A predictive compression method for remote visualization of 3D models

Framing remote visualization of 3D models as image transmission upon user's demand is a new and intriguing problem with a number of advantages presently balanced by a number of open questions. Among the advantages this approach doesn't require the complete download of the 3D model, thus avoiding long download times and the uncontrolled diffusion of the 3D data. Furthermore it permits the inspection of complex scenes on handheld devices such as mobile phones and PDAs. The key issue is how to best exploit the mathematical structure of the 3D model in order to take advantage of redundancy among its views. The presented work jointly uses prediction and image based rendering methods in order to exploit this redundancy. The proposed compression scheme takes advantage of depth information and uses it together with the available views in order to predict the following ones. As always in image compression we find that also in the case of remote visualization of 3D models compression efficiency comes at a premium with respect to visual quality.