Pose estimation of 3-D object by parametric egien space method using blurred edge images

Abstract

We propose a method to estimate pose of a three-dimensional object from two dimensional image based on a parametric eigen space method. In this method we use a Gaussian blurred edge image as the input image instead of the original image itself. There we compress the input images using K-L transformation, and show better pose estimation can be achieved than the method which use the original images as its inputs. The method has better efficiency especially in the cases of with occlusion, position shift and intensity changes of the objects.