In this paper we focus on using single frame videos from a moving camera with pure horizontal translation. We make use of the fact that "3D shape reconstruction in Euclidean space is not necessarily required, but information of dense matching points is basically enough to synthesize new viewpoint images". The scene geometry and camera motion can be inferred by factorization of feature coordinates over a series of frames. We consider zero convergence angle and unit translation for parameterization of Fundamental Matrix for pure translation as the basis for predicting the pairs. We generated a cost function that selects the best matching stereo from the given set of frames using Fundamental Matrix Estimation (FME). The predicted scenes are compared with existing methods on the basis of their Peak Signal to Noise Ratio and graphically displayed. The generated frames are also compared using a disparity map and the results are explained in the paper.
This document provides an overview of the 2017 Stereoscopic Displays and Applications conference (the 28th in the series) and an introduction to the conference proceedings.