next up previous
Next: Alternative Representations of Image Up: Developing Flexible Models for Previous: Introduction

Finding Good Starting Points for an ASM

An Active Shape Model is a local search technique, needing a rough starting point which it seeks to improve upon to locate a target object accurately. We have developed a method of using features on the object to help find an approximate starting position. Statistical models are generated for a set of distinctive features on the object. These are run across any target image to generate candidate positions for each feature. Plausible combinations of candidates can be found using a statistical shape model, and used to start an ASM. The approach can be used for locating any object for which we have trained a model [4].



Tim Cootes
12/9/1997