Image Shape Features

Abstract

Shape description or representation is an important issue in image analysis for object recognition and classification. The descriptions are given in terms of properties of objects contained in images and in terms of relationships between such objects. These properties correspond to characteristics of objects’ position, size and shape. Each shape or image to be stored in the database is processed to obtain the shape features. Shape features are then used by the different shape representation techniques (see Chapters 4 and 7) for organizing the useful shape information in index structures for efficient retrieval. For example, boundaries (connected edges) capture the characteristics of the shape object. Therefore, shapes can be processed to obtain their shape boundaries. Then, the shape boundaries are automatically decomposed into a set of boundary points (interest points) that are commonly used in machine vision techniques for shape matching. The set of shape features is by no means unique. A given set of features can give acceptable results in retrievals for a specific set of applications. However, they may fail to give acceptable results for other set of applications. Therefore, any shape representation technique should extract the shape features that experts may deem appropriate for the application domain.