A 4 - week project in Active Shape and Appearance Models

This paper describes a four-week project in active shape and appearance modeling, including experiments in shape and appearance modeling and a brief look at the fundamental theory. Shape modeling derives a statistical shape model from a set of example objects annotated with landmark points to generate new, similar object shapes. Shape modeling can be extended by gray-level modeling, where a similar technique is used to derive a statistical gray-level model by sampling the gray levels from example images. Shape and gray-level models can be combined into an appearance model, which describes the way shape and gray levels of an object vary from image to image. We used MATLAB to design and implement an appearance model based on preannotated images. Our primary goal was to design a model that can be used to generate new, similar images. Our secondary goal was to use the model for segmentation by fitting the model to a new unknown image. A key component of this project is our research into the field of active appearance modeling, in the interest of learning more about the field and apprising ourselves of the recent work of scientists in this area. T.F. Cootes and C.J. Taylor of the University of Manchester have provided definitive work in statistical shape and appearance modeling for the past few years [3], thereby enriching the fields of computer vision and medical image analysis. We have sought to explore their methodology by creating our own basic shape and appearance modeling system that can be adapted to many different types of objects.