Hybrid Gan and Spectral Angular Distance for Cloud Removal

This paper aims to present a new algorithm to remove thin clouds and retain information in corrupted images without the use of auxiliary data. By injecting physical properties into the cycle consistent generative adversarial network (GAN), we were able to convert a cloudy multispectral image to a cloudless image. To recover information beneath clouds and shadows we create a synthetic multispectral space to obtain illumination invariant features. Multispectral vectors were transformed from Cartesian coordinates to Polar coordinates to obtain spectral angular distance (SAD) then we employed them as input to train the deep neural network (DNN). Afterward, the outputs of DNN were transformed to Cartesian coordinates to obtain shadow and cloud-free multispectral images. The proposed method, Hybrid GAN-SAD yields trustworthy reconstructed results because of exploiting transparent information from certain multispectral bands to recover uncorrupted images.

1. Introduction

Remote Sensing (RS) imagery data provides an excellent opportunity in Earth Observation (EO) to analyze and obtain information to understand the Earth's resources and physical phenomena parameters. However, clouds cover more than half of the Earth, according to statistics. This is a common issue with optical RS images causing information to be obscured by clouds and their associated shadows. As a result, we cannot capture reliable information from corrupted images unless we use clear sky images for the same time that they are not available. Therefore, one reasonable solution is improving the networks by leveraging trustworthy and transparent physical properties. Our proposed method is exploiting spectral angular distance (SAD) to train cycle-consistent adversarial networks with illumination invariant features that is illustrated in Figure 1.