Affine Invariant CSS Shape Retrieval

With the coming of the information age, shape retrieval techniques have drawn moreand more attention as an important part of information retrieval. Curvature scale space(CSS) shape retrieval technique is developed by the scale space concept and theory. Itstands out among many shape retrieval techniques because of its simplicity, highefficiency and intuitive concept, and has been widely studied and applied in various fields.This paper firstly conducts a further research of CSS shape retrieval technique, andproposes an affine invariant CSS shape retrieval technique according to solve theinstability problem of the technique under affine transformations. The main idea of thenew technique is to introduce an affine camera model to CSS. The model can predicatethe variations of the database shapes under affine transformations, and compute thepossible results of the corresponding CSS maps, with which an extended shape databaseis built up. As a consequence, the new technique has a better affine invariant effect thanbefore. The second part of this paper is to study regularities and characters of CSS mapunder affine transformations, with which the computational efficiency of the affineinvariant CSS shape retrieval is improved.The main contributions and innovations of this paper include:1) To analyze and summarize the existing CSS shape retrieval technique and itsimproved versions.2) To introduce an affine camera model and propose affine invariant CSS shaperetrieval technique.3) To study the regularities and characters of CSS map under affine transformations.4) Based on the regularities obtained in3) to propose a rapid shape retrieval technique,which has a high computational efficiency.