Image defogging method based on adversarial network and multi-scale dense feature fusion

[0002] After searching, Chinese Patent No. CN102930514A discloses a fast image defogging method based on the atmospheric physical scattering model. Although the invention can perform image defogging through the atmospheric physical scattering model, the defogging is unnatural and has noise; in recent years, China's urban The fog and haze in the weather is gradually increasing, and the intensity of the light emitted by the scene will be attenuated when it passes through the haze and reaches the photosensitive element of the camera. In addition, the light scattering effect of the haze on the light will greatly affect the imaging effect of the camera. Based on image analysis Advanced computer vision technology has been widely used in the fields of military defense, artificial intelligence, and automatic control. Therefore, effective image defogging technology is required as a pre-processing process for computer vision applications to reduce or eliminate the impact of haze. In the haze image imaging model adopted, the atmospheric light intensity, the transmittance of the scene, and the image to be restored are all unknown quantities, and the only known haze image is the image obtained by the camera. Therefore, the problem of dehazing a single image is mathematically It is an ill-posed problem. The key problem of image defogging is to obtain the transmittance image of the haze image, and the transmittance is a function of image depth. In order to solve this ill-posed problem, the final method of image defogging is mainly divided into a priori Knowledge-based dehazing methods and deep learning-based dehazing methods, these methods basically use general-purpose network structures, but these structures have some limitations; therefore, an image dehazing method based on confrontational networks and multi-scale dense feature fusion is invented. Fog methods become especially important;

[0003] Most of the existing image defogging methods use general-purpose network structures. These structures often only use low-level local features of relatively shallow networks. Image defogging is unnatural and noisy; therefore, we propose an image defogging method based on adversarial networks and multi-scale dense feature fusion