Representing Shapes for Recognition
David G. Marr · Vision · 2010
This chapter is concerned with the transformation of shapes from a representation for recognition, and specifically highlights what creating a shape representation suitable for recognition entails. It discusses what recognition is and how it comes about, addressing the requirements for a representation for shape recognition in terms of the criteria of accessibility, scope and uniqueness, and stability and sensitivity. The chapter shows that a modular organization is important for a shape description, and then analyzes the interaction between the process that derives a three-dimensional model description and the recognition process. It is suggested that a shape representation for recognition should use an object-centered coordinate system, include volumetric primitives of various sizes, and have a modular organization.