Energy Minimization for Cirrus and Cumulus Cloud Separation in Atmospheric Images

Abstract

Multi-angle Imaging Spectro-Radiometer (MISR) instrument provides the multi-angle images of aerosols and clouds. There are a multitude of challenges for accurate stereo imaging of clouds and aerosols including the high variation of radiative properties of aerosols and clouds within an image. In this work, we adapt an image model to separate two specific types of clouds frequently appearing in MISR images. Specifically, we separate cirrus and cumulus clouds in the two-dimensional MISR single-channel images. We characterize these two cloud types according to their spatial variations and optical brightness. Cirrus clouds appear smooth and optically thin, while cumulus clouds present high optical oscillations and appear brighter. We adapt the additive piecewise-smooth (APS) model for this cloud separation task. We describe the differences between our results and the results of the previous joint work of the second author on cloud separation.