Removing snow from an image via image decomposition

Abstract

Snowfall removal from an image is a challenging problem. In this paper, we propose a snowfall removal framework via image decomposition based on Morphological component analysis. The proposed methods first decompose an image into low frequency (LF) and high frequency (HF) parts using bilateral filter. The high frequency part is then decomposing into “snow component” and “non snow component” by performing dictionary learning and sparse coding.