PedCut: an iterative framework for pedestrian segmentation combining shape models and multiple data cues

Person segmentation is a key computer vision problem in a number of application domains, such as image editing, surveillance and intelligent vehicles. This paper presents an iterative, EM-like framework for accurate pedestrian segmentation, combining generative shape models and multiple data cues. It is able to cope with a large variation of pedestrian appearances across cluttered backgrounds. In the E-step, shape priors are introduced in the unary terms of a Conditional Random Field (CRF) formulation, joining other data terms derived from color, texture and disparity cues. In the M-step, the resulting segmentation is used to adapt an Active Shape Model (ASM) [2]. The EM process alternates until the CRF-based segmentation does not appreciably change any more or a maximum number of iterations is reached. Fig. 1 illustrates our framework.