancement image, acquiring a corresponding low-frequency component under each scale and determining high-frequency dominant features; (4) Fusing the low-frequency component and the high-frequency dominant feature under each scale to obtain a multi-scale fusion feature; (5) And performing image reconstruction based on the multi-scale fusion features to generate a haze-free remote sensing image. Further, the multi-frequency exposure enhancement process in step (1) includes: And respectivel