Radial-Distance Based Shape Descriptor for Image Retrieval

Abstract

Shape analysis is used in many application fields including emerging virtual environments or 3D model market, security applications, medical imaging and many more. A Shape descriptor (or Signature) is the simplified representation of images. These shape descriptors carry important image information to store and makes easy the comparing of different shapes. The proposed shape descriptor is based on radial-distances. The type of shape descriptor used here is contour-based shape descriptor. Distance from center of bounding box encompassing the edge image to farthest point on the edge is calculated. A circle is drawn using the distance mentioned above as radius. The ratio of Euclidean distances of an edge pixel and the radius is considered as a feature. A set of such ratios for all the edge pixels forms a shape descriptor. The descriptor is divided into segments so as to avoid global distribution. A rotational matching scheme ensures invariance to rotation. As the computation of feature set is compact, implementation of this method results in quick retrieval of images invariant to scaling, translation and rotation.