Cloud removal based on dual tree complex wavelet transformation and improved atmosphere transmission

Abstract

Clouds produce obstacles which hinder satellites from collecting clear ground information. Removing clouds from satellite images becomes a promising way to recover valuable information efficiently. It reduces the demands for weather and improves the image collecting flexibility. In this paper, a cloud removal algorithm from single images is studied. Firstly, the fundamental principle of the dual tree complex wavelet transformation (DTCWT) is introduced briefly. The scheme of dividing the clouds from the scenery and the background roughly according to their frequencies after the image is decomposed by DTCWT is discussed. Secondly, a new method to estimate the atmosphere transmission coefficient is designed after analyzing the atmosphere radiation model and the dark channel priori theory. Then, a cloud removal algorithm from single images is proposed by combining DTCWT and the improved atmosphere transmission. Its implement procedures are described completed. Image processing experiments are carried out and evaluated. The results prove the proposed algorithm is satisfactory and superior to algorithms based on the weighted wavelet coefficients and the dark channel prior.