Article ID Journal Published Year Pages File Type
527259 Image and Vision Computing 2009 16 Pages PDF
Abstract

In this paper, we propose a methodology to automatically carry out registration of hands out of conventional X-ray images. The registration method we describe here will be referred to as “articulated registration”; the method is a landmark-based elastic registration procedure in which individual bones are affinely registered and soft tissues are elastically registered so that long skeletal structures are maintained straight while a continuous and smooth transformation is obtained all over the image. In order for the method to be fully automatic, the landmarks used for the registration are detected using a number of image processing algorithms. An optimization step for the refinement of the landmarks locations is included within the registration algorithm; the algorithm is based on an iterative procedure to maximize a local similarity measure. A final procedure to correct bone width has also been performed. We show that the articulated registration described here is robust and outperforms alternatives based on the thin-plate splines (TPS) algorithm. The algorithm for automatic landmark position finding has been tested using registered images with landmarks manually selected by an expert.

Related Topics
Physical Sciences and Engineering Computer Science Computer Vision and Pattern Recognition
Authors
, , , , , ,