Pseudo Stereoscopic Vision System Using a Piece of Human Group Picture －Depth Estimation by Face Recognition; Super pixel and K-means Segmentation－

Abstract

A pseudo stereoscopic vision system using a piece of human picture is discussed. Faces from a group photo are detected using a face recognition system, and the distance from camera lens (depth) of each person is estimated using each face size. Each person in the picture is clipped out by super pixels and k-means classification. To generate our pseudo stereoscopic vision, we create two images for right-eye and left-eye as each region of the person is moved according to its depth, and convert to an anaglyph image.