Abstract
In this article, we propose a method for extracting an image region of the object gazed by each participant in meeting space. In the method, we assume that the gazed object is located on a plane surface at a certain distance from the face of the participant, with its surface normal along with the gaze direction. Since images of the same plane from different viewpoints can be transferred into the image from the same viewpoint by a homography matrix corresponding to the plane, the image region of the plane can be obtained as the common region between the images transferred by the matrix. Based on the idea above, we first estimate the position and orientation of the face of each participant in the meeting space from the images of the meeting space. Second, we calculate the homography matrix corresponding to the plane, at a certain distance from the position and with the normal along with the orientation, of the face. The image region of the gazed object is extracted as the common region between the images transferred from different viewpoints into the same viewpoint using the calculated homography matrix.
