IMAGE DECOMPOSITION AND REPRESENTATION IN LARGE IMAGE DATABASE-SYSTEMS

To an increasing extent, applications demand the capability of retriev al based on image content, As a result, large image database systems n eed to be built to support effective and efficient accesses to image d ata on the basis of content. In this process, significant features mus t first be extracted from image data in their pixel format. These feat ures must then be classified and indexed to assist efficient retrieval of image content. However, the issues central to automatic extraction and indexing of image content remain largely an open problem, Tools a re not currently available with which to accurately specify image cont ent for image database uses. In this paper, we investigate effective b lock-oriented image decomposition structures to be used as the represe ntation of images in image database systems. Three types of block-orie nted image decomposition structures, namely, quad-, quin-, and nona-tr ees, are compared, In analyzing and comparing these structures, wavele t transforms are used to extract image content features, Our experimen tal analysis illustrates that nona-tree decomposition is the most effe ctive of the three decomposition structures available to facilitate ef fective content-based image retrieval. Using nona-tree structure to re present image content in an image database, various types of content-b ased queries and efficient image retrieval can be supported through no vel indexing and searching approaches. We demonstrate that the nona-tr ee structure provides a highly effective approach to supporting automa tic organization of images in large image database systems. (C) 1997 A cademic Press.