作者: Sadegh Abbasi , Farzin Mokhtarian , Josef Kittler
关键词:
摘要: In many applications, the user of an image database system points to image, and wishes retrieve similar images from database. Computer vision researchers aim capture information in feature vectors which describe shape, texture color properties image. These are indexed or compared one another during query processing find This paper is concerned with problem shape similarity retrieval databases. Curvature scale space (CSS) representation along a small number global parameters used for this purpose. The CSS consists several arch-shape contours representing inflection as it smoothed. maxima these represent shape. method then tested on 1100 marine creatures. A classified subset evaluate compare other methods. results show promising performance its superiority over Fourier descriptors moment invariants.