Abstract

This paper describes a method for stereoscopic rectification with geometric distortion minimisation, to generate suitable image pairs for 3D viewing applications. The current state of the art technique is not optimal as it lacks appropriate mathematical constraints. We present a new approach that enforces the same distortion minimisation criterion with more computational e±ciency whilst also achieving superior distortion removal. Detailed mathematical expressions have been developed that fully constrain the system to facilitate the use of faster and more accurate non-linear optimisation algorithms. Appropriate rectification transforms can then be defined at speeds suitable for real-timeimplementations