A remote sensing image defogging method based on edge sharpening recurrent generative adversarial network

Abstract

The present invention proposes a remote sensing image defogging method based on edge sharpening cyclic generation adversarial network. This method uses a deep generative adversarial network to process foggy remote sensing images, and can automatically and massively restore the feature information of foggy remote sensing images. Secondly, the present invention proposes a cyclic generative adversarial network model that adds an image sharpening mechanism. This model It improves the unclear texture of the pictures generated by the generative adversarial network and improves the ability to distinguish blurred images. Finally, the present invention improves the VGG16 network pre-training process for the model to calculate the perceptual consistency loss function, and uses remote sensing images to pre-train the VGG16 network. This allows remote sensing image features to be better extracted and makes the calculation of perceptual loss errors more accurate. Compared with existing remote sensing image dehazing methods, the present invention has the advantages of intelligence, batching and automation, can significantly improve the quality of remote sensing images, protect detailed information such as remote sensing image textures, and achieve good restoration effects.

Description

technical field [0001] The invention relates to the field of image processing, and more specifically, to a method for defogging remote sensing images based on deep learning. Background technique [0002] With the development of national aerospace technology, remote sensing images are used in all aspects of society. Due to natural, man-made and other reasons, fog has become a very common noise factor in remote sensing images. Fog will reduce the visibility of the atmosphere, blur the images obtained by optical equipment, especially based on the remote sensing platform, and make it impossible to obtain clear ground object information from the obtained remote sensing images. This will bring great difficulties to related applications based on remote sensing images, such as monitoring, automatic navigation, and target extraction. Researching an effective remote sensing image defogging method can make remote sensing images serve the current application more effectively. [0003...

AI Technical Summary

Problems solved by technology

Among them, the non-model-based defogging method is also called image enhancement method; this method does not need to consider the cause and model of image degradation when processing, and only needs to use conventional image enhancement methods to improve the image quality according to specific requirements. Visual effects are not based on the essence of optical imaging to achieve image defog processing; but the cloud thinning algorithm based on image enhancement needs to perform multi-scale transformation and inverse transformation, so the calculation is large and relatively complicated

The model-based defogging method, also known as the image restoration method, uses the principle of image degradation to achieve image defogging from the essence of optical imaging; however, due to the physical model it uses, it often requires the known scene depth and atmospheric conditions and other prior information, but for remote sensing images, it needs to obtain more information, so the use of image restoration methods brings a certain degree of difficulty to the restoration of remote sensing images

Images