Color image segmentation based on an iterative graph cut algorithm using time-of-flight cameras

Abstract

This work describes an approach to color image segmentation by supporting an iterative graph cut segmentation algorithm with depth data collected by time-of-flight (TOF) cameras. The graph cut algorithm uses an energy minimization approach to segment an image, taking account of both color and contrast information. The foreground and background color distributions of the images subject to segmentation are represented by Gaussian mixture models, which are optimized iteratively by parameter learning. These models are initialized by a preliminary segmentation created from depth data, automating the model initialization step, which otherwise relies on user input. © 2011 Springer-Verlag.