Active Shape Models: Evaluation of a Multi-Resolution Method for Improving Image Search

We describe a multi-resolution technique for locating for variable structures in images.
\nThis is an extension of work on Active Shape Models (ASMs) - statistical models which
\niteratively deform to match image data. An ASM consists of a shape model controlling a
\nset of landmark points, together with a statistical model of the grey-levels expected around
\neach landmark. Both the shape model and the grey-level models are trained on sets of
\nlabelled example images. In order to apply a coarse-to-fine search strategy it is necessary
\nto train a set of grey-level models for each landmark, one for every level of a multi-resolution image pyramid. During image search the model is started on the coarsest resolution
\nimage. As the search progresses it moves to finer and finer resolutions until no further
\nimprovement can be made. We describe an automatic technique for deciding when to
\n:ss has converged. We demonstrate the
\nntitative experiments which show a sigi speed and quality of fit compared to previous methods.