Dense depth map estimation for multiple view coding

Ozkalayci, Burak
Alatan, Abdullah Aydın
In this paper the basics of a proposed method that handles the stereo and especially multiple view coding problem in a geometrical way, are explained. For this purpose, estimation of the depth maps of the multiple views, captured by fully calibrated cameras, are done. In depth map estimation problem Markov Random Field modelling is used to have a depth map in a desired smoothness and in an efficient coding fashion. The geometric structure which is acquired by the depth map estimation, is used to reconstruct the all camera views and the PSNR values of the reconstructed images are calculated for the coding aspects.