Automatic modelling image represented objects using a statistic based approach

Abstract

This paper presents new methodologies to automatically extract significant points, from an object represented in images, useful to construct Point Distribution Models. Each model consists of a flexible shape template, describing how significant points of the object can vary, and a statistical model of the expected grey levels in regions around each model point. This information can be used to search objects in new images: Active Shape and Active Appearance Models. Both use PDMs for image analysis, to locate structures modeled in new images, or in a classifier, an estimate can be made of how likely the example in cause is a member of the class of shapes described by the model build. We present results for two objects: a hand and a face.