View alignment with dynamically updated affine tracking
Fernando Alonso Vaca De La Torre, Shaogang Gong, Stephen James McKenna · 2002
The authors propose a framework for fast view alignment using adaptive affine tracking. They address the issue of modelling both shape and texture information in eigenspace for view alignment. They present an effective bootstrapping process based on colour segmentation and selective attention. They recover affine parameters with dynamic updates to the eigenspace using most recent history and perform predictions in parameter space. Experimental results are given to illustrate their approach.