Removing foreground objects by using depth information from multi-view images

Abstract

In this thesis, we present a novel method for removing foreground objects in multi-view images. Unlike the conventional methods, which locate the foreground objects interactive way, we intend to develop an automated system. To design such a system, we use the depth information obtained from multi-view images. The proposed algorithm consists of two modules: 1) foreground object detection and removal, and 2) filling the detected foreground referring to background information from other images. The depth information of multi-view images is a critical cue adopted in this algorithm. By multi-view images, we do not mean images captured by a multi-camera equipped system employed in 3D broadcasting system. We use only one digital camera and take photos by hand not by tripod. Although it may cause a little incorrect matching result, it is shown that the resultant coarse depth information is sufficient to detect and remove the foreground objects. Foreground removed image is restored with the pixel information of multi-view images. The experimental results indicate that the proposed algorithm provides an effective tool, which can be used in various applications, such as digital camera, photo-realistic scene generation, digital cinema and so on.