Thin Cloud Removal Using Local Minimization And Image Transformation In HSI Color Space

Abstract

In the use of satellite images, clouds can be an obstacle due to their opacity property that can block the visibility of ground objects in case of thick clouds and can also be blended with the underlying details in case of thin clouds. In this research, we propose a novel method to remove thin clouds and retrieve the actual information by taking the advantage of HSI color space instead of RGB color space. The proposed method aims to estimate the cloud appearance called the scattering light, remove thin clouds from intensity channel to avoid an effect to the original color, adjust the intensity with gamma correction to recover some information accidentally removed from the previous step, and restore obscure details by using contrast limited adaptive histogram equalization to enhance the intensity result. Furthermore, since thin clouds also affect saturation channel, we propose to increase saturation reduced as the result from thin clouds by using logarithm image transformation as well. From the results, the proposed method can remove clouds that are not extremely opaque and preserve the actual information such as color and texture according to the lowest mean squared error values between the outputs and their cloud-free images in the same position taken at the closest time, and the output images and their contrast gained values when compared to those obtained from other single-image methods.