Image Defogging Method based on Multi-Scale and Frequency Domain Features
Haobo Wang, Yinghua Zhou · 2023
Due to the existence of atmospheric particles in the air (such as dust, colloids, raindrops, etc.), the quality of images taken in outdoor scenes is usually poor, and the visual effect will decline a lot. The phenomena caused by the absorption and refraction of atmospheric light by these particles include fog, smoke, etc. Images of these scenes usually show brightness increase and contrast decrease, which are not conducive to the identification and extraction of image features and increase the difficulty of subsequent image processing. In order to solve this problem, this paper proposes an image defogging method based on a deep neural network, which employs convolutions, multi-scale feature attention mechanism and frequency domain feature extraction to extract and fuse features of different scales. The details of the image are further restored by using the details extracted from the high-frequency components of the image. In order to verify the performance of the proposed method, experiments are conducted on RESIDE datasets. The experiments show that the model is superior to other models in terms of peak signal to noise ratio, structural similarity and subjective performance.