Automatic Foreground-Background Segmentation Based on Depth from Coded Aperture

Foreground-background segmentation is an important problem in computer vision, and it has many applications. We propose a technique of Automatic foreground-background segmentation based on depth from coded aperture. This method first calculates a coarse depth map using technique of coded aperture depth extraction, then estimate the general area of foreground. At last, in order to get the foreground, we use the GrabCut algorithm to segment the image. The entire progress is fully automatic, without any manual intervention. Experiments have proved its effectiveness. And we also did some simple examples for application.