Shadow Modeling and Detection

Abstract

Shadows are the common phenomena in most outdoor scenes, which bring many problems in image processing and computer vision. In this chapter, we present two novel methods focusing on modeling and extracting shadows from a single outdoor image, i.e., tricolor attenuation model and tricolor linear model. The tricolor attenuation model (TAM) describes the attenuation relationship between shadow and its non-shadow background, and it is derived based on image formation theory. The parameters of the TAM are fixed by using the spectral power distribution (SPD) of daylight and skylight, which are estimated according to Planck’s blackbody irradiance law. Based on the TAM, a multi-step shadow detection algorithm is proposed to extract shadows. The shadow linearity model is a linear ratio model which is derived from the perspective of physical imaging. The parameters of the linearity model are also fixed by the SPD of daylight and skylight, which are estimated according to our SPD calculation method proposed in Chap. 2, and they do not depend on the reflection curve of an object and the model can be applied to different scenes. Besides, we find additional new physical properties of shadows based on the tricolor linear model and take these shadow properties as features to design an effective shadow detection method. Finally, we conduct an extensive evaluation of eight widely adopted shadow features for shadow detection. Through sets of experiments, several useful and interesting conclusions are obtained.