Skip to Main Content
Spatial structure of objects is an important feature of images and making use of this feature properly is very important for the user in the retrieval system. However, only a third of the existing image retrieval systems use spatial features and those features are low precision features which are variant to rotation and shift. In this paper, a polygon-based spatial feature model is proposed to describe the spatial structure of objects in image. This model is a precise description and is rotation-invariant, scale-invariant and shift-invariant. The experiments show the proposed method has good performance.