Issues involving efficient data storage and retrieval have become increasingly important in the design of image database systems. The two dimensional b-string, a kind of symbolic picture representation, has been utili...
详细信息
Issues involving efficient data storage and retrieval have become increasingly important in the design of image database systems. The two dimensional b-string, a kind of symbolic picture representation, has been utilized for both characterizing the spatial knowledge embedded in images and for describing images with either partially or completely overlapping objects, without the need to partition objects. A new more efficient data structure, called the Relative Coordinates Oriented Symbolic string (the RCOS string), is proposed in this paper. Two decision trees are constructed so as to obtain the exact spatial relationship between any two objects. The feasibility of the proposed method is confirmedby performing some empirical analyses.
Recently, Yang and Lee(13) proposed a powerful data structure, the 2d b-string, to express the spatial knowledge embedded in iconic images. The data structure can be used to describe images with partly- or completely-...
详细信息
Recently, Yang and Lee(13) proposed a powerful data structure, the 2d b-string, to express the spatial knowledge embedded in iconic images. The data structure can be used to describe images with partly- or completely- overlapping objects without the need for partitioning the objects. We propose a signature file method for similarity retrieval of symbolic images described in 2d b-strings. The signatures of images are extractedby the disjoint coding and the superimposed coding techniques. A linear hashing function is adopted to organise the signature file. The experimental results show that the method has better performance than the one proposedby Lee, Yang and Chen.(11)
暂无评论