A Color Correlation and Texture-Based Shadow Elimination Method for Rainy Highway Videos

Abstract

This paper presents a novel method for elimination of cast shadow and water tailing in rainy highway videos. Based on color and texture characteristics of vehicles, background, cast shadow and water tailing region, a fusion method is proposed. First, color correlation is estimated between pixels through dot product of HSI vectors. Second, the texture difference based on gradient of current region and background region is calculated. Then, these two results are fused to both eliminate cast shadow and water tailing region. Experiments show that our method is simple and efficient, also helpful to extract accurate traffic information.